A Conserved Interaction between a C-Terminal Motif in Norovirus VPg and the HEAT-1 Domain of eIF4G Is Essential for Translation Initiation

Abstract

Translation initiation is a critical early step in the replication cycle of the positive-sense, single-stranded RNA genome of noroviruses, a major cause of gastroenteritis in humans. Norovirus RNA, which has neither a 5´ m7G cap nor an internal ribosome entry site (IRES), adopts an unusual mechanism to initiate protein synthesis that relies on interactions between the VPg protein covalently attached to the 5´-end of the viral RNA and eukaryotic initiation factors (eIFs) in the host cell. For murine norovirus (MNV) we previously showed that VPg binds to the middle fragment of eIF4G (4GM; residues 652-1132). Here we have used pull-down assays, fluorescence anisotropy, and isothermal titration calorimetry (ITC) to demonstrate that a stretch of ~20 amino acids at the C terminus of MNV VPg mediates direct and specific binding to the HEAT-1 domain within the 4GM fragment of eIF4G. Our analysis further reveals that the MNV C terminus binds to eIF4G HEAT-1 via a motif that is conserved in all known noroviruses. Fine mutagenic mapping suggests that the MNV VPg C terminus may interact with eIF4G in a helical conformation. NMR spectroscopy was used to define the VPg binding site on eIF4G HEAT-1, which was confirmed by mutagenesis and binding assays. We have found that this site is non-overlapping with the binding site for eIF4A on eIF4G HEAT-1 by demonstrating that norovirus VPg can form ternary VPg-eIF4G-eIF4A complexes. The functional significance of the VPg-eIF4G interaction was shown by the ability of fusion proteins containing the C-terminal peptide of MNV VPg to inhibit in vitro translation of norovirus RNA but not cap- or IRES-dependent translation. These observations define important structural details of a functional interaction between norovirus VPg and eIF4G and reveal a binding interface that might be exploited as a target for antiviral therapy.

Fig 1. MNV VPg interacts with eIF4GI via its HEAT-1 domain.

(A) Schematic representation of eIF4GI (NCBI accession AAM69365.1), one of two eIF4G paralogues expressed in humans, and the paralog DAP5 (NCBI accession NP_001036024.3). Positions of domains that interact with other proteins of the translation initiation machinery are indicated, as are the cleavage sites of FMDV L protease and Rhinovirus 2A protease. The principal eIF4G fragments that were sub-cloned for use in this study are also indicated. (B) SDS PAGE analysis of glutathione affinity pull-down assays that were performed to map the locus of MNV VPg binding. GST-fusions of various eIF4GI fragments (shown in panel A) were used as bait and His-tagged-MNV VPg(1–124) as prey. Left panel: protein mixtures applied to the glutathione-sepharose 4B beads (lanes 1–6). Right panel: proteins eluted with 10 mM glutathione (lanes 7–12). (C) SDS PAGE analysis of cobalt affinity pull-down assays to confirm that binding of MNV VPg occurs primarily through the eIF4G HEAT-1 domain. GST-fusions of various eIF4GI fragments with C-terminal His-tags were used as bait and untagged MNV VPg as prey. Lanes 1–4: protein mixtures applied to the cobalt resin; lanes 5–8: proteins eluted with 250 mM imidazole. (D) SDS PAGE analysis of cobalt affinity pull-down assays performed to confirm the eIF4G HEAT-1 domain as the locus of MNV VPg binding. His-tagged-MNV VPg(1–124) or His-tagged-FCV VPg(1–111) were used as bait proteins and GST-fusions of various eIF4GI fragments as prey. Left panel: protein mixtures applied to the cobalt resin (Lanes 1–6). Right panel: proteins eluted with 250 mM imidazole (lanes 7–12).

Fig 2. MNV VPg interacts with eIF4GI and eIF4GII HEAT-1 domains via the C-terminal residues 104–124.

(A) SDS PAGE analysis of cobalt affinity pull-down assay using a GST-eIF4GI HEAT-1 construct with a C-terminal His-tag as bait and either untagged MNV VPg(1–124) or MNV VPg(1–85) as prey. Protein mixtures are shown in lanes 1 and 2 (blue labels); bound proteins eluted with 250 mM imidazole are in lanes 3 and 4 (red labels). (B) Sequence alignment showing the location of amino acid substitutions introduced into the C terminus of His-tagged MNV VPg. (C) SDS PAGE analysis of cobalt affinity assays using His-tagged MNV mutants (panel B) as bait and either GST-eIF4GI-HEAT-1 (top panel) or GST-eIF4GII-HEAT-1 as prey. Lanes 1–2: input proteins; lanes 3–11: eluted proteins. (D) SDS PAGE analysis of cobalt affinity pull-down assay using either His-tagged GFP-VPg(1–124) wild-type, GFP-VPg(1–124) F123A, a His-tagged GFP-VPg-NS6 fusion (containing the inactivating C139A mutation of the protease active site Cys (NS6 numbering)) or a His-tagged GFP-VPg-NS6-4´ fusion (containing just the first 4 amino acids of NS6) as bait, and untagged eIF4GII HEAT-1 as prey. Lanes 1–5: protein mixes; lanes 6–10: proteins eluted with 250 mM imidazole. (E) Structure of the VPg-NS6 junction showing that the N-terminus of NS6 is tightly folded into the body of the protease. Structure shown is the crystal structure of VPg-NS6 from human Norwalk virus, which is very similar to the structure of MNV NS6 [40, 41]. Note that due to disorder in the crystal, only the C terminus of VPg is visible [41].

Fig 3. MNV VPg 104–124 interacts with the HEAT-1 domains of eIF4GI, eIF4GII and DAP5 with low micromolar affinity.

(A-B) FITC-labelled peptides– MNV VPg(104–124) and MNV VPg(108–124)–were used in fluorescence anisotropy binding assays with unlabelled HEAT-1 domains of eIF4GI (748–993), eIF4GII (745–1003) and DAP5 (61–323) in order to measure the affinity of the interaction. The normalised change in fluorescence anisotropy (relative to a no-protein control), ΔFA, is plotted against protein concentration. Error bars in ΔFP indicate the standard deviation of 5 (eIF4GI) or 10 (eIF4GII) independent measurements. The solid lines indicate the fit to a single-site binding model. (A) Comparison of the binding of MNV VPg(104–124) and MNV VPg(108–124) to eIF4GI and eIF4GII HEAT-1 domains. (B) Binding of MNV VPg(104–124) to the HEAT-1 domain of DAP5. The fit is calculated with the fluorescence anisotropy data from three independent experiments (all included in the graph). (C) ITC experiments in which unlabelled eIF4GII (745–1003) was titrated into MNV VPg(1–124). Top panel: raw data obtained for a representative experiment from 20 injections (firstly with a volume of 0.5 μL followed by 19 injections of 2 μL of eIF4GII HEAT-1). Bottom panel: the integrated data with a best-fit curve for the representative experiment generated for a single-site binding model using the Origin software package.

Fig 4. The C-termini of Noroviral VPg proteins contain a conserved sequence motif that binds to eIF4G HEAT-1.

(A) Top: Amino acid sequence alignments of representative sequences of the C-termini of all 6 genogroups of Norovirus (GI-GIV). Second from top: Alignment of the C-termini of Human Astrovirus 4 (HuAST4) VPg with MNV (GV) VPg. Third from top: Alignment of the C-termini of FCV VPg with MNV (GV) VPg. Bottom: Alignment of the C terminus of MNV (GV) VPg with Rice Yellow Mottle Virus (RYMV) VPg. The representative strains used in the alignments are GI—Hu/GI/Norwalk/1968/US, (NCBI accession AAC64602), GII—Lordsdale virus Hu/GII/Lordsdale/1993/UK (NCBI accession P54634), GIII—Bo/GIII/B309/2003/BEL (NCBI accession ACJ04905.1), GIV—Hu/GIV.1/LakeMacquarie/NSW268O (NCBI accession AFJ21375), GV—Mu/NoV/GV/MNV1/2002/USA (NCBI accession ABU55564.1), GVI—dog/GVI.1/HKU_Ca035F/2007/HKG (NCBI accession FJ692501), FCV—F9 strain (NCBI accession P27409.1), HuAst4—Human astrovirus 4 (NCBI accession Q3ZN07), Rice yellow mottle virus isolate CI4 (NCBI accession NC_001575). Sequence alignments were performed by ClustalW [6] and BioEdit (http://www.mbio.ncsu.edu/BioEdit/bioedit.html). ^#Denotes sequences that were used in pull-down assays (see panel B). (B) The indicated VPg sequences in panel A were fused to the C terminus of GST for use as prey in cobalt affinity pull-down assays in which His-tagged eIF4GI HEAT-1 was used as bait, and analysed by SDS PAGE. Lanes 1–3: input samples of some of the proteins used in pull-down experiments (black labels); lanes 4–11: eluted proteins (red labels; bands at ~27 kDa indicate GST-VPg C-terminal constructs that bound to His-eIF4G HEAT-1); lanes 12–19 (blue labels)–protein mixes used in the pull-down experiments. (C, D) Mutational analysis of the C-terminal sequences of MNV VPg were performed using GST-MNV VPg C-terminal fusions in the same way as the experiment presented in panel B. Labelling is colour-coded as in panel B. (E) Quantification of the pull-down results shown in panels C and D. (F) Helical wheel representation of MNV VPg 102–124 generated using http://heliquest.ipmc.cnrs.fr/ [44]. Residues mutated for the pull-down assays are indicated by bold-face labels and colour-coded by the effect of the substitution on binding to eIF4G HEAT-1: red–no binding; black—binding at or near wild-type levels.

Fig 5. NMR chemical shift mapping reveals the binding site for MNV VPg on the eIF4GI HEAT-1 domain.

(A) Superposition of the ¹H¹⁵N TROSY HSQC spectra of ¹⁵N-labelled eIF4GI HEAT-1 (748–993) obtained in the absence (black) and presence (green) of 0.35 molar equivalents of unlabelled MNV VPg(104–124) peptide. (¹H¹⁵N TROSY NMR spectra for all of the points in the titration are given in S3 Fig). The residues for which the intensity of amide ¹H¹⁵N signals exhibit the greatest reductions as the MNV VPg(104–124) peptide concentration is increased are indicated by their residue number. The assignments were taken from the Bio Magnetic Resonance Database (BMRB id 18738). (B) Plot of the ratio of the absolute values of non-overlapping peak volumes for assigned residues obtained at 0 and 0.35 molar equivalents of the MNV VPg(104–124) peptide. High values of the ratios indicate the residues in eIF4GI HEAT-1 most affected by MNV VPg binding. The red dashed line indicates the expected volume ratio (R_v) for unperturbed residues: 0.76, based on the relative number of scans performed in each HSQC experiment. Points above the green dashed line indicate residues for which the peak volume ratio was at least six times higher than the baseline for unperturbed amides. (C) Model of the complex of eIF4GI HEAT-1 (grey, surface representation) and eIF4A (cyan, cartoon representation) indicating the predicted location of the VPg binding site. Surface residues in eIF4GI HEAT-1 that exhibit the greatest changes in R_v in the presence of the MNV VPg(104–124) peptide are coloured yellow. The model of eIF4GI HEAT-1 was generated using SWISS-MODEL [46]; the complex was created by superposing this model on the eIF4G HEAT-1 component of the yeast eIF4G-eIF4A co-crystal structure [47]. (D) SDS PAGE analysis of pull-down assays to test the effect of mutations in the putative VPg-binding site on eIF4G HEAT-1 on binding to the viral protein. His-tagged MNV VPg(1–124) was used as bait; eIF4GI HEAT-1 wild-type or mutant proteins were used as prey, and the pull-down buffer contained 150 mM NaCl (top panel) or 300 mM NaCl (bottom panel). Lanes 1–8 (blue labels): input protein mixtures; lanes 9–16 (red labels): eluted proteins. (E) Graphical representation of the results of the cobalt affinity pull-down assays performed in the presence of 150 or 300 mM NaCl, plotted as the ratio of the optical densities of the eIF4GI HEAT-1 band to that of the His-MNV VPg band in eluted fractions (lanes 9–16 in panel D). Band densities were quantified using ImageJ (http://imagej.nih.gov/ij/) and the ratios were normalised to the wild-type control.

Fig 6. MNV VPg forms a ternary complex with eIF4G and eIF4A.

(A) SEC profiles monitored by differential refractive index (which is proportional to protein concentration) for eIF4A, eIF4G HEAT-1 and MNV VPg as well as for the binary complexes formed by mixing approximately equimolar quantities of eIF4A and eIF4G HEAT-1 (4A-4G), eIF4G HEAT-1 and MNV VPg (4G-VPg), and the ternary complex obtained from eIF4A, eIF4G HEAT-1, and MNV VPg (4A-4G-VPg). SDS PAGE analysis of peak fractions of the binary and ternary complexes obtained in the SEC experiments are also shown in A. (B) SEC-MALLS analysis of the molar mass distributions of the binary and ternary complexes plotted against the SEC profiles shown in A.

Fig 7. GST-MNV VPg 102–124 inhibits MNV VPg-mediated translation in vitro but not cap-dependent or IRES-dependent translation.

In vitro translation reactions programmed with VPg-linked MNV RNA were performed in the presence of increasing concentrations of GST-MNV VPg(102–124) WT protein or the GST-MNV VPg(102–124) F123A mutant that binds very poorly to eIF4G. Protein synthesis was monitored by autoradiography of SDS PAGE analysis of incorporation of ³⁵S-methionine in translation reactions. (A) Top panel: Effect of exogenous GST-MNV VPg(102–124) proteins on translation from VPg-linked MNV RNA; bottom panel: quantitative analysis of the level of ³⁵S-methionine incorporation. This result was confirmed by independent replication–see S5A Fig. (B) Top panel: Effect of exogenous GST-MNV VPg(102–124) proteins on translation from capped bi-cistronic mRNA constructs containing the FMDV IRES between the first (CAT) and second (Luc) cistrons; bottom panel: quantitative analysis of the level of ³⁵S-methionine incorporation.

Fig 8. mCherry-MNV VPg(102–124) co-immunoprecipitates with translation initiation proteins.

(A) RFP-Trap immunoprecipitation of lysates of BV2 cells expressing mCherry, mCherry-MNV VPg(102–124) WT or mCherry-MNV VPg(102–124) F123A. Pre-purification lysates (Input), purified fractions (IP) and unbound fractions (Flow through) were analysed by SDS PAGE and western blotting. (B) Time course of BV2 infection (MOI 0.01 TCID₅₀ units/cell) with MNV1 in the presence of mCherry-VPg proteins. Prior to infection BV2 cells were transduced with lentiviruses expressing mCherry, mCherry-MNV VPg(102–124) WT or mCherry-MNV VPg(102–124) F123A. The progress of infection was monitored by measuring the number of viral cDNA copies generated from quantitative RT-PCR analysis of whole-cell RNA. (C) Fluorescence microscopy analysis of cell penetration of the FITC-TAT-MNV VPg(102–124) peptides. The images shown are merged images of DAPI stained nuclear DNA (blue) and wild-type or F123A versions of the cell penetrating peptides (green). (D) Time course of BV2 infection (MOI 0.01 TCID₅₀ units/cell) with MNV1 following pre-treatment for 150 minutes with 100 μM of wild-type or F123A versions of the cell penetrating peptides prior to infection. The progress of infection was monitored by RT-PCR analysis (as in panel B).