|Home | About | Journals | Submit | Contact Us | Français|
The initiation of protein synthesis plays an essential regulatory role in human biology. At the center of the initiation pathway, the 13-subunit eukaryotic translation initiation factor 3 (eIF3) controls access of other initiation factors and mRNA to the ribosome by unknown mechanisms. Using electron microscopy (EM), bioinformatics and biochemical experiments, we identify two highly conserved RNA-binding motifs in eIF3 that direct translation initiation from the hepatitis C virus internal ribosome entry site (HCV IRES) RNA. Mutations in the RNA-binding motif of subunit eIF3a weaken eIF3 binding to the HCV IRES and the 40S ribosomal subunit, thereby suppressing eIF2-dependent recognition of the start codon. Mutations in the eIF3c RNA-binding motif also reduce 40S ribosomal subunit binding to eIF3, and inhibit eIF5B-dependent steps downstream of start codon recognition. These results provide the first connection between the structure of the central translation initiation factor eIF3 and recognition of the HCV genomic RNA start codon, molecular interactions that likely extend to the human transcriptome.
The initiation of eukaryotic protein synthesis is one of the most highly regulated steps in gene expression. Cap-dependent translation requires ≥12 initiation factors (eIFs) to associate with the 7-methyl guanosine cap at the 5′ end of mRNA and the 40S ribosomal subunit, leading to the formation of a 43S preinitiation complex that scans to the correct initiation codon and positions the initiating methionyl-tRNA (Met-tRNAi). The resulting 48S preinitiation complex then associates with the 60S ribosomal subunit to form an active 80S ribosome (1–3). In viruses such as hepatitis C virus (HCV), structured RNA sequences called internal ribosome entry sites (IRESs) located in the genomic 5′-untranslated region trigger an alternative cap-independent mechanism of assembling the eukaryotic protein synthesis machinery at the correct start codon (3,4). The HCV IRES drives translation of its genomic RNA, helping the virus to evade the host immune response that suppresses canonical cap-dependent translation (4). However, the specific molecular events that accompany preinitiation complex formation in an mRNA cap- or IRES-dependent manner remain poorly understood.
Translation initiation factor eIF3 is central to the assembly of competent translation preinitiation complexes (3). Evidence now indicates that eIF3 stimulates most of the reactions in cap-dependent translation initiation, including initiation factor binding to the 40S subunit in the 43S preinitiation complex, mRNA recruitment to the 43S preinitiation complex, mRNA scanning for the start codon (2) and release of eIF2–guanosine diphosphate (GDP) after start codon recognition (5). Human eIF3 also extends the length of the mRNA-binding channel of the 40S subunit in initiation complexes (6). In humans, eIF3 is important for cell homeostasis, and misregulation of eIF3 subunit expression correlates with cancer progression (7). In addition, human eIF3 plays an important role in translation initiation mediated by the HCV IRES. Defined secondary and tertiary structural elements of the HCV IRES (8) bind specifically to eIF3 and the 40S ribosomal subunit (4,9–11). These structural interactions position the genomic RNA at the correct start codon without the need for the 5′-cap binding translation initiation machinery and mRNA scanning (12,13). HCV IRES–eIF3 association has been proposed to substitute for translation initiation factor eIF4G, which is required for cap-dependent translation (14). However, the molecular consequences of HCV IRES interactions with eIF3 are unknown.
Despite the biological significance of eIF3, its molecular contributions to translation initiation remain unclear owing to the scarcity of structural information. Human eIF3 is an 800 kDa complex containing 13 proteins (eIF3a–eIF3m) (15). The cryo-electron microscopic reconstruction of endogenously purified human eIF3 revealed a five-lobed structure with anthropomorphic features that binds to the platform of the 40S ribosomal subunit (14). Remarkably, our recent 3D reconstruction of a bacterially reconstituted octameric core complex of eIF3 showed it to be structurally similar to the larger native eIF3, likely due to the inherent flexibility of large regions within the complex beyond the central core (16).
In the case of HCV IRES-mediated translation, cryo-EM reconstructions of the binary HCV IRES–eIF3 complex revealed extensive interactions between the IRES RNA and eIF3. However, the specific regions of contact could not be identified definitively (14). In cross-linking experiments, eIF3 subunits a, b, d and f were shown to be located near to or contacting the HCV IRES (10,17). The functional importance of these interactions remains obscure owing to the paucity of structural information about them.
Here we used EM, biochemical and bioinformatics approaches to determine the critical domains responsible for eIF3 binding to the HCV IRES and 40S ribosomal subunit, and for the formation of translation initiation complexes. Together, these results provide fundamental mechanistic insights into eIF3 function during HCV IRES-mediated translation, identify possible new targets for HCV therapeutic intervention and reveal motifs in eIF3 that may be important for the control of translation initiation on cellular mRNAs.
All the transfer and coexpression vectors used for generating eIF3 subcomplexes are listed in Supplementary Tables S1 and S2. Quikchange mutagenesis (Agilent Technologies) was used to mutate the predicted helix-loop-helix (HLH) motifs in subunits a and c. Expression and purification of eIF3 subcomplexes are described in the Supplementary Materials and Methods.
Human eIF3 and 40S ribosomal subunits, the HCV IRES and IIIabc domains were purified as described (18). The sequence of the IRES IIIabc domain is that used previously (18). The IRES used in rabbit reticulocyte lysate (RRL) initiation complex formation includes nucleotides 39–352 of the HCV subtype 1 b genomic RNA (16). The HCV IRES and IIIabc domain were labeled at the 3′ end for use in native gel mobility shift assays, as described (16).
Binding reactions to the HCV IRES or IIIabc domain contained reconstituted human eIF3 (or eIF3 subcomplex), labeled HCV IRES or IIIabc domain at 20 nM and THEMK buffer (34 mM Tris, 66 mM HEPES, 0.1 mM EDTA, 2.5 mM MgCl2, 75 mM KCl, pH 7.8). Reactions were carried out at 25°C for 15 min and then resolved by native 1% agarose gels at 4°C for 45 min in the same buffer (18). Gels were visualized by fluorescence imaging using a Typhoon Scanner (Amersham Biosciences). The excitation and emission wavelengths to detect the HCV IRES were 494 and 518 nm, respectively. Binding reactions with the 40S ribosomal subunit were carried out at 25°C for 15 min in ribosome binding buffer (20 mM Hepes, pH 7.5, 100 mM KCl, 2.5 mM MgCl2). Native agarose gel electrophoresis used to monitor binding was carried out at 4°C in THEMK buffer with frequent buffer exchanges.
The full-length IRES within the context of 5′ and 3′ flanking sequences were transcribed, purified and annealed as described in the Supplementary Materials and Methods (19). Folded RNA samples were placed at room temperature for 10 min before reaction with 1-methyl-7-nitroisatoic anhydride (1M7) (20). The folded RNA was incubated with the eIF3 subcomplexes or eIF3 storage buffer at room temperature for 10 min before reaction with 1M7. The eIF3 subcomplexes were in 1.5-fold excess over IRES RNA, at concentrations >250-fold the Kd of the interaction to ensure binding (9). SHAPE modification, reverse transcription, capillary electrophoresis and data processing and normalization were performed as previously described (19) with minor modifications (Supplementary Materials and Methods).
The in vitro translation assays were carried out as described (16). To check the formation of preinitiation complexes, 2 mM 5′-Guanylyl imidodiphosphate (GMPPNP) (Sigma) was added to the reactions. The reactions were run on 10–25% sucrose gradients as described (16). To check the formation of 80S complexes, 100 µg/ml cycloheximide (Sigma) was added to the reactions. The reactions were run on 10–40% sucrose gradient as described (16). The DNA oligonucleotide 5′-AACGATCAGAGTAGTGGTATTTCACC-3′ was used as the probe for detecting human 28S rRNA (21). All the western blots and northern blots were carried out as described in (16).
C-flat 400-mesh grids containing 2 µm holes with a spacing of 2 µm (Protochips) were glow-discharged in a nitrogen atmosphere for 45 s using an Edwards Carbon Evaporator. Grids were immediately loaded into a Vitrobot (FEI company) whose climate chamber had equilibrated to 4°C and 100% humidity. Four microliters of aliquots of the purified sample were placed onto the grids. The grids were blotted for 3.5 s at an offset of −1 mm, and plunged into liquid ethane. Because certain views of the eIF3 core particle were missing for the samples using holey grids, the purified complex was also frozen using grids onto which a thin carbon film was floated. Data were acquired on a Gatan 4 K × 4 K camera using a Tecnai F20 electron microscope operating at 120 keV at a nominal magnification of 100 000× using a dose of ~20 e−/Å2 and a defocus range from −1.2 to −2.5 µm.
Data processing (2D) was performed using programs and utilities contained within the Appion processing environment (22). After some initial preprocessing, particle image stacks were subjected to several rounds of reference-free 2D classification by iterative multivariate statistical analysis and multireference alignment using IMAGIC (23). The cryo-EM reconstructions were performed using either an iterative projection matching approach in EMAN2 (24,25) or a maximum-likelihood approach (26) within the XMIPP package (27). We used as an initial reference the negative-stain structure of the human eIF3 core we previously solved (16), filtered to 60 Å resolution. More details can be found in the Supplementary Materials and Methods.
A prior cryo-EM reconstruction showed the HCV IRES extending across eIF3, from the left arm to the right leg (11,14). In that study, the IRES IIIabc domain was modeled near the right leg of eIF3. However, independent corroborating evidence for the positioning of domains within the HCV IRES in the ternary complex with eIF3 and the 40S ribosomal subunit is lacking. The octameric core of eIF3 binds to the HCV IRES or IIIabc domain RNA (Figure 1A) with an affinity similar to that of native eIF3 (16). We therefore collected cryo-EM data for the eIF3 octameric core bound to the IRES IIIabc domain to better orient the IRES RNA when bound to eIF3 (Supplementary Materials and Methods). Although conformational flexibility and substoichometric binding (Supplementary Figure S1) hampered 3D reconstructions of the eIF3 core/IIIabc RNA complex at high resolution, a multimodel refinement approach (26) yielded one 3D reconstruction with clear extra density, when compared with the unbound eIF3 core, which emanates from the left arm of eIF3, the position of the N-terminus of eIF3a (28) (Figure 1A and Supplementary Figure S1C). The presence of this extra density was confirmed in 3D difference maps (Figure 1A) and agreed with the conclusions from reference-free 2D class average analysis (Supplementary Figure S1). Recent EM labeling experiments have shown this region in the eIF3 core structure to be occupied by subunits a and c (28). Interestingly, limited proteolysis and UV cross-linking experiments revealed that eIF3 subunits a* and c* (the N-terminally truncated versions of eIF3a and eIF3c used in the octameric core reconstitution) become protected by binding of the IIIabc domain RNA, and that they form cross-links with 4-thio-U substituted HCV IRES RNA (Supplementary Figure S2A–2C; Supplementary Materials and Methods). Taken together, these results reveal that the IIIabc domain of the HCV IRES is positioned near the left arm of eIF3 where it interacts with eIF3 subunits a* and c*, and not by the right leg (14).
To determine the regions of eIF3 subunits a and c that bind the HCV IRES IIIabc domain, the sequences of these two subunits were analyzed computationally with BindN (29) and Phyre2 (30).These analyses predicted the presence of RNA-binding HLH motifs near amino acid 40 of subunit a and 340 of subunit c, both near the N-termini of the a* and c* truncated subunits used in the eIF3 reconstitution (Figure 1B). The predicted HLH motif in subunit c is highly conserved in all eukaryotic eIF3s, and that in subunit a is highly conserved in metazoans (Figure 1C, Supplementary Figure S2D). Notably, an N-terminal deletion in subunit a* (residues 5–46) that includes its predicted HLH motif reduced the apparent binding of the eIF3 octamer to the HCV IRES over 30-fold (Figure 2A). Further deletion of amino acids 302–343 from subunit c*, which includes its predicted HLH motif, reduced the apparent affinity of the eIF3 octameric core for the IRES over 100-fold compared with wild-type octamer (Figure 2A), indicating that amino acids 302–343 in subunit c* also contribute to HCV IRES binding by the eIF3 octamer. Importantly, truncations of the predicted HLH motif in eIF3a, or of both HLH motifs (Figure 1B), did not interfere with octamer assembly (Supplementary Figure S3A). Similar binding results were obtained with the HCV IRES IIIabc domain RNA (Supplementary Figure S3B). Taken together, these results indicate that both predicted HLH motifs in eIF3 subunits a and c directly bind the IIIabc domain in the HCV IRES.
We next sought to identify the specific amino acids in eIF3 subunits a and c that are critical for eIF3 binding to the HCV IRES RNA. Both HLH motifs in eIF3 subunits a and c are rich in arginines and lysines in the predicted loop region (Figure 1B). Interestingly, mutations of amino acids 340–343 in subunit c* from RGKK to NGEE or of amino acids 36–39 in subunit a* from KSKK to NSEE, neither of which affected complex assembly (Supplementary Figure S3A) effectively mimicked the reduced affinity of the N-terminal truncations for HCV IRES RNA described above. This was true in the context of fully assembled eIF3 octamer (Figure 2B) and purified c* monomer or a*c* dimer (Figure 2C). Similar results were observed in binding experiments with the HCV IRES IIIabc domain RNA (Supplementary Figure S3C). Thus, the loop residues of both HLH motifs in eIF3 subunits a and c are critical for eIF3 binding to the HCV IRES RNA.
Subunits eIF3b and eIF3d have also been proposed to be involved in eIF3 binding to the HCV IRES (10,17). To test the contribution of subunits beyond eIF3a and eIF3c to HCV IRES binding, we reconstituted eIF3 complexes containing the octameric core plus eIF3d (9-mer), plus eIF3d and eIF3b (10-mer) or with all eIF3 subunits except eIF3j (12-mer) (Supplementary Materials and Methods). As observed for octamer assembly, mutations in subunits a* and c* did not interfere with eIF3 9-12 subunit reconstitutions (Supplementary Figure S3A). Comparisons of the eIF3 octamer with 9-mer and 10-mer assemblies, each incorporating the above mutations in the HLH motifs in subunits a* and/or c*, revealed that subunits b and d contribute only modestly to HCV IRES binding (Figure 2B). In the context of 12-mer eIF3 complexes (octamer plus subunits b, d, g and i), mutations in the c* HLH motif weakened HCV IRES binding modestly, whereas mutations in the HLH domain of subunit a* reduced the apparent affinity of the 12-mer for the IRES by ~20-fold (16) (Figure 2D). These data therefore indicate that subunits a, b, c and d all contribute to eIF3 binding to the HCV IRES, with subunit eIF3a playing the dominant role through its HLH motif.
To identify the regions within the HCV IRES that contact the predicted HLH motifs in eIF3 subunits a* and c*, we used SHAPE (selective 2′-hydroxyl acylation analyzed by primer extension) chemistry to analyze the flexibility of each nucleotide in the HCV IRES in the absence and presence of wild-type eIF3 octamer, 10-mer or with mutations in the predicted HLH motif of subunit c* (19) (Supplementary Materials and Methods). The eIF3 octamer or 10-subunit complex strongly protected the HCV IRES IIIb stem-loop from SHAPE reactivity (Figure 3), consistent with the IIIb region serving as the main binding site for eIF3 (9). The eIF3 octamer with mutations in the HLH motif of subunit eIF3c, which slightly reduced IRES binding (Figure 3), also protected the IIIb stem-loop from SHAPE reactivity, albeit less strongly (Figure 3). Together with the mutational analysis above (Figure 2), these results indicate that the HLH motif of subunit eIF3a makes the most extensive contacts with the HCV IRES by binding the IRES IIIb stem loop. Subunit eIF3c may augment this primary interaction rather than contacting an entirely separate region of the IRES, whereas subunits eIF3b and eIF3d may contribute to additional contacts near the lower portions of domain III and the pseudoknot regions of the IRES (Figure 3).
The Proteasome, COP9, eIF3/Mpr1-Pad1 N terminal (PCI/MPN) octameric core of eIF3 binds the 40S ribosomal subunit with only 2–3-fold weaker affinity compared with native eIF3 (16). Considering the importance of the N-termini of eIF3 subunits a* and c* for HCV IRES binding, as well as their proximity to the 40S subunit (14), we tested whether mutations in the HLH motifs in subunits a and c affect the interaction of eIF3 with the 40S ribosomal subunit. Remarkably, eIF3 octameric core complexes with mutations in the HLH motifs bound the 40S ribosomal subunit with severely reduced affinity (Supplementary Figure S4). The addition of subunits b, d, g and i restored 40S ribosomal subunit binding of eIF3 complexes with mutations in the HLH motifs to some degree (Figure 4). These results indicate that the HLH motifs in eIF3a and eIF3c are important for eIF3 binding to the 40S ribosomal subunit, as well as to the HCV IRES.
Because mutations in the HLH motifs of eIF3 subunits a* and c* impair the association of eIF3 with the HCV IRES and 40S ribosomal subunit individually, we tested whether the mutations would impair the formation of preinitiation complexes in HCV IRES-dependent translation. We first tested the formation of preinitiation complexes on the 40S ribosomal subunit in the presence of reconstituted and glutathione S-transferase (GST)-tagged eIF3 or its mutants in rabbit reticulocyte lysates (RRLs) programmed with an HCV IRES-containing model mRNA (Supplementary Materials and Methods). In RRL reactions with 2 mM of the non-hydrolyzable GTP analogue GMPPNP to stall preinitiation complex formation, wild-type reconstituted eIF3 formed preinitiation complexes containing the 40S ribosomal subunit, eIF2/Met-tRNAi and the HCV IRES-containing mRNA (Figure 5A and B and Supplementary Figure S5B) (16). Whereas mutations in the HLH motif of eIF3c only slightly impaired preinitiation complex formation (Figure 5B, fractions 2 and 3), mutations in the HLH motif of eIF3a drastically reduced the formation of preinitiation complexes containing the HCV IRES mRNA and eIF2/Met-tRNAi (Figure 5B).
We also tested the formation of 80S ribosome complexes stalled in the presence of cycloheximide (32) with wild-type reconstituted eIF3 and the above mutant forms in RRL programmed with the HCV IRES-containing model mRNA (Figure 5C and D, and Supplementary Figure S5C). In RRL reactions containing cycloheximide to stall translation after formation of programmed 80S ribosomes (32), mutation of the HLH motif in eIF3c did not impact the formation of 48S preinitiation complexes (Figure 5D, fractions 2 and 3), but decreased the formation of functional 80S complexes containing initiator tRNA (Figure 5D, fraction 4). Interestingly, defects in the formation of programmed 80S ribosomes due to the mutations in eIF3c* correlated with defects in the association of translation initiation factor eIF5B with preinitiation complexes (Figure 5D). In eIF2-mediated translation initiation, eIF5B is generally required after start codon recognition and eIF2–GDP release, in a second GTP-dependent step that leads to 60S subunit joining (33,34). Consistent with the experiments with GMPPNP, mutation of the HLH motif in eIF3a resulted in even more severe defects, eliminating the ability of reconstituted eIF3 to mediate Met-tRNAi loading into 80S complexes (Figure 5D). Taken together, these results indicate that the HLH motif in eIF3a contributes to eIF2-dependent Met-tRNAi binding to preinitiation complexes, whereas the HLH motif in eIF3c is important for eIF5B association with preinitiation complexes.
Although eIF3 is essential for the function of the HCV IRES in translation initiation (12,35,36) and binds tightly to the HCV IRES in isolation (9,10), the molecular mechanism by which it stimulates initiation from HCV IRES has not been clear. Using cryo-EM data as a guide, we were able to predict the location of HLH RNA-binding motifs in eIF3 subunits a and c, and to show that these HLH motifs make direct contacts with the HCV IRES. Furthermore, these HLH motifs are important for eIF3 binding to the 40S ribosomal subunit. The contacts between the HLH motifs to the HCV IRES and 40S ribosomal subunit are augmented by contributions from subunits b and d, consistent with previous cross-linking results (10,17).
Based on the experiments in translation extracts, the HLH motif in eIF3a serves as a lynchpin in assembling translation initiation complexes on the HCV IRES. The HLH motif in eIF3a contributes to early steps of HCV IRES-mediated translation initiation dependent on eIF2, before start codon selection, whereas the eIF3c HLH motif is important for later events dependent on eIF5B, after start codon selection but before 60S subunit joining (Figure 6). Intriguingly, the HLH motif in eIF3a is highly conserved across metazoans and is present in most eukaryotes with available genome sequences (Figure 1C, Supplementary Figure S2D), suggesting that this motif may play a role in loading of the eIF2–GTP–Met-tRNAi ternary complex into translation preinitiation complexes on cellular mRNAs as well as the HCV genomic RNA. The HLH motif in eIF3c is highly conserved in all eukaryotes (Figure 1C), implying that its role in eIF5B exchange with eIF2 after start codon selection to catalyze 60S ribosomal subunit joining (33,37) may be a nearly universal feature of eukaryotic translation initiation. Interestingly, the conformation of the eIF3 octameric core closely resembles the proteasome lid subcomplex once it is bound to the 26S proteasome holoenzyme, which causes the N-terminus of Rpn5 to adopt an extended conformation, similar to the conformation seen here for eIF3a (38). The a* and c* arms are the most mobile elements in the eIF3 core, and correspond to the eIF3 regions in closest proximity to the mRNA binding cleft of the 40S subunit (14,28). The flexibility of these arms in eIF3 may be important for their function in translation initiation complex assembly. Based on its location in eIF3, the eIF3c HLH motif would be positioned near the head domain of the 40S ribosomal subunit (14,28). Contacts between the eIF3c HLH domain and the head of the 40S subunit, which is thought to adopt a closed conformation after start codon selection (11,39), may serve to stimulate eIF2 exchange with eIF5B after start codon recognition (40).
The present experiments identified a surprisingly localized contact between an HLH motif in eIF3a and the IIIb stem-loop of the HCV IRES RNA that controls the efficiency of IRES-mediated preinitiation complex assembly and functional 80S ribosome formation. The local nature of this contact could serve as a new target for the development of small molecule therapeutics to inhibit HCV IRES-mediated translation initiation. In conditions of cellular stress when active eIF2 pools are depleted, the eIF3c HLH motif may also play an important role in an alternative eIF5B-mediated translation initiation pathway using the HCV IRES (36). In this alternative pathway, translation initiation occurs without the requirement for eIF2. It will be interesting in future experiments to investigate how these two HLH motifs in eIF3 contribute to regulating translation initiation on cellular mRNAs in different physiological contexts.
The Cryo-EM map of the eIF3 PCI/MPN octameric complex and maps used to generate the difference map with the IIIabc RNA have been deposited in the Electron Microscopy Data Bank under accession codes EMD-2166, EMD-2198 and EMD-2199.
Supplementary Data are available at NAR Online: Supplementary Tables 1 and 2, Supplementary Figures 1–5, Supplementary Materials and Methods and Supplementary References [41–52].
National Institutes of Health (NIH) [R56-AI095687 to J.H.D.C.; P50-GM102706 to J.A.D. and J.H.D.C.]; Spanish Ministry of Education through the Programa Nacional de Movilidad de Recursos Humanos del Plan Nacional de I-D+i 2008-2011 (to E.A.-P.). J.A.D. and E.N. are Howard Hughes Medical Institute Investigators. Funding for open access charge: NIH [P50-GM102706].
Conflict of interest statement. None declared.
The authors thank G. Lander for invaluable technical help, P. Grob and T. Houweling for electron microscopy and computational support, respectively, and D. Smith for helpful discussions on the manuscript.