Viral and Cellular Proteins Containing FGDF Motifs Bind G3BP to Block Stress Granule Formation

The C-terminal L/ITFGDFD repeat motifs, which constitute the G3BP-binding site of nsP3, are well conserved in the Old World alphaviruses [30] with particularly strong conservation of the phenylalanine residues at positions 3 and 6 of the seven residue motif and of the glycine residue at position 4. In order to identify essential residues in the SFV sequence for G3BP binding, we constructed mutants in which each residue from T2 to D7 in both repeats was exchanged for alanine (Fig. 1A). These were constructed in the context of the pEGFP-nsP3-31 construct, containing amino acids 447–477 of SFV nsP3 fused to the C-terminus of EGFP. The empty pEGFP-C1 vector, encoding EGFP with a C-terminal extension of 21 residues was used as a control. We previously demonstrated that EGFP-nsP3-31 and an analogous EGFP fusion protein containing amino acids 475–523 of the CHIKV nsP3 efficiently bind G3BP in transient transfection experiments [27,28]. BHK cells were transfected with vectors encoding EGFP alone, EGFP-nsP3-31 or each of the alanine mutants. Lysates were immunoprecipitated with G3BP-1 antibody and immunoblotted with antibodies against G3BP-1, GFP or actin (Fig. 1B, upper panels). As expected, EGFP-nsP3-31, containing the wild type (wt) nsP3 sequence, efficiently bound G3BP-1, but EGFP alone did not. The construct carrying a mutation of the threonine residue T2 to an alanine exhibited weak but readily detectable binding to G3BP. However, mutation of either of the residues F3, G4 or F6 completely disrupted binding. Finally, while mutation of the aspartate residues at D5 to alanine substantially reduced binding, mutation of D7 had no effect. Lysates were also immunoprecipitated with anti-GFP and immunoblotted for G3BP-1, GFP or actin (Fig. 1B, middle panels), with similar results. Furthermore, GFP-immunoprecipitates were immunoblotted for G2BP-2 (S1 Fig.). These results show that both alternatively-spliced isoforms of G3BP-2 (a and b) interact with EGFP-31 constructs in a similar manner to G3BP-1. We conclude from these results that residues F3, G4 and F6 are essential for G3BP-binding and that T2 and D5 contribute significantly.

Previously we have shown that in Old World alphavirus-infected cells, the C-terminal repeat sequences of nsP3 bind and sequester both G3BP-1 and G3BP-2 into foci of viral protein accumulation including cytopathic vacuoles (CPV) [27,28,31]. In order to test the effects of the non-G3BP-binding mutations on the life cycle of SFV, we mutated both nsP3 residues F451 and F468 to alanines in the infectious clone of SFV (corresponding to the mutation F3A in the two (L/I)TFGDFD motifs, hereafter referred to as SFV-F3A). Previously, we showed that SFV-Δ789, lacking nsP3 residues 449–472, including the G3BP-binding sites displayed slower processing of the P34 precursor [28]. To determine if SFV-F3A also displays slow processing phenotype, we infected BHK cells with wt SFV, SFV-F3A or SFV-Δ789 and compared levels of P34. Lysates from WT SFV or SFV-F3A-infected cells did not contain detectable levels of P34 (S2A Fig.), and we concluded that the F3A mutations do not result in defective P34 processing. To demonstrate that nsP3-F3A does not bind G3BP-1 in the context of a viral infection, we infected BHK cells with wt SFV or SFV-F3A at MOI 10. Lysates were immunoprecipitated with G3BP-1 antibody and analyzed by immunoblotting for nsP3, G3BP-1 and actin. As expected, nsP3 from wt SFV-infected cells coprecipitated with G3BP-1, but nsP3 F3A did not (Fig. 2A). To determine whether SFV-F3A nsP3 colocalized with G3BP, cells were infected at low MOI with wt SFV or SFV-F3A. Cells were fixed and stained for SFV nsP3, G3BP-1 and TIA-1, another SG marker. As expected, wt SFV nsP3 colocalized well with G3BP-1, with very little colocalization of G3BP-1 with TIA-1 (Fig. 2B). However, nsP3 in SFV-F3A infected cells did not colocalize with G3BP-1, confirming immunoprecipitation data indicating that nsP3 F3A does not interact with G3BP-1. Instead, SFV-F3A infected cells often contained foci of G3BP-1 and TIA-1 co-staining, suggesting longer persistence of SGs in cells infected with the SFV-F3A mutant.

Previously we demonstrated that at late times in SFV infection, when most of the infection-induced SG have been disassembled, SGs cannot be re-induced with phospho-eIF2α- or eIF4A-dependent stress inducers (sodium arsenite or pateamine A (Pat A), respectively) because G3BP is sequestered by nsP3 [28,29]. We next tested whether cells infected with wt SFV or SFV-F3A could mount an SG response to an exogenous stress inducer. Cells were infected with wt SFV or SFV-F3A and stressed with Pat A at 7 hpi for 1 h before fixation and staining for TIA-1 to detect SGs and nsP3 to detect SFV infection (Fig. 2C). Pat A was used since sodium arsenite has little effect in SFV-infected cells as these already display sustained eIF2α phosphorylation [28]. Approximately 100% of mock-infected cells responded to Pat A by forming SGs. After infection with wt SFV, 5.2% of cells had SGs, and this number was not significantly altered by Pat A treatment, confirming that wt SFV–infected cells cannot mount a stress response to secondary stress signals. SFV-F3A–infected cells had a significantly higher number of SG-positive cells than wt SFV–infected cultures (18.4%), and this proportion was increased to 49.2% by the Pat A treatment, indicating that in the absence of nsP3/G3BP interaction, infection-induced SGs persist longer and additional SGs can be stress-induced in SFV-F3A-infected cells. To determine whether the specific ablation of G3BP recruitment affects viral replication, we performed single-step and multi-step growth curves comparing wt SFV and SFV-F3A in MEFs. These data revealed that SFV-F3A propagated to titers between 1.5 and 2 orders of magnitude lower that wt SFV after both low and high MOI infection (Fig. 2D) indicating that blocking the nsP3/G3BP interaction by point mutation of the N-terminal phenylalanine residue in both FGDF motifs attenuates viral infection. Single and multi-step growth curve experiments were also performed in BHK cells with similar results (S2B Fig.). Finally, to address whether the F3A mutations in SFV-nsP3 affect other functions than SG inhibition by G3BP sequestration, we performed viral growth curves in eIF2α-AA MEF cells [32]. These cells contain a mutation at the phosphorylation site (S51A) in eIF2α gene, such that eIF2α cannot be phosphorylated and SGs are not induced. SFV-F3A replication was very similar to WT SFV in those cells (Fig. 2E). This result strongly suggests that the attenuation of SFV-F3A in WT cells is due to the action of a G3BP-dependent process downstream of eIF2α phosphorylation, most likely SG formation. Hence, inhibition of SG formation (in response to infection-induced eIF2α phosphorylation) is probably the primary function of the FGDF motifs in Old World alphavirus nsP3.

We have previously demonstrated that at late times in infection with SFV, a large proportion of cellular G3BP is recruited to sites of viral protein accumulation [28]. We wondered whether this interaction displaces USP10, a cellular binding partner of G3BP that also binds to the NTF2-like domain [8,25]. When we determined the localization of USP10 in SFV-infected MEFs, we did not detect any significant overlap with G3BP (S3 Fig.). This suggests that USP10 is excluded from the complex with G3BP by the nsP3 interaction, possibly by competition for the same site on G3BP. Recently, the G3BP-binding domain of USP10 was shown to be located within the N-terminal 76 amino acid residues [16]. Having shown that the G3BP-binding motif of nsP3 consists of two FGDF motifs, we asked whether a similar motif exists in the G3BP-binding region of USP10. When we aligned the sequences of the N-terminal 80 residues of USP10 from several species (Fig. 3A), we identified several well conserved phenylalanine residues (F10, F13, F18, F21, F22 and F59 in the human sequence), as well as a conserved FG(D/E)F motif at residues 10–13. We observed particularly high species conservation in the N-terminal 40 residues. To determine whether this USP10 region binds G3BP, we fused the N-terminal 40 residues of human USP10 to EGFP (EGFP-USP10_1–40). BHK cells were transfected with vectors encoding EGFP alone or EGFP-USP10_1–40. Lysates were immunoprecipitated with anti-GFP and immunoblotted with sera to G3BP-1, G3BP-2, GFP and actin (Fig. 3B, left panels). Indeed, both G3BP proteins efficiently coprecipitated with EGFP-USP10_1–40 but not with EGFP alone. Lysates were also immunoprecipitated with G3BP-1 antibody to verify reciprocal coimmunoprecipitation of EGFP-USP10_1–40 (Fig. 3B, middle panels). To determine whether this binding is dependent on the FGDF motif at residues 10–13, a panel of mutated EGFP-USP10_1–40 variants was created in which each residue from F10 to F13 was substituted to alanine. Cells were transfected with vectors encoding wt EGFP, EGFP-USP10_1–40 or each variant. Lysates were immunoprecipitated with anti-GFP and immunoblotted with sera to G3BP-1, G3BP-2, GFP and actin (Fig. 3C). Alanine substitution of F10, G11 or F13 completely disrupted binding, while the D12A mutation allowed weak binding to both G3BP-1 and G3BP-2. Similar results were obtained after G3BP-1 immunoprecipitation (S4 Fig.). These results are remarkably similar to those obtained following mutagenesis of the G3BP-binding domain of SFV nsP3 (Fig. 1), suggesting that both proteins indeed bind G3BP via their FGDF-motifs.

To test if the overexpression and binding of full-length USP10 to G3BP affects the SG response, BHK cells were transfected with EGFP alone, EGFP-USP10 wt or EGFP-USP10-F10A. Transfected cells were mock stressed or stressed with sodium arsenite for 1 h before fixation and staining for G3BP-1 and TIA-1 to detect SGs. Approximately 97% of the mock transfected cells and 71% of the EGFP transfected cells had SGs upon sodium arsenite treatment (Fig. 3D). Overexpression of wt EGFP-USP10 efficiently blocked the ability of the cells to form SGs (just 1% of cells showed SGs), whereas 59% cells transfected with the G3BP-nonbinding mutant EGFP-USP10-F10A responded to sodium arsenite stress with SGs. Representative images are presented in S5 Fig.. Taken together, the results in Fig. 3 show that USP10 acts as a negative regulator of G3BP-dependent SG formation and that this regulation is dependent on G3BP binding by the FGDF motif at residues 10–13.

In contrast to USP10, nsP3 contains two FGDF motifs (residues 451–454 and 468–471). We therefore hypothesized that a minimal G3BP-binding motif could consist of only one FGDF motif flanked by surrounding residues, which would provide enough space between the two motifs to allow nsP3 to bind two G3BP molecules. Based on our finding that both G3BP-1 and G3BP-2 bind FGDF motifs, we further hypothesized that if EGFP-nsP3-31 binds two molecules of G3BP, it would bind any combination of G3BP-1 and -2, with frequencies depending on the relative abundance of the two proteins and potential differences in affinity. Therefore, G3BP-2 should be detectable in nsP3 complexes immunoprecipitated with anti-G3BP-1 and vice versa, but USP10 should form a complex with only one molecule of either G3BP-1 or -2 (Fig. 4A). To test this, BHK cells were transfected with vectors encoding EGFP, EGFP-USP10_1–40 or EGFP-nsP3-31. When lysates were immunoprecipitated with anti-GFP, both G3BP-1 and G3BP-2 were detected in complex with EGFP-nsP3-31 and EGFP-USP10_1–40 (Fig. 4B, left panels). However, when G3BP-1 was immunoprecipitated, G3BP-2 could be detected in EGFP-nsP3-31, but not in EGFP-USP10_1–40 bound complexes (Fig. 4B, middle panels). This suggests that the nsP3 sequence, containing two FGDF motifs, can form a ternary complex with both G3BP-1 and G3BP-2 simultaneously, while the USP10 sequence, containing only one FGDF motif, can bind either G3BP-1 or G3BP-2, but not both.

Next, we endeavored to confirm the 2:1 stoichiometry of nsP3:G3BP using biophysical in vitro measurements. We employed isothermal titration calorimetry (ITC) to determine the affinity, stoichiometry and thermodynamic signature of the interaction between the purified NTF2-like domain of human G3BP-1, expressed in E. coli (G3BP-NTF2), and a peptide spanning residues 449–473 of SFV nsP3 (nsP3-25), containing two FGDF motifs. Injection of nsP3-25 into a G3BP-NTF2 protein solution resulted in concentration-dependent exothermic heat changes (Fig. 4C, upper panel). The binding isotherm of the integrated heat changes was fitted using a simple one-to-one binding model, yielding an affinity value of 7 μM as well as 2.4 G3BP binding sites per nsP3-25 peptide molecule (Fig. 4C, lower panel). This indicates that one peptide efficiently binds two G3BP molecules. The interaction affinity was 16-fold stronger than the previously determined affinity value of 115 μM for the interaction between a similar G3BP-NTF2 fragment and a DSGFSFGSK peptide [33]. Injection of a control peptide in which both FGDF motifs were exchanged for AGDA (nsP3-25-mut) only produced injection-related heat fluctuations at baseline level (Fig. 4D).

Engagement of two G3BP molecules with each nsP3-25 peptide was further supported by analytical size exclusion chromatography (SEC) analysis. It is known from crystal structures and SEC analysis that the NTF2-like domains of rat and human G3BP form homo-dimers [33]. G3BP-NTF2 alone was eluted with an apparent molecular weight of 30 kDa, slightly lower than the calculated molecular weight of the dimer (36 kDa), while pre-incubation of G3BP with a ten-fold molar excess of nsP3-25 peptide shifted the elution peak, corresponding to the formation of a 60 kDa complex (Fig. 4E). Taking into account the 2:1 ratio for the G3BP-NTF2:nsP3-25 interaction as derived from ITC (Fig. 4C), we suggest that the formed 60 kDa complex comprised four G3BP-NTF2 molecules (two dimers) and two nsP3-25 peptides. There was no evidence for higher-order oligomers.

The crystal structures of the NTF2-like domain of G3BP-1 in its apo form and complexed with the nucleoporin (Nup)-derived peptide DSGFSFGSK have recently been determined [33], revealing that the ligand is bound in an extended conformation within a long and deep groove on the surface of G3BP (S6A Fig.). The peptide-binding site is amphipathic; while the base of the groove is highly hydrophobic, both walls lining the cleft of the protein and the positively charged N-terminus are polar with several basic residues (R32, K5, K123). Both phenylalanine residues of the peptide protrude towards well-defined pockets within the hydrophobic groove. While the side chain of residue F4 of the Nup-derived peptide is buried in a deep pocket formed by the G3BP-1 residues F15, F33 and F124, the side chain of residue F6 is localized in a shallower pocket formed by the G3BP-1 residues F124, V11 and L10 [33].

Our binding and mutagenesis studies reveal that G3BP-1 binds approximately 16 times stronger to FGDF-containing peptides compared to the previously described interaction with the DSGFSFGSK peptide. Both these sequences contain two crucial phenylalanines, which we hypothesized to be bound similarly in both peptides. We assessed this by creating a molecular model in which the octapeptide LTFGDFDE was manually docked into the peptide-binding groove of G3BP-1, using the crystal structure of the G3BP-NTF2 complexed with the DSGFSFGSK peptide as a template (Figs. 5 and S6B). The conformational flexibility of the additional glycine residue within the FGDF motif would allow the phenylalanine side-chains to take similar orientations as in the SGFSF-peptide. Furthermore, the molecular model indicates that the negatively charged residues D5, D7 and E8 in LTFGDFDE could interact with the positively charged side chains of residues K123 and K5 as well as with the positively charged N-terminal region of G3BP (Fig. 5B). It should be noted that these two lysine residues are conserved in human, mouse and Xenopus G3BP-1 and are also present in human G3BP-2 (S7 Fig.). The binding mode suggested by the molecular model of the LTFGDFDE/G3BP-1 complex is in agreement with our biochemical analyses, which showed that mutation of the conserved phenylalanine residues at positions 451, 454, 468 and 471 of nsP3 and 10 and 13 of USP10 eliminated G3BP-binding (Figs. 1 and 3). The model might also explain the elimination of binding of both nsP3 and USP10 to G3BP upon mutation of the glycine residue to alanine, which would restrict flexibility and thus hinder adequate positioning of the phenylalanine side-chains in the FGDF motifs.

In order to validate the proposed binding model, residue F33, localized within the hydrophobic pocket of G3BP and predicted to be proximal to residue F3 of the bound LTFGDFDE peptide, was mutated to tryptophan (G3BP-F33W). Residue F33 is buried at the bottom of the peptide-binding cleft, and we hypothesized that its substitution to tryptophan would reduce the size of the pocket, thus hindering adequate positioning of the benzene ring of peptide residue F3 into the cleft (Fig. 6A, upper panel). As a control, residue F124 was also mutated to tryptophan, since this phenylalanine residue is solvent-accessible and not localized within the binding groove (Fig. 6A, lower panel). Notably, F33 is conserved in human, mouse, Xenopus and Aedes mosquito G3BP-1, while F124 conserved in human, mouse and Xenopus, but a tyrosine residue in the Aedes sequence (S7A Fig.). Both are conserved in human G3BP-2 (S7B Fig.). To determine whether mutation of these residues affects binding of G3BP with FGDF motifs, nsP3 (tagged with a biotin acceptor peptide, BAP) was co-expressed with either pEGFP-C1, pEGFP-G3BP-wt, pEGFP-G3BP-F33W or pEGFP-G3BP-F124W and binding was analyzed by immunoprecipitation with anti-GFP (Fig. 6B). Consistent with our structural model, EGFP-G3BP-wt and the F124W variant efficiently bound to nsP3-BAP, while EGFP-G3BP-F33W did not detectably interact. Similarly to results with cotransfected nsP3, endogenous USP10 was also coimmunoprecipitated with EGFP-G3BP-wt and EGFP-G3BP-F124W but not with EGFP-G3BP-F33W (Fig. 6C).

The extreme N-terminus of G3BP (residues 1–11) forms a large part of the hydrophobic pocket for positioning of the peptide residue F6 and also contains a lysine residue (K5), possibly interacting with the negative charges of acidic residues downstream of the FGDF motif. The model predicts that a truncated version of G3BP (Δ1–11) would not be capable of binding the peptide (Fig. 6D). To evaluate this, 293T cells were transfected with pEGFP-C1, pEGFP-G3BP-wt or pEGFP-G3BP-∆1–11 and analyzed by immunoprecipitation with anti-GFP and immunoblotting for USP10, GFP or actin (Fig. 6E). Endogenous USP10 coprecipitated with EGFP-G3BP-wt but not with EGFP-G3BP-∆1–11. Taken together, the results in Fig. 6 strongly support our structural model of the G3BP/FGDF complex.

Next, we asked whether the FGDF motif, mediating G3BP binding, is present in proteins other than SFV nsP3 and USP10. To identify such proteins, we searched the UniProtKB human and virus databases for all proteins containing FGxF motifs, where x = D, E or S, and also containing at least two acidic residues within the downstream 5 residues (as in both SFV nsP3 and USP10). Glutamic acid was permitted in the third position due to its chemical similarity to aspartic acid, while serine was permitted since Sindbis virus nsP3, shown by several researchers to bind G3BP-1 [25,26], contains a serine at that position. We identified 34 human (S1 Table) and 32 viral (S2 Table) sequences that meet these criteria. In one of those proteins, the infected cell protein (ICP)8 of herpes simplex virus (HSV)-1, the FGDF motif is located between residues 1144 and 1147 in the C-terminal region of the 1196 aa-long protein that is not predicted to adopt any specific secondary structure [34] and contains 3 acidic residues within the downstream 5 residues. This is reminiscent of the context of the FGDF motifs in Old World alphavirus nsP3, an otherwise quite different protein. ICP8 is a single stranded DNA binding protein that is expressed in the lytic cycle of herpes simplex viral replication. It is one of seven viral proteins that are necessary for viral DNA replication [35]. The FGDF motif in ICP8 is conserved in several HSV-1 strains and also in HSV-2 (S2 Table). To determine whether ICP8 binds G3BP-1, 293T cells were cotransfected with expression plasmids encoding HSV-1 ICP8 together with plasmids encoding EGFP-C1, EGFP-G3BP-wt, -F33W or -F124W. Lysates were immunoprecipitated with anti-GFP and immunoblotted for ICP8, GFP or actin (Fig. 7A). The results indicate that ICP8 indeed forms a complex with EGFP-G3BP-wt and EGFP-G3BP-F124W, but not with EGFP-C1 or EGFP-G3BP-F33W. These results are strikingly similar to the binding profiles of SFV nsP3 and USP10 (Fig. 6B and C), and strongly suggest that HSV-1 ICP8 forms a complex with G3BP-1 via binding of the FGDF motif at residues 1144–1147.

Since the binding of nsP3 and USP10 to G3BP inhibits SGs, we hypothesized that ICP8 binding might do the same. To investigate this, BHK cells were transfected with an ICP8 expression plasmid, stressed with sodium arsenite and stained for ICP8 as well as for G3BP-1 and TIA-1 to detect SGs. We observed that in 59% of the transfected cells, ICP8 accumulated mainly in the cytoplasm, while in the remaining 41%, ICP8 was localized in the nucleus. In neither case was there a detectable change in the localization of either G3BP-1 or TIA-1 (S8 Fig.), with these proteins exhibiting diffuse cytoplasmic or nuclear staining, respectively. After sodium arsenite stress however, only 10% of cells with predominantly cytoplasmic ICP8 contained TIA-1 and G3BP-1-positive SGs, while 96% of the cells with nuclear ICP8 and a similar proportion of mock-transfected cells had SGs (Fig. 7B). Representative images are provided in S8 Fig., showing that, when localized in the cytoplasm, ICP8 blocks the induction of SGs after sodium arsenite treatment. Due to the diffuse staining pattern of cytoplasmic ICP8, it is not possible to discern the proportion that is interacting with G3BP, but the inhibition of SG assembly was profound. Taken together with the results in Fig. 7A, this suggests that ICP8, like the cellular protein USP10 and SFV nsP3, interacts with G3BP via its FGDF-motif in a manner which blocks the formation of SGs

Expression vectors for EGFP-nsP3-31 [28], nsP3-BAP [27,42], G3BP-NTF2 [27], and HSV-1 ICP8 [43] were described previously. pEGFP-USP10_1–40 wt sequences and corresponding alanine mutants were obtained from GeneArt, and ligated between the BglII and EcoRI sites of pEGFP-C1. Construction of the infectious clone pCMV-SFV-F3A: The PCR product derived from primers 1 and 2 (S3 Table) and the PCR product derived from primers 3 and 4 were fused by a one-step PCR in a molar ratio of 1:1. The DNA was denatured and annealed at 46˚C for 2 min. These partially double-stranded molecules were made fully double stranded by extension at 72˚C for 3 min. The fusion DNA was then amplified by using primer 1 and 4 for 25 cycles of PCR consisting of treatment at 95˚C for 30 s, 69˚C for 30 s, and 72˚C for 2 min, followed by a final extension at 72˚C for 5 min. The derived PCR product was purified and subcloned into pTZ57R/T plasmid (Thermo Scientific). The resulting pTZ57R/T-F3A plasmid was digested with XhoI and BglII and religated to the similar digested pCMV-SFV4 vector [44]. The presence of mutations was confirmed by sequencing. Construction of pEGFP-G3BP-F33W and -F124W: The PCR product derived from primers 5 and 6 (F33W) or primers 5 and 7 (F124W) and the product derived from primers 8 and 10 (F33W) or primers 9 and 10 (F124W) were fused by a one-step PCR in a molar ratio of 1:1. The DNA was denatured and annealed at 33˚C for 2 min. These partially double-stranded molecules were made fully double stranded by extension at 72˚C for 3 min. The fusion DNAs containing the F33W mutation or F124W mutation were then amplified by using primers 5 and 10 for 25 cycles of PCR consisting of treatment at 95˚C for 30 s, 56˚C for 30 s, and 72˚C for 2 min, followed by a final extension at 72˚C for 5 min. The derived PCR product was purified and subcloned into pTZ57R/T plasmid (Thermo Scientific). The resulting pTZ57R/T-G3BP1-F33W, -F124W plasmid was digested with BglII and EcoRI and religated to the similar digested pEGFP-C1-G3BP1 vector. The presence of mutations was confirmed by sequencing.

All cell lines were maintained as previously described [28,45,46]. Where indicated, cells were stressed by addition of sodium arsenite (0.5 mM) or pateamine A (100 nM) in complete medium for 60 min. Cells were transfected with Lipofectamine 2000 (Invitrogen) reagent according to the manufacturer’s instructions. Virus titration was performed by plaque assay, as previously described [28]

Immunofluorescence, immunoprecipitations and immunoblotting were performed as described previously (Panas et al., 2012). For details of all antibodies used, see. S4 Table

His-tagged G3BP-NTF2 was expressed in E. coli BL21 T7 Express cells and purified using HisTrap columns (GE Healthcare). Before ITC, G3BP-NTF2 was eluted from a Superdex 75 HiLoad 16/60 (GE Healthcare) SEC column equilibrated in ITC-buffer (25 mM HEPES, 150 mM NaCl, 10 mM MgCl₂, 10% glycerol, pH 7.5) at a retention volume of 60 mL. The peptides nsP3-25 and nsP3-25-mut were dissolved in ITC buffer at a concentration of 500 μM and dialyzed extensively against the same buffer. ITC measurements were performed using an ITC200 titration calorimeter (GE Healthcare). The cell temperature was set to 37°C, the reference power to 7 μCal/sec and the syringe stirring speed to 1000 rpm. G3BP-NTF2 was loaded into the cell. The peptides were titrated in 48 injections, each injection with a volume of 750 nL, a duration time of 1.5 sec and a waiting time between the injections of 150 sec. The first injection was performed using a volume of 300 nL, a duration time of 0.6 sec and a spacing time of 120 sec. Background measurements were performed with buffer injected into the protein solution, and peptide into the buffer solution. Data were analyzed using the Origin software as included in the instrument package.

Before analysis, 40 μM G3BP-NTF2 and 400 μM nsP3-25 were mixed and incubated for 1h in SEC-buffer (25 mM HEPES, 300 mM NaCl, 5 mM MgCl₂, 10% glycerol, pH 7). A sample volume of 100 μL was injected onto the Superdex 75 10/300 GL column equilibrated in SEC-buffer. G3BP-NTF2 alone was used a control. The apparent molecular weights of the eluted proteins were calculated from their retention volumes as described in the gel filtration LMW/HMW calibration kits assuming similar globular shapes for the analyzed proteins and calibration standards (GE Healthcare).

.

All relevant data are within the paper and its Supporting Information files.