Crystal structure of eukaryotic translation initiation factor 2B

Journal name:
Nature
Volume:
531,
Pages:
122–125
Date published:
DOI:
doi:10.1038/nature16991
Received
Accepted
Published online

Eukaryotic cells restrict protein synthesis under various stress conditions, by inhibiting the eukaryotic translation initiation factor 2B (eIF2B)1, 2. eIF2B is the guanine nucleotide exchange factor for eIF2, a heterotrimeric G protein consisting of α-, β- and γ-subunits. eIF2B exchanges GDP for GTP on the γ-subunit of eIF2 (eIF2γ), and is inhibited by stress-induced phosphorylation of eIF2α. eIF2B is a heterodecameric complex of two copies each of the α-, β-, γ-, δ- and ε-subunits3; its α-, β- and δ-subunits constitute the regulatory subcomplex4, while the γ- and ε-subunits form the catalytic subcomplex5. The three-dimensional structure of the entire eIF2B complex has not been determined. Here we present the crystal structure of Schizosaccharomyces pombe eIF2B with an unprecedented subunit arrangement, in which the α2β2δ2 hexameric regulatory subcomplex binds two γε dimeric catalytic subcomplexes on its opposite sides. A structure-based in vitro analysis by a surface-scanning site-directed photo-cross-linking method identified the eIF2α-binding and eIF2γ-binding interfaces, located far apart on the regulatory and catalytic subcomplexes, respectively. The eIF2γ-binding interface is located close to the conserved ‘NF motif’, which is important for nucleotide exchange. A structural model was constructed for the complex of eIF2B with phosphorylated eIF2α, which binds to eIF2B more strongly than the unphosphorylated form. These results indicate that the eIF2α phosphorylation generates the ‘nonproductive’ eIF2–eIF2B complex5, which prevents nucleotide exchange on eIF2γ, and thus provide a structural framework for the eIF2B-mediated mechanism of stress-induced translational control.

At a glance

Figures

  1. Overall structure of S. pombe eIF2B and mapping of human VWM mutations and eIF2γ cross-links.
    Figure 1: Overall structure of S. pombe eIF2B and mapping of human VWM mutations and eIF2γ cross-links.

    a, The crystal structure of S. pombe eIF2B (wall-eyed stereo view). The α-, β-, γ-, δ- and ε-subunits are coloured blue, cyan, orange, green and pink, respectively. The NF motifs in the ε-subunits are shown by red sticks. The visible C-terminal Cα atoms of the ε-subunits are shown by spheres, and the HEAT domains, whose electron densities were not observed in our structure, extend outwards. b, Mapping of the S. pombe eIF2B residues corresponding to VWM-causing missense mutations of human eIF2B (Supplementary Table 1) as spheres on the S. pombe eIF2B structure. Their environments are colour-coded on the spheres (green, solvent exposed; yellow, subunit interface; brown, structural core). c, Mapping of the S. pombe eIF2B residues corresponding to the solvent-exposed VWM-causing missense mutations of human eIF2B (green) and to the pBpa cross-links with K. pastoris eIF2γ, through the interaction between eIF2B and eIF2 (teal), on the surface model. The NF motif is coloured red.

  2. Photo-cross-linking between eIF2α and the central cavity of the eIF2B regulatory subunit.
    Figure 2: Photo-cross-linking between eIF2α and the central cavity of the eIF2B regulatory subunit.

    a, Mapping of the cross-linked (orange) eIF2B sites with S. pombe eIF2α on the crystal structure (surface model) (see Extended Data Fig. 5a). Asn100(2Bβ) is disordered in this structure. Inset: the same view as in Fig. 1a (surface model). b, Mapping of the eIF2α sites cross-linked with eIF2Bα and eIF2Bβ on the human eIF2α structure21 (ribbon model; see Extended Data Fig. 7a, b). The sites cross-linked with eIF2Bα, eIF2Bβ and both eIF2Bα and eIF2Bβ are represented as blue, cyan and purple spheres, respectively. The phosphorylated residue Ser51(2α) is represented as the red sphere. The KGYID sequence is coloured brown. c, Docking of the eIF2B and eIF2α structures, based on our photo-cross-linking experiments. Inset: the same view as in a.

  3. Model of the eIF2–eIF2B interactions.
    Figure 3: Model of the eIF2–eIF2B interactions.

    a, Docking of the eIF2 and eIF2B structures, by fixing eIF2α-NTD in the central cavity of eIF2B. eIF2α-NTD is placed in the same manner as in Fig. 2c. The structures of aIF2α-CTD and aIF2γ from PDB accession number 2QMU24, the archaeal homologues of the corresponding segments of eIF2, are positioned so they extend towards the NF motif (red) from the C terminus of eIF2α-NTD. b, Schematic representations of the eIF2–eIF2B interactions proposed in the present study.

  4. Biochemical properties of recombinant S. pombe eIF2B.
    Extended Data Fig. 1: Biochemical properties of recombinant S. pombe eIF2B.

    a, Guanine-nucleotide exchange catalysed by S. pombe eIF2B, and its inhibition by eIF2 phosphorylation. K. pastoris eIF2 was labelled with [3H]GDP and incubated with ATP and activated PKR (eIF2(αP); red line), or with ATP only (eIF2; blue line). Reactions were started by the addition of excess unlabelled GDP with eIF2B (solid line) or buffer (dashed line). Individual data points of triplicate analyses are shown by dots and boxes. The dissociation rates of GDP from eIF2 are summarized in the table. b, The SEC profiles of respective proteins (S. pombe eIF2B (green), K. pastoris eIF2 (cyan), eIF2(αP) (pink)) (top), eIF2B with K. pastoris eIF2 (middle) and eIF2B with eIF2(αP) (bottom). The chromatograms of the absorbance at 280 nm and the Coomassie blue stained SDS–PAGE gels are shown. For gel source data, see Supplementary Fig. 1. c, ITC measurements between S. pombe eIF2B (as the decamer) and S. pombe eIF2α (top), eIF2B and P-eIF2α (bottom). We used the eIF2α subunit alone, rather than the trimeric eIF2, in these experiments because of the difficulty in the preparation of highly concentrated eIF2. Representative thermograms are shown. Two runs were concatenated for analysis. The binding stoichiometry (N) and the dissociation constant (Kd) were calculated from triplicate analyses (mean ± s.d.).

  5. Architecture of the eIF2B subcomplexes.
    Extended Data Fig. 2: Architecture of the eIF2B subcomplexes.

    a, b, Two different views of the crystal structure of S. pombe eIF2B (wall-eyed stereo view), coloured as in Fig. 1a and from the same view as in Fig. 1b and Fig. 1c, respectively. The assembly of the subcomplexes is primarily mediated by the β–ε and δ–γ interactions. In this regard, the results of the two prior mutational analyses12, 13 are consistent with the present structure, except that one suggested that the β-helical region of the γ-subunit is involved in the inter-subunit interactions. ck, Ribbon models of the structures of S. pombe eIF2B subcomplexes (ce, g, h, j) and proteins with structural similarity (f, i, k). c, d, The α2 homodimer (c) and the βδ heterodimer (d) in the eIF2B decamer. The conformations of each subunit are similar to those in the human α2 homodimer43 and the Chaetomium thermophilum βδ heterotetramer3, even though the relative orientations of the subunits in the dimers are slightly different from those represented in these partial structures. e, f, The regulatory subcomplex in the eIF2B decamer (e) and ribose-1,5-bisphosphate (R15Pi) homohexamer44 (PDB 3A11) (f). The architecture of the regulatory subcomplex is an assembly of three similarly shaped dimer moieties: one homodimer of the α-subunit and two heterodimers of the β- and δ-subunits. The arrangement of the regulatory subunits resembles that in the C. thermophilum βδ heterotetramer3, and shares some similarity to that in the homohexameric structure of R15Pi44. g, The γ-subunit of eIF2B. h, The ε-subunit of eIF2B. i, Potato tuber ADP-glucose pyrophosphorylase (AGP) (PDB 1YP3)45. The dimerization interfaces between the catalytic subunits are coloured in deeper shades (g, h). j, k, The subunit heterodimerization mode in the catalytic subcomplex of eIF2B (j) and the subunit homodimerization mode in the potato tuber AGP (k). The dimerization manner of the γ- and ε-subunits is novel: each of their structures resembles the subunit structure of the AGP homotetramer45, but they dimerize through their N-terminal domains, in a different manner than the AGP homotetramer45.

  6. Mapping of the residues corresponding to missense VWM mutations on the subunit interfaces and the distal face of eIF2B.
    Extended Data Fig. 3: Mapping of the residues corresponding to missense VWM mutations on the subunit interfaces and the distal face of eIF2B.

    The eIF2B residues corresponding to VWM-causing missense mutations in human (Supplementary Table 1) are mapped on the S. pombe eIF2B structure, with the same subunit colouring as in Fig. 1. The S. pombe eIF2B residues corresponding to VWM-causing missense mutations are shown in parentheses. a, VWM-related residues are mapped as spheres on the overall structure (ribbon model). The environments of the residues are colour-coded (green, solvent-exposed; yellow, subunit interface; brown, structural core) on the spheres. b, VWM-related residues are mapped on the surface model of the inter-subcomplex interfaces on the catalytic subcomplex side (left), with the interfaces for the α-, β- and δ-subunits coloured blue, cyan and green, respectively, and on the βδ dimer side (right), with the interfaces for the γ- and ε-subunits coloured orange and pink, respectively. VWM-related residues around the βδ dimerization interface are shown as spheres in the inset. The mutations in the regulatory subunits are clustered around the dimerization interface between the β- and δ-subunits, as mentioned in ref. 3. Our structure further revealed that the binding site for the ε-subunit is formed by the correct interaction between the β- and δ-subunits, thus explaining the abundance of mutations around this interface. c, VWM-related residues in the α2 homodimer are located around the homodimerization interface, and shown as spheres in the inset. These VWM-related mutations around the subunit interfaces (b, c) may cause appreciable degrees of subunit dissociation from the eIF2B decamer, leading to incomplete complexes, destabilization of eIF2B resulting in aggregation/degradation, and/or changes in the conformation and activity of the intact eIF2B decamer. d, VWM-related residues on the distal face of the catalytic subcomplex are mapped on the surface model. The NF motif is shown in red. Several VWM-related residues are located near the NF motif, including Arg111(2Bε), corresponding to the human Arg136His(2Bε) mutation, for which the mouse model is available46. e, The ε-subunit further contains several exposed missense VWM mutations, especially in the β-helical domain. f, VWM-related residues in the central cavity. Only one residue, corresponding to the human Lys110Glu(2Bα) mutation, is exposed to the solvent.

  7. Photo-cross-linking between pBpa-labelled eIF2B and eIF2.
    Extended Data Fig. 4: Photo-cross-linking between pBpa-labelled eIF2B and eIF2.

    a, The S. pombe eIF2B variants bearing a single site-specific pBpa substitution in the catalytic subunits were mixed with K. pastoris eIF2, and irradiated with ultraviolet (365 nm) for 5 min on ice. Since eIF2γ harbours the Flag-tag at the C terminus, the products cross-linked with eIF2γ were detected by western blotting with an anti-Flag antibody. Site-specific slow-migrating bands that appeared after ultraviolet irradiation were judged as cross-linked bands. The relevant bands are indicated with teal dots. b, The eIF2γ-cross-linked sites are shown in teal, except for the selected ones explained below, and the cross-link-negative sites are shown in grey, on the surface model of the catalytic subcomplex, coloured in the same manner as in Fig. 1. The NF motif is shown in red. c, Time-course analysis of cross-linking with eIF2γ. Four selected sites on the distal face (Gln117(2Bε) (light green in b), Leu257(2Bε) (blue), Glu204(2Bε) (purple) and Ser258(2Bγ) (violet)) were examined by time courses of the cross-linking with eIF2 or eIF2(αP). The ratio of the band intensity of the eIF2B γ- or ε-subunit cross-linked with eIF2(αP) to that with unphosphorylated eIF2 at each time point is shown below the lane. ND means that no band of the eIF2B subunit cross-linked with eIF2(αP) was detected at the time point. For gel source data, see Supplementary Fig. 1.

  8. Photo-cross-linking between pBpa-labelled eIF2B and eIF2α.
    Extended Data Fig. 5: Photo-cross-linking between pBpa-labelled eIF2B and eIF2α.

    a, The S. pombe eIF2B variants bearing a single site-specific pBpa substitution in the regulatory subunit were mixed with S. pombe P-eIF2α, and irradiated with ultraviolet (365 nm) for 5 min. Cross-linking was detected as described in Extended Data Fig. 4. The relevant bands are indicated with orange dots. b, The eIF2α-cross-linked sites are shown in orange, and the cross-link-negative sites are shown in grey, on the surface model of the overall structure of eIF2B, with the same subunit colouring as in Fig. 1. c, The eIF2B variants were similarly mixed with unphosphorylated eIF2α and irradiated with ultraviolet. The bands that were also observed in a are indicated with orange dots, and the unphosphorylated eIF2α-specific bands are indicated with magenta dots. d, Arg84(2Bβ) and Gln91(2Bβ), which exhibited unphosphorylated eIF2α-specific cross-links, are shown in magenta. The view is the same as in Fig. 2a. For gel source data, see Supplementary Fig. 1.

  9. Solvent-exposed Gcn– mutations located in the central cavity and buried Gcn– mutations clustered around the trimerization interface.
    Extended Data Fig. 6: Solvent-exposed Gcn mutations located in the central cavity and buried Gcn mutations clustered around the trimerization interface.

    a, The residues corresponding to Gcn mutations19, 20 are mapped in blue on the surface model of the S. pombe eIF2B structure, with the same subunit colouring as in Fig. 1. b, c, The residues corresponding to exposed Gcn mutations are mainly located in the central cavity (b), and their locations coincide with the P-eIF2α cross-link sites shown in orange (c, Extended Data Fig. 5b). d, e, The residues corresponding to Gcn mutations on the α2 homodimer (d) and the βδ heterodimer (e). The residues of S. pombe eIF2B corresponding to S. cerevisiae Gcn mutations are indicated in parentheses. The interfaces for the trimerization of the regulatory dimers are coloured grey. Most of the Gcn-related residues are mapped only on one face of the subunits, as predicted43, 47. The present eIF2B structure revealed that these ‘mutation-rich’ faces are used for the assembly of the regulatory subunits to form the subcomplex, and the Gcn-related residues are clustered on the interface for the trimeric assembly.

  10. Cross-linking of pBpa-labelled eIF2α with the central cavity formed by the eIF2B regulatory subunits.
    Extended Data Fig. 7: Cross-linking of pBpa-labelled eIF2α with the central cavity formed by the eIF2B regulatory subunits.

    a, b, The S. pombe eIF2α variants bearing a single site-specific pBpa residue were mixed with epitope-tagged S. pombe eIF2B, and irradiated with ultraviolet (365 nm) for 5 min on ice. Cross-linking was detected as described in Extended Data Fig. 4. Cross-links with eIF2Bα were detected with an anti-myc antibody (a), and those with eIF2Bβ were detected with an anti-HA antibody (b). The relevant bands are indicated with blue dots in a, and cyan dots in b. These cross-linked sites are mapped on the human eIF2α structure21 in Fig. 2b. For gel source data, see Supplementary Fig. 1. c, The model of the eIF2B–eIF2α complex, built on the basis of the cross-linking experiments, is shown in Fig. 2c. The phosphorylated residue Ser51(2α) is highlighted with the magenta circle. d, The electrostatic surface potential of eIF2B, from the same viewpoint as in c. Red and blue colours represent negative and positive potentials, respectively, of ±10kT/e. The cavity has no positively charged patch; therefore, the mechanism underlying the enhanced affinity for P-eIF2α is still unclear. One possible mechanism is a cation-mediated recognition of the phosphoserine residue. The phosphorylated Ser51(2α) may coordinate a cation together with the negatively charged residues at the bottom of the central cavity, although we did not observe any electron density for such cations in the cavity. Another possibility is a phosphorylation-induced conformational change of the Ser51-flanking loop. The phosphorylation of Ser51(2α) may induce the rearrangement of adjacent arginine residues, and enable a stronger interaction with the negatively charged residues at the bottom of the central cavity.

  11. Analyses of eIF2B Gcn– mutations.
    Extended Data Fig. 8: Analyses of eIF2B Gcn mutations.

    a, The locations of the Gcn-related residues mutated for the SEC and ITC analyses (Glu57(2Bα) and Asp248(2Bδ)) are indicated on the S. pombe eIF2B structure, with the same view and colouring as in Extended Data Fig. 6b. b, The SEC analyses of the interaction between K. pastoris eIF2(αP) and S. pombe eIF2B, bearing Gcn-related mutations. The chromatograms of the absorbance at 280 nm and the SDS–PAGE gels of each run are shown. The green bar represents the elution point of free eIF2B (Extended Data Fig. 1b). For gel source data, see Supplementary Fig. 1. c, ITC measurements between S. pombe eIF2α or P-eIF2α and eIF2B, bearing Gcn-related mutations. Representative thermograms for the ITC experiments are shown. d, e, The nucleotide exchange activities of S. pombe eIF2B, bearing Gcn-related mutations, on K. pastoris eIF2(αP) (d) and eIF2 (e) were examined as described in Extended Data Fig. 1a (αGlu57Lys mutant, blue line; δAsp248Lys mutant, green line; wild type, grey line). Individual data points of triplicate analyses are shown by dots.

  12. The locations of ISRIB-resistant mutations on the eIF2B structure.
    Extended Data Fig. 9: The locations of ISRIB-resistant mutations on the eIF2B structure.

    Residues corresponding to the ISRIB-resistant mutations26 are mapped onto the eIF2B structure in red. The residue corresponding to the Arg171Gln(2Bδ) mutation (Lys112(2Bδ)) is in the disordered region, at the N terminus of the δ-subunit. The disordered N-terminal segment of the δ-subunit is indicated by the dotted green line.

Tables

  1. Data collection and refinement statistics
    Extended Data Table 1: Data collection and refinement statistics

Main

In eukaryotic translation initiation, eIF2 in the GTP-bound form delivers an initiator methionyl-tRNA (Met-tRNAiMet) to the ribosome, and then dissociates in the GDP-bound form2. For the next round of translation initiation, eIF2B catalyses the exchange of the eIF2γ-bound GDP for GTP. This guanine nucleotide exchange activity requires the HEAT domain at the carboxy (C) terminus and the NF motif in the amino (N)-terminal region of the ε-subunit, as well as the formation of the catalytic γε subcomplex5, 6, 7. The NF motif consists of consecutive Asn–Phe residues conserved in eIF2Bε (Asn237 and Phe238 in S. pombe eIF2Bε), and their mutations reduce the nucleotide exchange activity drastically, down to the level of a HEAT-domain fragment7. Under stressful conditions, the phosphorylation of eIF2α at Ser51 induces its stronger binding to the regulatory αβδ subcomplex of eIF2B, which results in the inhibition of eIF2B4, 8, 9. This inhibition limits the supply of Met-tRNAiMet to the ribosome, leading to global translational repression and de-repression of the translation of stress-induced mRNAs1, 2, 8. However, the mechanisms of the catalysis and the inhibition and their mutual relationship have remained elusive, and information about the overall structure of eIF2B has long been awaited. Importantly, a variety of mutations of the human eIF2B subunits are related to the neurodegenerative disease leukoencephalopathy with vanishing white matter (VWM) or childhood ataxia with central nervous system hypomyelination (CACH)10. In patients with this disease, white matter lesions and neurological disorders severely deteriorate during recovery after exposure to stresses, and the eIF2B activities are generally lower than normal11. Thus, the structure of eIF2B would also provide information underlying the pathogenesis of VWM/CACH.

We prepared S. pombe eIF2B, by co-expressing all subunits in Escherichia coli (Kashiwagi et al., submitted). We examined the nucleotide exchange activity against Komagataella pastoris (Pichia pastoris) eIF2, which shares high sequence identity with S. pombe eIF2 (α: 66%; β: 47%; γ: 76%). This recombinant eIF2B molecule exhibited the nucleotide exchange activity on K. pastoris eIF2, and was inhibited by the phosphorylated eIF2 protein (eIF2(αP)) (Extended Data Fig. 1a). The eIF2B molecule showed stronger binding to both the trimeric eIF2(αP) complex and the phosphorylated eIF2α subunit (P-eIF2α) (Extended Data Fig. 1b, c). Therefore, the recombinant eIF2B molecule displayed the characteristic biochemical properties of the natural eIF2B.

We determined the crystal structure of eIF2B at 3.0-Å resolution, and assigned almost all regions except for the ε-subunit HEAT domains (Fig. 1a, Extended Data Fig. 2a, b and Extended Data Table 1). The crystal structure revealed an unprecedented arrangement of the subunits: the hexameric regulatory subcomplex resides at the centre, with the two heterodimeric catalytic subcomplexes bound on opposite sides. The assembly of the subcomplexes is primarily mediated by the β–ε and δ–γ interactions (Fig. 1a and Extended Data Fig. 2a, b). This decameric structure is consistent with many of the numerous previously reported results of subunit interactions12, 13, 14, 15, 16 (Extended Data Fig. 2). Mapping of the residues corresponding to the missense mutations causing VWM disease (Fig. 1b and Supplementary Table 1) revealed that many mutations are located within or around various subunit interfaces (yellow in Fig. 1b and Extended Data Fig. 3a–c). These interface mutations may cause substantial effects on the structural and biochemical properties of eIF2B.

Figure 1: Overall structure of S. pombe eIF2B and mapping of human VWM mutations and eIF2γ cross-links.
Overall structure of S. pombe eIF2B and mapping of human VWM mutations and eIF2γ cross-links.

a, The crystal structure of S. pombe eIF2B (wall-eyed stereo view). The α-, β-, γ-, δ- and ε-subunits are coloured blue, cyan, orange, green and pink, respectively. The NF motifs in the ε-subunits are shown by red sticks. The visible C-terminal Cα atoms of the ε-subunits are shown by spheres, and the HEAT domains, whose electron densities were not observed in our structure, extend outwards. b, Mapping of the S. pombe eIF2B residues corresponding to VWM-causing missense mutations of human eIF2B (Supplementary Table 1) as spheres on the S. pombe eIF2B structure. Their environments are colour-coded on the spheres (green, solvent exposed; yellow, subunit interface; brown, structural core). c, Mapping of the S. pombe eIF2B residues corresponding to the solvent-exposed VWM-causing missense mutations of human eIF2B (green) and to the pBpa cross-links with K. pastoris eIF2γ, through the interaction between eIF2B and eIF2 (teal), on the surface model. The NF motif is coloured red.

Two key motifs for the nucleotide exchange activity, the HEAT domain and the NF motif of the ε-subunit, both reside in the ‘distal’ region of the structure, and several missense VWM mutations are mapped near the NF motif (Fig. 1c and Extended Data Fig. 3d). Therefore, nucleotide exchange on eIF2γ occurs on the distal face of eIF2B. To examine how eIF2γ binds to this face, we performed surface-scanning photo-cross-linking experiments (see Methods). Thirty-two variants of eIF2B, labelled with p-benzoyl-L-phenylalanine (pBpa), were ultraviolet-irradiated in the presence of K. pastoris eIF2 (Extended Data Fig. 4a, b). Photo-cross-linking was detected between eIF2γ and the distal face of eIF2B, when pBpa was incorporated at one of the ten sites distributed over a large area, extending from eIF2Bε to eIF2Bγ (Fig. 1c). The phosphorylation of eIF2 retarded cross-linking at the sites near the NF motif (Gln117(2Bε) and Leu257(2Bε)), demonstrating that the interaction of the γ-subunit of eIF2(αP) with the NF motif-surrounding region of eIF2Bε is much less efficient than that of the unphosphorylated form (Extended Data Fig. 4b, c). This observation may indicate why eIF2(αP) is a poor substrate for nucleotide exchange by eIF2B. Meanwhile, cross-linking at the sites distant from the NF motif (Glu204(2Bε) and Ser258(2Bγ)) was hardly affected by the eIF2 phosphorylation (Extended Data Fig. 4b, c). The different effects of the eIF2 phosphorylation may reflect the dual functions of the catalytic subcomplex on eIF2 (ref. 17). The catalytic subcomplex also catalyses the displacement of eIF5, which detaches from the ribosome together with eIF2 and inhibits the dissociation of eIF2γ-bound GDP (ref. 18). This activity is insensitive to phosphorylation and impaired by eIF2Bγ mutations17. Therefore, the observed phosphorylation-insensitive cross-linking at the distant sites is likely to have trapped the interaction for this step. Furthermore, several VWM mutations are mapped on the same surface area of the eIF2Bγ subunit as the phosphorylation-insensitive cross-linking (Fig. 1c and Extended Data Fig. 3d), suggesting that defects in the eIF5 displacement are relevant to VWM disease in these cases.

We next investigated the interaction between the eIF2B regulatory subcomplex and P-eIF2α. pBpa was incorporated at 80 sites in the regulatory α-, β- and δ-subunits, and 13 of them were cross-linked with S. pombe P-eIF2α. The cross-linked sites are distributed in all three regulatory subunits, and are clustered within the cavity-like regions of the eIF2B regulatory subcomplex (Fig. 2a and Extended Data Fig. 5a, b). The cavities are formed around the centre of one set of the α-, β- and δ-subunits (‘central cavity’) on the top/bottom ends of the hexameric regulatory subcomplex (Fig. 1a). Thus, eIF2B possesses two P-eIF2α-binding sites, in agreement with the ITC results (Extended Data Fig. 1c). This cavity bears some residues with mutations that have been isolated as Gcn (general control non-depressible) mutations19, 20, which prevent cells from inducing translational control upon eIF2 phosphorylation, and their positions overlap with the eIF2α-cross-linking sites (Extended Data Fig. 6a–c). In addition, the majority of the Gcn mutation positions are located at the interfaces between the regulatory dimers (Extended Data Fig. 6d, e), demonstrating that the correct assembly of the regulatory subunits is requisite for the strong binding with P-eIF2α and the induction of translational control. In contrast, fewer missense VWM mutations were identified in the central cavity (Extended Data Fig. 3f). Therefore, it seems that VWM disease occurs in most cases by mechanisms unrelated to the inhibition of the GEF activity of eIF2B by the eIF2α phosphorylation. Intriguingly, a similar cross-linking pattern was observed without eIF2α phosphorylation (Extended Data Fig. 5c), although the binding was weaker. The difference was that additional cross-links occurred at Arg84(2Bβ) and Gln91(2Bβ) when eIF2α was not phosphorylated, and both of these residues exist in the interior of the cavity (Extended Data Fig. 5d). This suggests that the central cavity can accommodate both phosphorylated and unphosphorylated eIF2α in similar manners, but the interaction with eIF2α is somewhat delocalized in the absence of phosphorylation.

Figure 2: Photo-cross-linking between eIF2α and the central cavity of the eIF2B regulatory subunit.
Photo-cross-linking between eIF2α and the central cavity of the eIF2B regulatory subunit.

a, Mapping of the cross-linked (orange) eIF2B sites with S. pombe eIF2α on the crystal structure (surface model) (see Extended Data Fig. 5a). Asn100(2Bβ) is disordered in this structure. Inset: the same view as in Fig. 1a (surface model). b, Mapping of the eIF2α sites cross-linked with eIF2Bα and eIF2Bβ on the human eIF2α structure21 (ribbon model; see Extended Data Fig. 7a, b). The sites cross-linked with eIF2Bα, eIF2Bβ and both eIF2Bα and eIF2Bβ are represented as blue, cyan and purple spheres, respectively. The phosphorylated residue Ser51(2α) is represented as the red sphere. The KGYID sequence is coloured brown. c, Docking of the eIF2B and eIF2α structures, based on our photo-cross-linking experiments. Inset: the same view as in a.

To delineate how eIF2α is accommodated in the central cavity, we introduced pBpa at 27 sites in the N-terminal domain of S. pombe eIF2α (eIF2α-NTD) and examined the cross-linking with the eIF2B regulatory subunits. We successfully detected the cross-linking of eIF2α with eIF2Bα and eIF2Bβ (Extended Data Fig. 7a, b). Mapping of the cross-linked sites onto the human eIF2α structure21 revealed that the tip of eIF2α-NTD is located close to eIF2Bα and eIF2Bβ (Fig. 2b). On the basis of these complementary experiments, the structures of eIF2α and eIF2B were docked in accordance with the cross-linking results (Fig. 2c), which indicated that eIF2α-NTD is buried in the central cavity along with the notch between the β- and δ-subunits of eIF2B. The eIF2α KGYID sequence (residues 79–83), which is important for the strong binding to eIF2B9, 22, closely faces eIF2Bα (Fig. 2c, inset). The residues Ser48, Leu84 and Val89 of eIF2α, whose mutations suppress the inhibitory effect of the eIF2α phosphorylation23, are located at the bottom of the cavity. Therefore, our docking model seems to be reasonable. In the model, the phosphorylated residue Ser51(2α) is also located at the bottom of the cavity (Fig. 2c, inset). The cavity has no positively charged patch suitable for the accommodation of a phosphate group (Extended Data Fig. 7c, d). Therefore, the mechanism underlying the enhanced affinity for P-eIF2α is still unclear.

Considering our finding that the cross-linking at the sites around the NF motif is impaired by the eIF2 phosphorylation, as described above, eIF2γ seems to minimally contact the NF motif when P-eIF2α-NTD binds to the central cavity. On the basis of the docking model of P-eIF2α-NTD captured by the central cavity, we further docked the trimeric aIF2, the archaeal homologue of eIF2 (ref. 24), on the eIF2B structure. This docking model indicates that it is difficult for the aIF2/eIF2 captured by the central cavity to simultaneously interact with the NF motif (Fig. 3a). Therefore, the central-cavity-captured state is the ‘nonproductive’ state, which is distinct from the ‘productive’ state for efficient nucleotide exchange (Fig. 3b, left)5. When eIF2α is not phosphorylated, eIF2γ-bound GDP is exchanged with GTP by the HEAT domain and the NF motif, because the nonproductive state is not stable (Fig. 3b, left). The stress-induced eIF2α phosphorylation stabilizes the nonproductive state by the stronger interaction in the central cavity, thereby impeding the nucleotide exchange on the bound eIF2 and the entry of another eIF2 (Fig. 3b, right). According to this model, the Gcn mutations, which alleviate translational control upon eIF2 phosphorylation, should destabilize the strong interaction in the central cavity. We selected two Gcn mutations in the central cavity, Glu57Lys(2Bα) and Asp248Lys(2Bδ) (Glu44Lys(2Bα) and Glu377Lys(2Bδ) in Saccharomyces cerevisiae) (Extended Data Fig. 8a). These eIF2B mutations abrogated its strong interactions with eIF2(αP) and P-eIF2α (Extended Data Fig. 8b, c). The GDP exchange assay also revealed the alleviation of the inhibition by eIF2(αP) (Extended Data Fig. 8d). Therefore, these Gcn mutations diminish the strong interaction in the central cavity, thus restoring efficient nucleotide exchange (Fig. 3b).

Figure 3: Model of the eIF2–eIF2B interactions.
Model of the eIF2–eIF2B interactions.

a, Docking of the eIF2 and eIF2B structures, by fixing eIF2α-NTD in the central cavity of eIF2B. eIF2α-NTD is placed in the same manner as in Fig. 2c. The structures of aIF2α-CTD and aIF2γ from PDB accession number 2QMU24, the archaeal homologues of the corresponding segments of eIF2, are positioned so they extend towards the NF motif (red) from the C terminus of eIF2α-NTD. b, Schematic representations of the eIF2–eIF2B interactions proposed in the present study.

These analyses additionally revealed that the Gcn mutations also affect the interaction with unphosphorylated eIF2α, and enhance the nucleotide exchange activity of eIF2B (Extended Data Fig. 8c, e). The nucleotide exchange activity of eIF2B is also enhanced by a small molecule called integrated stress response inhibitor (ISRIB)25, 26, which was identified as a compound that blocks the stress response27. The ISRIB-resistant mutations26 are located near the pseudo-twofold rotational axis of eIF2B (Extended Data Fig. 9). Therefore, our structure is consistent with the proposed models of action of ISRIB, in which this symmetric molecule stabilizes the eIF2B decamer25, 26.

The current study of eIF2B provides the structural basis not only for an overall understanding of its function in translational control, but also of the mechanisms of VWM. We have focused so far on the structural integrity and the catalytic activities of nucleotide exchange and eIF5 displacement, as the causes of this disease. However, VWM disease can reportedly result from alterations in other, as yet undescribed, eIF2B functions28, 29. Mapping of missense VWM mutations on the structure actually revealed the existence of some exposed mutations that may not be explained by defects in the known functions (for example, γArg312Gln and γIle346Thr in Extended Data Fig. 3b and the mutations in Extended Data Fig. 3e). The present decamer structure is expected to contribute to the elucidation of the currently undescribed mechanisms of this disease.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Expression and purification of recombinant S. pombe eIF2B

The detailed methods for expression and purification of the native eIF2B protein are described elsewhere (Kashiwagi et al., submitted). For the production of the selenomethionine (SeMet)-substituted protein, the cells were grown in M9 medium until the absorbance, A600 nm, reached 0.4, supplemented with SeMet and amino acids, grown for 30 min, induced with 0.5 mM IPTG, and further grown at 18 °C overnight30. The SeMet-derivative protein was also purified in the same manner as described, except the DTT concentration was increased to 10 mM during the purification and the Ni-Sepharose flow-through step was omitted. For cross-linking experiments, the epitope tags were fused to the regulatory subunits (α: myc-tag at the N terminus; β: HA-tag at the N terminus; δ: Strep-tag at the C terminus) using a PrimeSTAR Mutagenesis Basal Kit (Takara).

Purification of K. pastoris eIF2

We used K. pastoris eIF2 for nucleotide exchange assays and SEC analyses, because a high quality sample of K. pastoris eIF2 can be prepared more efficiently than S. pombe eIF2, by the following methods. First, the DNA sequence encoding the THF (Thrombin-His6-3 × Flag) tag was inserted into the K. pastoris genomic locus corresponding to the C terminus of eIF2γ, by homologous recombination31. The resultant K. pastoris strain producing the THF-tagged eIF2γ was grown in medium containing 1% yeast extract, 2% tryptone and 5% (v/v) glycerol at 30 °C. Harvested cells were suspended in 75 mM Tris-OAc buffer (pH 7.5), containing 300 mM KOAc, 10% (v/v) glycerol, 5 mM Mg(OAc)2, 1 mM EDTA, 2 mM DTT and protease inhibitors, and were disrupted using an FPG12800 Pressure Cell Homogenizer (Stansted Fluid Power). After centrifugation, the supernatant was purified by chromatography on Q-Sepharose, Ni-Sepharose, HiTrap Heparin, MonoS and Superdex 200 (GE Healthcare) columns. The final buffer was 20 mM HEPES-KOH buffer (pH 7.5), containing 150 mM KOAc, 5% (v/v) glycerol, 5 mM Mg(OAc)2, 0.1 mM EDTA and 1 mM DTT.

Expression and purification of recombinant S. pombe eIF2α

S. pombe eIF2α was produced as the N-terminally GST-tagged and C-terminally His6-Flag-tagged protein, in E. coli Rosetta2 (DE3). Transformed cells were grown in lysogeny broth (LB) medium at 37 °C. After the addition of 0.3 mM IPTG when the culture reached A600 nm = 0.4, the cells were grown at 18 °C overnight. The cells were lysed by sonication in 20 mM HEPES-KOH buffer (pH 7.5), containing 150 mM KCl, 10% (v/v) glycerol, 1 mM DTT and protease inhibitors, and the lysate was fractionated on Ni-Sepharose and HiTrap Heparin columns. After overnight cleavage of the GST tag with the HRV 3C protease at 4 °C, eIF2α was purified on the HiTrap Heparin and Superdex 200 columns in 20 mM HEPES-KOH buffer (pH 7.5), containing 150 mM KCl, 5% (v/v) glycerol and 1 mM DTT.

In vitro phosphorylation of the α-subunit of eIF2

The active human PKR protein was prepared as previously described, with some modifications32, 33. Briefly, the full-length PKR protein was expressed in E. coli Rosetta2 (DE3) cells as the N-terminally His6-tagged form, and dephosphorylated in vivo by co-expression with λ protein phosphatase. Transformed cells were grown in LB medium supplemented with 0.2% glucose at 37 °C. After the addition of 0.3 mM IPTG when the culture reached A600 nm = 0.6, the cells were grown at 20 °C for 24 h. The cells were lysed by sonication in 20 mM HEPES-NaOH buffer (pH 7.5), containing 50 mM NaCl, 10% (v/v) glycerol, 1 mM DTT and protease inhibitors, and the lysate was centrifuged. The supernatant was fractionated on Heparin-Sepharose, Ni-Sepharose, HiTrap SP and Superdex 200 (GE Healthcare) columns. For the preparation of P-eIF2α and eIF2(αP), PKR and S. pombe eIF2α were dialysed against 20 mM HEPES-KOH buffer (pH 7.5), containing 50 mM KCl, 5 mM MgCl2, 0.1 mM EDTA and 1 mM DTT, and K. pastoris eIF2 was dialysed against the same buffer containing 100 mM KCl. PKR was concentrated to 2 mg ml−1, and activated by an incubation with 0.5 mM ATP at 30 °C for 1 h. eIF2α and eIF2 were phosphorylated by an incubation with the activated PKR and 0.5 mM ATP at 25 °C. Phosphorylation was confirmed by Phos-tag SDS–PAGE (Wako).

Guanine-nucleotide exchange assay

The eIF2–[3H]GDP binary complex was formed by incubating 90 pmol of K. pastoris eIF2 with 37.5 pmol of [3H]GDP in assay buffer (20 mM HEPES-KOH buffer (pH 7.5) containing 100 mM KCl, 10% (v/v) glycerol, 0.1 mM EDTA, 5 mM NaF, 1 mM DTT and 2 mg ml−1 BSA), at 21 °C for 10 min. After the addition of 3 mM MgCl2 and 0.1 mM ATP, the eIF2–[3H]GDP complex was further incubated with or without 1 μl of the activated PKR for 5 min, and kept on ice. After a 5 min incubation at 15 °C, measurements were started by the simultaneous additions of a 100-fold amount of GDP and 22.5 pmol of eIF2B. At each time point, samples were transferred into 2.5 ml of ice-cold wash buffer (20 mM HEPES-KOH buffer (pH 7.5) containing 100 mM KCl and 5 mM MgCl2), and immediately vacuum-filtered through mixed cellulose ester filters (Advantec). After two washes with 2.5 ml of the ice-cold wash buffer, the filters were dried and the radioactivity was quantitated by liquid scintillation counting.

Size-exclusion chromatography analysis

A mixture of 500 pmol of S. pombe eIF2B and 1 nmol of K. pastoris eIF2 was incubated in 20 mM HEPES-KOH buffer (pH 7.5), containing 300 mM KOAc, 5% (v/v) glycerol, 3 mM Mg(OAc)2, 0.1 mM EDTA and 1 mM DTT, at 4 °C. After the sample volume was adjusted to 200 μl, it was fractionated on a Superose 6 column (GE Healthcare).

Isothermal titration calorimetry

S. pombe eIF2B and eIF2α were applied to Sephacryl S-300 and Superdex 200 columns, respectively, in 20 mM HEPES-KOH buffer (pH 7.5), containing 300 mM KOAc, 5% (v/v) glycerol, 3 mM Mg(OAc)2 and 0.1 mM EDTA. The measurements were performed with a MicroCal Auto-iTC200 calorimeter (Malvern). The titration was performed by injecting 2 μl of eIF2α (200 μM) into the eIF2B solution (20 μM), 19 times. To calculate the binding stoichiometry and the dissociation constants, two runs were concatenated.

Crystallization and structure determination

The native and SeMet-derivative eIF2B samples were concentrated to 5–6 mg ml−1, and their crystals were obtained and cryoprotected in similar manners to those described in Kashiwagi et al. (submitted). The final data sets were collected at BL41XU of SPring-8 (Hyogo, Japan). Data collection was performed at 100 K, and the wavelengths were 1.0000 Å for the native data set and 0.9792 Å for the SeMet-derivative data set. The data sets were processed with XDS34 and SCALA35. Data collection statistics are summarized in Extended Data Table 1. The initial phases were determined from the SeMet-derivative data set by the single-wavelength anomalous dispersion (SAD) method, by the use of the autoSHARP pipeline36. The molecular model was built automatically with Buccaneer37. The model was modified manually in Coot38, and refined by PHENIX39 against the native data set. In the Ramachandran plot, 93.4% of the residues in the model are in the favoured region, and 6.0% and 0.6% of the residues are in the allowed and disallowed regions, respectively. The electrostatic surface potential was determined with APBS40.

Surface-scanning site-directed photo-cross-linking assays

The incorporation of pBpa was performed by the expanded genetic code method: pBpa was incorporated site-specifically into proteins using an E. coli RFzero strain, in which the UAG codon is reassigned to pBpa41, 42. For the expression of pBpa-labelled S. pombe eIF2B, the codon at a specified position in the expression constructs (Kashiwagi et al., submitted) was changed to a TAG triplet, using a PrimeSTAR Mutagenesis Basal Kit. For the pBpa-labelled S. pombe eIF2α, the GST tag of the expression construct was replaced by MBP, and one codon was changed to TAG. The proteins labelled with pBpa were produced with the BL21 (DE3)-based RFzero strain, expressing pBpaRS and UAG-decoding tRNA41, 42. The cells were grown in LB medium supplemented with 0.2% glucose and 1 mM pBpa, at 37 °C. When the culture reached A600 nm = 0.8, 0.5 mM IPTG was added, and the cells were grown at 20 °C for 24 h. The pBpa-labelled eIF2B and eIF2α were purified with Amylose Resin (New England Biolabs), and the MBP tags were cleaved with HRV 3C protease. For the cross-linking experiments between eIF2B and eIF2α, the proteins were dialysed against 20 mM HEPES-KOH buffer (pH 7.5), containing 150 mM KCl, 5% (v/v) glycerol and 1 mM DTT. For the cross-linking experiments between eIF2B and K. pastoris eIF2, the proteins were dialysed against 20 mM HEPES-KOH buffer (pH 7.5), containing 150 mM KOAc, 5% (v/v) glycerol, 3 mM Mg(OAc)2, 0.1 mM EDTA and 1 mM DTT. The pBpa-labelled protein was divided into aliquots corresponding to the amount purified from 50 ml LB culture, and mixed with 100 pmol of the cross-linking target protein, and then the total volume was adjusted to 110 μl. The final concentrations of the pBpa-labelled proteins were about 0.5–1.0 μM. A series of protein variants, each with pBpa incorporated at one specified surface site, was ultraviolet-irradiated in the presence of the cross-linking target protein. Ultraviolet irradiation at 365 nm wavelength was performed on ice for 5 min. The cross-linked products were detected by western blotting. The Flag-tag, myc-tag, HA-tag and Strep-tag were detected using Anti-DDDDK-tag mAb-HRP-DirecT (M185-7, MBL), Anti-c-Myc-Peroxidase antibody (A5598, Sigma), Monoclonal Anti-HA−Peroxidase antibody (H6533, Sigma) and Strep•Tag II Antibody HRP Conjugate (71591, Novagen), respectively. For the time-course analysis of the cross-linking of the pBpa-labelled eIF2B with eIF2 or eIF2(αP), ATP alone or with the activated PKR was added to the dialysed eIF2, kept at 21 °C for 5 min and then mixed with eIF2B.

Accession codes

Primary accessions

Protein Data Bank

References

  1. Pavitt, G. D. eIF2B, a mediator of general and gene-specific translational control. Biochem. Soc. Trans. 33, 14871492 (2005)
  2. Jackson, R. J., Hellen, C. U. T. & Pestova, T. V. The mechanism of eukaryotic translation initiation and principles of its regulation. Nature Rev. Mol. Cell Biol. 11, 113127 (2010)
  3. Kuhle, B., Eulig, N. K. & Ficner, R. Architecture of the eIF2B regulatory subcomplex and its implications for the regulation of guanine nucleotide exchange on eIF2. Nucleic Acids Res. 43, 999410014 (2015)
  4. Yang, W. & Hinnebusch, A. G. Identification of a regulatory subcomplex in the guanine nucleotide exchange factor eIF2B that mediates inhibition by phosphorylated eIF2. Mol. Cell. Biol. 16, 66036616 (1996)
  5. Pavitt, G. D., Ramaiah, K. V. A., Kimball, S. R. & Hinnebusch, A. G. eIF2 independently binds two distinct eIF2B subcomplexes that catalyze and regulate guanine-nucleotide exchange. Genes Dev. 12, 514526 (1998)
  6. Gomez, E., Mohammad, S. S. & Pavitt, G. D. Characterization of the minimal catalytic domain within eIF2B: the guanine-nucleotide exchange factor for translation initiation. EMBO J. 21, 52925301 (2002)
  7. Gomez, E. & Pavitt, G. D. Identification of domains and residues within the ε subunit of eukaryotic translation initiation factor 2B (eIF2Bε) required for guanine nucleotide exchange reveals a novel activation function promoted by eIF2B complex formation. Mol. Cell. Biol. 20, 39653976 (2000)
  8. Wek, R. C., Jiang, H. Y. & Anthony, T. G. Coping with stress: eIF2 kinases and translational control. Biochem. Soc. Trans. 34, 711 (2006)
  9. Krishnamoorthy, T., Pavitt, G. D., Zhang, F., Dever, T. E. & Hinnebusch, A. G. Tight binding of the phosphorylated α subunit of initiation factor 2 (eIF2α) to the regulatory subunits of guanine nucleotide exchange factor eIF2B is required for inhibition of translation initiation. Mol. Cell. Biol. 21, 50185030 (2001)
  10. Pavitt, G. D. & Proud, C. G. Protein synthesis and its control in neuronal cells with a focus on vanishing white matter disease. Biochem. Soc. Trans. 37, 12981310 (2009)
  11. Fogli, A. & Boespflug-Tanguy, O. The large spectrum of eIF2B-related diseases. Biochem. Soc. Trans. 34, 2229 (2006)
  12. Wang, X., Wortham, N. C., Liu, R. & Proud, C. G. Identification of residues that underpin interactions within the eukaryotic initiation factor (eIF2) 2B complex. J. Biol. Chem. 287, 82638274 (2012)
  13. Reid, P. J., Mohammad-Qureshi, S. S. & Pavitt, G. D. Identification of intersubunit domain interactions within eukaryotic initiation factor (eIF) 2B, the nucleotide exchange factor for translation initiation. J. Biol. Chem. 287, 82758285 (2012)
  14. Wortham, N. C., Martinez, M., Gordiyenko, Y., Robinson, C. V. & Proud, C. G. Analysis of the subunit organization of the eIF2B complex reveals new insights into its structure and regulation. FASEB J. 28, 22252237 (2014)
  15. Bogorad, A. M. et al. Insights into the architecture of the eIF2Bα/β/δ regulatory subcomplex. Biochemistry 53, 34323445 (2014)
  16. Gordiyenko, Y. et al. eIF2B is a decameric guanine nucleotide exchange factor with a γ2ε2 tetrameric core. Nature Commun. 5, 3902 (2014)
  17. Jennings, M. D., Zhou, Y., Mohammad-Qureshi, S. S., Bennett, D. & Pavitt, G. D. eIF2B promotes eIF5 dissociation from eIF2•GDP to facilitate guanine nucleotide exchange for translation initiation. Genes Dev. 27, 26962707 (2013)
  18. Jennings, M. D. & Pavitt, G. D. eIF5 has GDI activity necessary for translational control by eIF2 phosphorylation. Nature 465, 378381 (2010)
  19. Vazquez de Aldana, C. R. & Hinnebusch, A. G. Mutations in the GCD7 subunit of yeast guanine nucleotide exchange factor eIF-2B overcome the inhibitory effects of phosphorylated eIF-2 on translation initiation. Mol. Cell. Biol. 14, 32083222 (1994)
  20. Pavitt, G. D., Yang, W. & Hinnebusch, A. G. Homologous segments in three subunits of the guanine nucleotide exchange factor eIF2B mediate translational regulation by phosphorylation of eIF2. Mol. Cell. Biol. 17, 12981313 (1997)
  21. Ito, T., Marintchev, A. & Wagner, G. Solution structure of human initiation factor eIF2α reveals homology to the elongation factor eEF1B. Structure 12, 16931704 (2004)
  22. Dey, M. et al. PKR and GCN2 kinases and guanine nucleotide exchange factor eukaryotic translation initiation factor 2B (eIF2B) recognize overlapping surfaces on eIF2α. Mol. Cell. Biol. 25, 30633075 (2005)
  23. Vazquez de Aldana, C. R., Dever, T. E. & Hinnebusch, A. G. Mutations in the α subunit of eukaryotic translation initiation factor 2 (eIF-2α) that overcome the inhibitory effect of eIF-2α phosphorylation on translation initiation. Proc. Natl Acad. Sci. USA 90, 72157219 (1993)
  24. Yatime, L., Mechulam, Y., Blanquet, S. & Schmitt, E. Structure of an archaeal heterotrimeric initiation factor 2 reveals a nucleotide state between the GTP and the GDP states. Proc. Natl Acad. Sci. USA 104, 1844518450 (2007)
  25. Sidrauski, C. et al. Pharmacological dimerization and activation of the exchange factor eIF2B antagonizes the integrated stress response. eLife 4, e07314 (2015)
  26. Sekine, Y. et al. Stress responses. Mutations in a translation initiation factor identify the target of a memory-enhancing compound. Science 348, 10271030 (2015)
  27. Sidrauski, C. et al. Pharmacological brake-release of mRNA translation enhances cognitive memory. eLife 2, e00498 (2013)
  28. Liu, R. et al. Severity of vanishing white matter disease does not correlate with deficits in eIF2B activity or the integrity of eIF2B complexes. Hum. Mutat. 32, 10361045 (2011)
  29. Wortham, N. C. & Proud, C. G. Biochemical effects of mutations in the gene encoding the alpha subunit of eukaryotic initiation factor (eIF) 2B associated with vanishing white matter disease. BMC Med. Genet. 16, 64 (2015)
  30. Van Duyne, G. D., Standaert, R. F., Karplus, P. A., Schreiber, S. L. & Clardy, J. Atomic structures of the human immunophilin FKBP-12 complexes with FK506 and rapamycin. J. Mol. Biol. 229, 105124 (1993)
  31. Higo, T. et al. Development of a hexahistidine-3 × FLAG-tandem affinity purification method for endogenous protein complexes in Pichia pastoris. J. Struct. Funct. Genomics 15, 191199 (2014)
  32. Matsui, T., Tanihara, K. & Date, T. Expression of unphosphorylated form of human double-stranded RNA-activated protein kinase in Escherichia coli. Biochem. Biophys. Res. Commun. 284, 798807 (2001)
  33. Lemaire, P. A., Lary, J. & Cole, J. L. Mechanism of PKR activation: dimerization and kinase activation in the absence of double-stranded RNA. J. Mol. Biol. 345, 8190 (2005)
  34. Kabsch, W. XDS. Acta Crystallogr. D 66, 125132 (2010)
  35. Evans, P. Scaling and assessment of data quality. Acta Crystallogr. D 62, 7282 (2006)
  36. Vonrhein, C., Blanc, E., Roversi, P. & Bricogne, G. Automated structure solution with autoSHARP. Methods Mol. Biol. 364, 215230 (2007)
  37. Cowtan, K. The Buccaneer software for automated model building. 1. Tracing protein chains. Acta Crystallogr. D 62, 10021011 (2006)
  38. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D 60, 21262132 (2004)
  39. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D 66, 213221 (2010)
  40. Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl Acad. Sci. USA 98, 1003710041 (2001)
  41. Mukai, T. et al. Genetic-code evolution for protein synthesis with non-natural amino acids. Biochem. Biophys. Res. Commun. 411, 757761 (2011)
  42. Chin, J. W., Martin, A. B., King, D. S., Wang, L. & Schultz, P. G. Addition of a photocrosslinking amino acid to the genetic code of Escherichia coli. Proc. Natl Acad. Sci. USA 99, 1102011024 (2002)
  43. Hiyama, T. B., Ito, T., Imataka, H. & Yokoyama, S. Crystal structure of the α subunit of human translation initiation factor 2B. J. Mol. Biol. 392, 937951 (2009)
  44. Nakamura, A. et al. Dynamic, ligand-dependent conformational change triggers reaction of ribose-1,5-bisphosphate isomerase from Thermococcus kodakarensis KOD1. J. Biol. Chem. 287, 2078420796 (2012)
  45. Jin, X., Ballicora, M. A., Preiss, J. & Geiger, J. H. Crystal structure of potato tuber ADP-glucose pyrophosphorylase. EMBO J. 24, 694704 (2005)
  46. Geva, M. et al. A mouse model for eukaryotic translation initiation factor 2B-leucodystrophy reveals abnormal development of brain white matter. Brain 133, 24482461 (2010)
  47. Dev, K. et al. Archaeal aIF2B interacts with eukaryotic translation initiation factors eIF2α and eIF2Bα: implications for aIF2B function and eIF2B regulation. J. Mol. Biol. 392, 701722 (2009)

Download references

Acknowledgements

We thank the staff of the beamline BL41XU at SPring-8 for their support. This work was performed with the approval of the Japan Synchrotron Radiation Research Institute (proposals 2012A1335 and 2012B1572). This work was supported by JSPS KAKENHI grants 23687013 and 25121737 (to T.I.), the Targeted Proteins Research Program (TPRP) and the Platform for Drug Discovery, Informatics and Structural Life Science, from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan (to S.Y.), and a research program of the UT-RIKEN Cooperation Laboratory of Structural Biology (to S.Y.).

Author information

Affiliations

  1. Graduate School of Science, The University of Tokyo, Bunkyo-ku, Tokyo 113-0033, Japan

    • Kazuhiro Kashiwagi,
    • Takuya B. Hiyama,
    • Takuhiro Ito &
    • Shigeyuki Yokoyama
  2. RIKEN Systems and Structural Biology Center, Tsurumi-ku, Yokohama 230-0045, Japan

    • Kazuhiro Kashiwagi,
    • Takuya B. Hiyama,
    • Toshiaki Higo,
    • Takashi Umehara,
    • Kensaku Sakamoto,
    • Takuhiro Ito &
    • Shigeyuki Yokoyama
  3. RIKEN Center for Life Science Technologies, Tsurumi-ku, Yokohama 230-0045, Japan

    • Kazuhiro Kashiwagi,
    • Mari Takahashi,
    • Madoka Nishimoto,
    • Toshiaki Higo,
    • Takashi Umehara,
    • Kensaku Sakamoto &
    • Takuhiro Ito
  4. RIKEN Structural Biology Laboratory, Tsurumi-ku, Yokohama 230-0045, Japan

    • Shigeyuki Yokoyama

Contributions

K.K. and T.I. designed and performed experiments, and wrote the manuscript. M.T., M.N. and T.B.H. purified the samples. T.H. and T.U. established the K. pastoris strain. K.S. provided the materials and technical advice for the surface-scanning photo-cross-linking experiments. S.Y. designed experiments and wrote the manuscript.

Competing financial interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to:

The atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB) under accession number 5B04.

Author details

Extended data figures and tables

Extended Data Figures

  1. Extended Data Figure 1: Biochemical properties of recombinant S. pombe eIF2B. (358 KB)

    a, Guanine-nucleotide exchange catalysed by S. pombe eIF2B, and its inhibition by eIF2 phosphorylation. K. pastoris eIF2 was labelled with [3H]GDP and incubated with ATP and activated PKR (eIF2(αP); red line), or with ATP only (eIF2; blue line). Reactions were started by the addition of excess unlabelled GDP with eIF2B (solid line) or buffer (dashed line). Individual data points of triplicate analyses are shown by dots and boxes. The dissociation rates of GDP from eIF2 are summarized in the table. b, The SEC profiles of respective proteins (S. pombe eIF2B (green), K. pastoris eIF2 (cyan), eIF2(αP) (pink)) (top), eIF2B with K. pastoris eIF2 (middle) and eIF2B with eIF2(αP) (bottom). The chromatograms of the absorbance at 280 nm and the Coomassie blue stained SDS–PAGE gels are shown. For gel source data, see Supplementary Fig. 1. c, ITC measurements between S. pombe eIF2B (as the decamer) and S. pombe eIF2α (top), eIF2B and P-eIF2α (bottom). We used the eIF2α subunit alone, rather than the trimeric eIF2, in these experiments because of the difficulty in the preparation of highly concentrated eIF2. Representative thermograms are shown. Two runs were concatenated for analysis. The binding stoichiometry (N) and the dissociation constant (Kd) were calculated from triplicate analyses (mean ± s.d.).

  2. Extended Data Figure 2: Architecture of the eIF2B subcomplexes. (1,112 KB)

    a, b, Two different views of the crystal structure of S. pombe eIF2B (wall-eyed stereo view), coloured as in Fig. 1a and from the same view as in Fig. 1b and Fig. 1c, respectively. The assembly of the subcomplexes is primarily mediated by the β–ε and δ–γ interactions. In this regard, the results of the two prior mutational analyses12, 13 are consistent with the present structure, except that one suggested that the β-helical region of the γ-subunit is involved in the inter-subunit interactions. ck, Ribbon models of the structures of S. pombe eIF2B subcomplexes (ce, g, h, j) and proteins with structural similarity (f, i, k). c, d, The α2 homodimer (c) and the βδ heterodimer (d) in the eIF2B decamer. The conformations of each subunit are similar to those in the human α2 homodimer43 and the Chaetomium thermophilum βδ heterotetramer3, even though the relative orientations of the subunits in the dimers are slightly different from those represented in these partial structures. e, f, The regulatory subcomplex in the eIF2B decamer (e) and ribose-1,5-bisphosphate (R15Pi) homohexamer44 (PDB 3A11) (f). The architecture of the regulatory subcomplex is an assembly of three similarly shaped dimer moieties: one homodimer of the α-subunit and two heterodimers of the β- and δ-subunits. The arrangement of the regulatory subunits resembles that in the C. thermophilum βδ heterotetramer3, and shares some similarity to that in the homohexameric structure of R15Pi44. g, The γ-subunit of eIF2B. h, The ε-subunit of eIF2B. i, Potato tuber ADP-glucose pyrophosphorylase (AGP) (PDB 1YP3)45. The dimerization interfaces between the catalytic subunits are coloured in deeper shades (g, h). j, k, The subunit heterodimerization mode in the catalytic subcomplex of eIF2B (j) and the subunit homodimerization mode in the potato tuber AGP (k). The dimerization manner of the γ- and ε-subunits is novel: each of their structures resembles the subunit structure of the AGP homotetramer45, but they dimerize through their N-terminal domains, in a different manner than the AGP homotetramer45.

  3. Extended Data Figure 3: Mapping of the residues corresponding to missense VWM mutations on the subunit interfaces and the distal face of eIF2B. (754 KB)

    The eIF2B residues corresponding to VWM-causing missense mutations in human (Supplementary Table 1) are mapped on the S. pombe eIF2B structure, with the same subunit colouring as in Fig. 1. The S. pombe eIF2B residues corresponding to VWM-causing missense mutations are shown in parentheses. a, VWM-related residues are mapped as spheres on the overall structure (ribbon model). The environments of the residues are colour-coded (green, solvent-exposed; yellow, subunit interface; brown, structural core) on the spheres. b, VWM-related residues are mapped on the surface model of the inter-subcomplex interfaces on the catalytic subcomplex side (left), with the interfaces for the α-, β- and δ-subunits coloured blue, cyan and green, respectively, and on the βδ dimer side (right), with the interfaces for the γ- and ε-subunits coloured orange and pink, respectively. VWM-related residues around the βδ dimerization interface are shown as spheres in the inset. The mutations in the regulatory subunits are clustered around the dimerization interface between the β- and δ-subunits, as mentioned in ref. 3. Our structure further revealed that the binding site for the ε-subunit is formed by the correct interaction between the β- and δ-subunits, thus explaining the abundance of mutations around this interface. c, VWM-related residues in the α2 homodimer are located around the homodimerization interface, and shown as spheres in the inset. These VWM-related mutations around the subunit interfaces (b, c) may cause appreciable degrees of subunit dissociation from the eIF2B decamer, leading to incomplete complexes, destabilization of eIF2B resulting in aggregation/degradation, and/or changes in the conformation and activity of the intact eIF2B decamer. d, VWM-related residues on the distal face of the catalytic subcomplex are mapped on the surface model. The NF motif is shown in red. Several VWM-related residues are located near the NF motif, including Arg111(2Bε), corresponding to the human Arg136His(2Bε) mutation, for which the mouse model is available46. e, The ε-subunit further contains several exposed missense VWM mutations, especially in the β-helical domain. f, VWM-related residues in the central cavity. Only one residue, corresponding to the human Lys110Glu(2Bα) mutation, is exposed to the solvent.

  4. Extended Data Figure 4: Photo-cross-linking between pBpa-labelled eIF2B and eIF2. (447 KB)

    a, The S. pombe eIF2B variants bearing a single site-specific pBpa substitution in the catalytic subunits were mixed with K. pastoris eIF2, and irradiated with ultraviolet (365 nm) for 5 min on ice. Since eIF2γ harbours the Flag-tag at the C terminus, the products cross-linked with eIF2γ were detected by western blotting with an anti-Flag antibody. Site-specific slow-migrating bands that appeared after ultraviolet irradiation were judged as cross-linked bands. The relevant bands are indicated with teal dots. b, The eIF2γ-cross-linked sites are shown in teal, except for the selected ones explained below, and the cross-link-negative sites are shown in grey, on the surface model of the catalytic subcomplex, coloured in the same manner as in Fig. 1. The NF motif is shown in red. c, Time-course analysis of cross-linking with eIF2γ. Four selected sites on the distal face (Gln117(2Bε) (light green in b), Leu257(2Bε) (blue), Glu204(2Bε) (purple) and Ser258(2Bγ) (violet)) were examined by time courses of the cross-linking with eIF2 or eIF2(αP). The ratio of the band intensity of the eIF2B γ- or ε-subunit cross-linked with eIF2(αP) to that with unphosphorylated eIF2 at each time point is shown below the lane. ND means that no band of the eIF2B subunit cross-linked with eIF2(αP) was detected at the time point. For gel source data, see Supplementary Fig. 1.

  5. Extended Data Figure 5: Photo-cross-linking between pBpa-labelled eIF2B and eIF2α. (612 KB)

    a, The S. pombe eIF2B variants bearing a single site-specific pBpa substitution in the regulatory subunit were mixed with S. pombe P-eIF2α, and irradiated with ultraviolet (365 nm) for 5 min. Cross-linking was detected as described in Extended Data Fig. 4. The relevant bands are indicated with orange dots. b, The eIF2α-cross-linked sites are shown in orange, and the cross-link-negative sites are shown in grey, on the surface model of the overall structure of eIF2B, with the same subunit colouring as in Fig. 1. c, The eIF2B variants were similarly mixed with unphosphorylated eIF2α and irradiated with ultraviolet. The bands that were also observed in a are indicated with orange dots, and the unphosphorylated eIF2α-specific bands are indicated with magenta dots. d, Arg84(2Bβ) and Gln91(2Bβ), which exhibited unphosphorylated eIF2α-specific cross-links, are shown in magenta. The view is the same as in Fig. 2a. For gel source data, see Supplementary Fig. 1.

  6. Extended Data Figure 6: Solvent-exposed Gcn mutations located in the central cavity and buried Gcn mutations clustered around the trimerization interface. (440 KB)

    a, The residues corresponding to Gcn mutations19, 20 are mapped in blue on the surface model of the S. pombe eIF2B structure, with the same subunit colouring as in Fig. 1. b, c, The residues corresponding to exposed Gcn mutations are mainly located in the central cavity (b), and their locations coincide with the P-eIF2α cross-link sites shown in orange (c, Extended Data Fig. 5b). d, e, The residues corresponding to Gcn mutations on the α2 homodimer (d) and the βδ heterodimer (e). The residues of S. pombe eIF2B corresponding to S. cerevisiae Gcn mutations are indicated in parentheses. The interfaces for the trimerization of the regulatory dimers are coloured grey. Most of the Gcn-related residues are mapped only on one face of the subunits, as predicted43, 47. The present eIF2B structure revealed that these ‘mutation-rich’ faces are used for the assembly of the regulatory subunits to form the subcomplex, and the Gcn-related residues are clustered on the interface for the trimeric assembly.

  7. Extended Data Figure 7: Cross-linking of pBpa-labelled eIF2α with the central cavity formed by the eIF2B regulatory subunits. (368 KB)

    a, b, The S. pombe eIF2α variants bearing a single site-specific pBpa residue were mixed with epitope-tagged S. pombe eIF2B, and irradiated with ultraviolet (365 nm) for 5 min on ice. Cross-linking was detected as described in Extended Data Fig. 4. Cross-links with eIF2Bα were detected with an anti-myc antibody (a), and those with eIF2Bβ were detected with an anti-HA antibody (b). The relevant bands are indicated with blue dots in a, and cyan dots in b. These cross-linked sites are mapped on the human eIF2α structure21 in Fig. 2b. For gel source data, see Supplementary Fig. 1. c, The model of the eIF2B–eIF2α complex, built on the basis of the cross-linking experiments, is shown in Fig. 2c. The phosphorylated residue Ser51(2α) is highlighted with the magenta circle. d, The electrostatic surface potential of eIF2B, from the same viewpoint as in c. Red and blue colours represent negative and positive potentials, respectively, of ±10kT/e. The cavity has no positively charged patch; therefore, the mechanism underlying the enhanced affinity for P-eIF2α is still unclear. One possible mechanism is a cation-mediated recognition of the phosphoserine residue. The phosphorylated Ser51(2α) may coordinate a cation together with the negatively charged residues at the bottom of the central cavity, although we did not observe any electron density for such cations in the cavity. Another possibility is a phosphorylation-induced conformational change of the Ser51-flanking loop. The phosphorylation of Ser51(2α) may induce the rearrangement of adjacent arginine residues, and enable a stronger interaction with the negatively charged residues at the bottom of the central cavity.

  8. Extended Data Figure 8: Analyses of eIF2B Gcn mutations. (397 KB)

    a, The locations of the Gcn-related residues mutated for the SEC and ITC analyses (Glu57(2Bα) and Asp248(2Bδ)) are indicated on the S. pombe eIF2B structure, with the same view and colouring as in Extended Data Fig. 6b. b, The SEC analyses of the interaction between K. pastoris eIF2(αP) and S. pombe eIF2B, bearing Gcn-related mutations. The chromatograms of the absorbance at 280 nm and the SDS–PAGE gels of each run are shown. The green bar represents the elution point of free eIF2B (Extended Data Fig. 1b). For gel source data, see Supplementary Fig. 1. c, ITC measurements between S. pombe eIF2α or P-eIF2α and eIF2B, bearing Gcn-related mutations. Representative thermograms for the ITC experiments are shown. d, e, The nucleotide exchange activities of S. pombe eIF2B, bearing Gcn-related mutations, on K. pastoris eIF2(αP) (d) and eIF2 (e) were examined as described in Extended Data Fig. 1a (αGlu57Lys mutant, blue line; δAsp248Lys mutant, green line; wild type, grey line). Individual data points of triplicate analyses are shown by dots.

  9. Extended Data Figure 9: The locations of ISRIB-resistant mutations on the eIF2B structure. (597 KB)

    Residues corresponding to the ISRIB-resistant mutations26 are mapped onto the eIF2B structure in red. The residue corresponding to the Arg171Gln(2Bδ) mutation (Lys112(2Bδ)) is in the disordered region, at the N terminus of the δ-subunit. The disordered N-terminal segment of the δ-subunit is indicated by the dotted green line.

Extended Data Tables

  1. Extended Data Table 1: Data collection and refinement statistics (198 KB)

Supplementary information

PDF files

  1. Supplementary Figure 1 (425 KB)

    This file contains full scanned gels in uncropped form.

Excel files

  1. Supplementary Table 1 (18 KB)

    This file contains the list of missense VWM mutations and corresponding residues.

Additional data