Structures of phlebovirus glycoprotein Gn and identification of a neutralizing antibody epitope

Overall Structure of the Gn Head Domain: SFTSV and RVFV.

Both SFTSV and RVFV Gn are type I transmembrane proteins with the N-terminal ectodomain binding on the cell surface and a C-terminal transmembrane helix anchored on the virus membrane (Fig. 1A). The ectodomain can be divided into the head and stem domains. To facilitate crystallization, the stem region of SFTSV Gn was removed by limited trypsin digestion due to an unsuccessful effort with the full-length ectodomain for crystallization. The last amino acid visible in the crystal structure is N340. However, the molecular weight measured by mass spectrometry (MS) indicates that the cleavage site is between residues K371 and S372, suggesting a conformational disorder in the C terminus of the truncated Gn (residues 340–372), instead of degradation during crystallization (Fig. 1A). We designed the construct of the RVFV Gn head domain based on the structure of the SFTSV Gn head domain, and succeeded in obtaining its crystal structure. Both the crystal structures of SFTSV and RVFV Gn head domains were determined at a resolution of 2.6 Å (Table S1). A Dali search within the Protein Data Bank (PDB) failed to identify any existing structures to the Gn of both SFTSV and RVFV, suggesting a novel fold. The structure of SFTSV Gn was solved by “antibody-walking” (Materials and Methods), in which the antibody provides model-based phasing information to determine the unknown structure. The RVFV Gn structure was solved by molecular replacement, using SFTSV Gn as a search model.

The Gn head domains of both SFTSV and RVFV fold into a similar triangular shape architecture consisting of three subdomains (Fig. 1 B and C). Subdomains I and II form the foundation bed supporting the subdomain III protruding on the top (Fig. 1 B and C and Fig. S1A). However, the three subdomains show a different arrangement between these two viruses, with an average rmsd of 5.010, 2.393, and 1.825 for subdomains I, II, and III, respectively (Fig. S1B). Specifically, in subdomain I of SFTSV, a five-stranded β-sheet (β2, β3, β4, β5, and β9), one helix (α3), and one 3₁₀-helix (η2) are located on the interface of subdomains I and II. Two α-helices (α1 and α2), three 3₁₀-helices (η1, η3, and η4), and two sets of small antiparallel β-strands (β1 and β8, β6 and β7) flank the β-sheet on the opposite side. A free cysteine (C99) on α3 is located in the interior of subdomain I. For the RVFV Gn subdomain I, a four-stranded β-sheet (β1, β2, β3, and β6) and three α-helices (α2, α3, and α4) can be found on the interface of subdomains I and II. One α-helix (α1), three 3₁₀-helices (η1, η2, and η3), and one pair of small antiparallel β-strand (β4 and β5) are on the opposite side. β-Strands are the major component in subdomain II and the pattern of β-strands between SFTSV and RVFV is the same (Fig. 1 D and E and Fig. S2). The core structure comprising the six β-strands is located next to subdomain I and a three-stranded β-sheet connects to subdomain III. The only different secondary element in subdomain II between these two is the α-helix. SFTSV contains an α4, whereas RVFV does not have it in subdomain II. The secondary elements in subdomain III between SFTSV and RVFV are similar. Four β-strands and three α-helices stabilize subdomain III, in which α5 in SFTSV is replaced by η5 in RVFV (Fig. 1 D and E). Although the Gn structures in the Phlebovirus genus display similar configurations, the overall structures between the Phlebovirus genus and Hantavirus genus are distinct (Fig. S1 C and D). Furthermore, the topology between these two genera is completely different (Fig. S1E).

The glycosylation sites are observed to be different between SFTSV and RVFV Gn. For SFTSV, two N-linked glycans (N33 and N63) can be observed in subdomain I (Fig. 1 A, B, and D and Fig. S2), which is consistent with theoretical predictions. For RVFV, although N438 is predicted to be an N-linked glycosylation site, no glycans can be observed in the solved crystal structure in the insect cell-expressed protein (Fig. 1 A, C, and E).

Both the SFTSV and RVFV Gn are cysteine-rich proteins. SFTSV Gn has 27 cysteines, whereas RVFV Gn has 28 cysteines. Twelve cysteines in the head domains are identical between these two Gn, and all of the other 10 cysteines in the stem domains are conserved for the 11 sequences of the available phleboviruses (Figs. S2 and S3A). The SFTSV Gn head domain contains eight disulfide bonds and a free cysteine. Specifically, two disulfide bonds (C26-C49 and C143-C156) and an unpaired cysteine (C99) are in subdomain I, one disulfide bond is in subdomain II (C206-C216), one disulfide bond is across subdomains II and III (C180-C327), and four disulfide bonds are in subdomain III (C258-C305, C266-303, C274-C280, and C287-C292). The RVFV Gn head domain is stabilized by nine disulfide bonds, four of which are in subdomain I (C179-C188, C229-C239, C250-C281, and C271-C284), one of which is in subdomain II (C322-C332), three of which are in subdomain III (C374-C434, C402-C413, and C420-C425), and one of which is across subdomains II and III (C304-C456). Six disulfide bonds (C271-C284 in subdomain I, C304-C456 in subdomain II, C322-C332 crossing subdomain II and III, C374-C434, C402-C413, and C420-C425 in subdomain III, RVFV Gn numbering) are conserved between the SFTSV and RVFV Gn head domains (Fig. S2).

Dimerization of Gn Through Its Stem Region.

The full-length SFTSV Gn ectodomain was produced using the baculovirus expression system. Results from size-exclusion chromatography using a Superdex 200 10/300 GL column showed the peaks eluted at 12.6 mL and 14.4 mL, corresponding to the dimer and monomer in solution, respectively (Fig. 2A). Nonreducing SDS/PAGE shows that the dimer is linked by disulfide bonds (Fig. S4A). Notably, the dimer was completely disassociated after a treatment of limited trypsin proteolysis (Fig. 2A and Fig. S4A), indicating that the C terminus of Gn is solely responsible for dimerization via disulfide bonds (Fig. 2A). To characterize the C terminus of Gn, we constructed the unstructured part of Gn (Gn-C), which corresponds to residues 338–452. The Gn-C protein forms a dimer under nonreducing conditions, which is consistent with the above-described result (Fig. S4B). Full-length RVFV Gn forms a dimer under nonreducing conditions as well (Fig. S4C).

To determine which cysteines are responsible for dimerization, we used MS and site-directed mutagenesis to analyze Gn-C and the full-length Gn protein of SFTSV. MS analysis identified four intermolecular disulfide bonds involving C430, C435, C438, and C447 from the Gn-C dimer but not the monomer (Table S2), suggesting that these cysteine residues may mediate Gn dimerization. Indeed, simultaneously mutating these four cysteine residues to alanine completely abolished dimer formation of the full-length Gn, leaving only the monomer (Fig. 2B). Furthermore, the C430A/C447A substitution shifted the equilibrium toward the Gn monomer, but failed to abolish the Gn dimer. The C435A/C438A substitution reduced the relative abundance of the dimer more substantially than the C430A/C447A substitution (Fig. 2B), suggesting that although C430, C435, C438, and C447 all contribute to Gn dimerization, the two intermolecular disulfide bonds mediated by C435 and C438 play a major role. MS analysis also identified a disulfide bond between C356 and C424 in both the Gn-C monomer and dimer (Table S2). Mutation of either C356 or C424 severely reduced the yield of the protein (Fig. S5), suggesting that the C356-C424 disulfide bond is critical for stabilizing Gn. The data described above support that the cysteines in the stem region are responsible for dimerization and the cysteines are highly conserved in the same genus (Figs. S3 and S6). It is noteworthy that the last four cysteines (C430, C435, C438, and C447, SFTSV numbering) responsible for dimerization are highly conserved across five genera (Phlebovirus, Hantavirus, Nairovirus, Orthobunyavirus, and Tospovirus) of the Bunyaviridae family, indicating that members in the whole Bunyaviridae likely share the same assembly organization with a Gn disulfide bond-linked dimer (Fig. S7).

Human Antibody MAb4-5 Specificity to SFTSV Gn.

Antibody MAb 4–5 was previously isolated using whole SFTSV virions as bait from a phage-display antibody library derived from the peripheral blood mononuclear cells of a patient that recovered from SFTS disease (30). Since only the sequence of the variable region was available, we constructed the Fab fragment with the variable region of MAb 4–5 heavy and light chains, combined with the constant region of a hemagglutinin antibody CR8020 (IgG1) (33). To verify neutralizing activity, we also constructed full-length MAb 4–5 with human Ig G1 (IgG1). The recombinant antibodies were expressed by 293T mammalian cells and the concentration required to obtain 50% neutralization of 100 TCID₅₀ SFTSV in vitro was 44.2 μg/mL (Fig. 3A). The interaction between Gn and MAb 4–5 was further demonstrated by surface plasmon resonance (SPR) assays with a dissociation constant (K_d) of 25.9 nM (Fig. 3B). We also measured the binding between RVFV Gn and MAb 4–5, but no binding or neutralization was observed, indicating the specific binding of MAb 4–5 to SFTSV (Fig. 3 C and D).

Complex Structure of the SFTSV Gn Head and Neutralizing MAb 4–5.

To elucidate the structural basis of virus neutralization, we further prepared the Gn head–MAb 4–5 Fab complex by mixing the two proteins in vitro and then purifying by size-exclusion chromatography (Fig. S8). Consistent with the high binding affinity between Gn head and MAb 4–5, the complex is stable and easily obtained. Crystals diffracting to 2.1 Å were grown from digested Gn in complex with MAb 4–5 Fab (Table S1). The complex structure was determined by molecular replacement using a Fab structure (PDB ID code 4RIR) as the search model followed by iterative rounds of model building and refinement. Automatic model extension of the missing domain of the Gn head was carried out with AUTOBUILD in PHENIX (34). Therefore, the SFTSV Gn structure described earlier was actually solved by “antibody-walking.” Structure analysis reveals that MAb 4–5 binds the membrane-distal head of Gn and the contact was mediated only by the heavy chain (Fig. 4A). The interaction surface between MAb 4–5 and Gn buries 614.8 Ų of the molecular surface. The primary region of the interaction is the complementarity-determining region (CDR) H3 (Fig. 4B), which penetrates the hydrophobic trough, consisting of α5, α6, and η5 in SFTSV Gn domain III. In this interaction, Y104 of the CDR H3 interacts with F256, F286, V289, and A290. Notably, α6 is the key element on the Gn head domain, which contacts all three CDR regions in the MAb 4–5 heavy chain. In particular, K288 is the most important residue located on α6. Specifically, the K288 main chain hydrogen-bonds the W33 side chain on the CDR H1 loop. The side chain of K288 can form salt bridges with the side chains of both D55 and D57 on the CDR H2 loop. K288 main chain hydrogen bonds the side chain of R101 on the CDR H3 loop. Moreover, two electrostatic patches are observed on the antigen–antibody interface. The basic patches, which consist of K100 and R102 on the CDR H3 loop, interact with E293 on Gn, while the acidic patch (D55 and D57) on the CDR H2 loop forms salt bridges with Gn K288 (Fig. 4C). To determine the critical role of residue 288 in MAb 4–5 binding, two substitutions (K288A and K288E) were constructed, expressed, and purified. Affinity assay indicates that these two substitutions dramatically decrease the binding between Gn and MAb 4–5 (Fig. S9).

The MAb 4–5 can recognize the denatured SFTSV Gn head but not the RVFV Gn head by Western blot (Fig. 5A). Structural analysis indicates that α6 is the major epitope recognized by MAb 4–5. Comparison of the epitope amino acids sequence between SFTSV and RVFV shows that only two amino acids are conserved (F286 and C287, SFTSV numbering) (Fig. 5B). The key residue K288 is replaced by serine in RVFV, which is the main reason leading the abolishment of binding to MAb 4–5. Specifically, the salt bridges between K288 and D55, K288 and D57 were lost. The hydrogen bonds between the Gn head and heavy chain were lost due to the shift of the main chain (Fig. 5C). Moreover, the hydrophobic pocket in SFTSV Gn can accommodate the side chain of F104 in the MAb 4–5 heavy chain. In contrast, the pocket becomes shallow (Fig. 5D), as A290 is replaced by Y423 in RVFV. Additionally, the side chain of K405 displays steric hindrance to CDR H3 loop. On the other hand, K405 in RVFV is a positive-charged amino acid, which shows repulsive force to the positive-charged R102 on the CDR H3 loop. Further conservation analysis of surface amino acids among 11 phleboviruses indicates that domain III is a variable region, which can be recognized by a specific neutralizing antibody (Fig. S10A). In contrast, domain II shows the relatively conserved epitope, which may be recognized by broadly neutralizing antibody (Fig. S10 B and C).

Protein Expression and Purification.

The SFTSV Gn ectodomain (GenBank accession no. JF906057.1, residues 20–452) followed by a C-terminal six-histidine purification tag was subcloned into the pFastBac1 vector (Invitrogen) modified with a gp67 signal sequence at the N terminus, as previously described (50–53). Transfection and virus amplification were conducted with sf9 cells, and the recombinant proteins were produced in High Five cells. The cell culture media were collected 60 h after infection and then purified by nickel affinity chromatography with a 5-mL HisTrap HP column (GE Healthcare) and size-exclusion chromatography with a Hiload 16/60Superdex 200-pg column in 20 mM Tris, pH 8.0, 50 mM NaCl. Gn protein was pooled and incubated with trypsin at a mass ratio of 300:1 at 277 K overnight, and further purified on a Superdex 200 column (GE Healthcare). Digested Gn was then concentrated to 7.5 mg mL⁻¹ for crystallization. The RVFV Gn head domain (GenBank accession no. JQ068143.1, residues 154–469) was constructed using the same strategy without trypsin digestion. The Gn head was concentrated to 10 mg mL⁻¹ for crystallization. To prove the C terminus of Gn playing an important role in dimerization, the Gn-C (residues 338–452) was constructed using the same strategy as the SFTSV Gn ectodomain described above. Moreover, the expression and purification strategies of Gn-C were the same as SFTSV Gn as well.

MAb 4–5 Fab was synthesized with its variable region (30) and the constant region of CR8020 IgG1 (33) into the pcDNA4 expression vector. A six-histidine tag was designed at the C terminus of the light chain for purification. The MAb 4–5 heavy chain was subcloned into a modified pcDNA4 with mouse IgG Fc at the C terminus. The full-length and the Fab fragment of MAb 4–5 were produced by HEK293T cells and purified by protein A and HisTrap HP column, respectively. MAb 4–5 Fab was further purified on a Superdex 200 column in 20 mM Tris, pH 8.0, 50 mM NaCl for crystallization, while the full-length MAb 4–5 was purified on a Superdex 200 column in PBS buffer for neutralization assay.

To obtain the complex of Gn and MAb 4–5, MAb 4–5 was mixed with the digested Gn at a molar ratio of 1:1 and incubated at 277 K for 1 h. The mixture was then loaded on a Superdex 200 column in 20 mM Tris, pH 8.0, 50 mM NaCl, and concentrated to 10 mg mL⁻¹ for crystallization.

Crystallization.

All crystallizations were performed using a vapor-diffusion sitting-drop method with 1 μL protein mixing with 1 μL reservoir solution. SFTSV Gn crystals were grown in the reservoir solution comprising of 20% (wt/vol) PEG 4000, 20% (vol/vol) 2-propanol, and 0.1 M sodium citrate, pH 5.5, at 291 K. RVFV Gn head crystals were grown in the reservoir solution of 0.2 M ammonium sulfate, 0.1 M Mes, pH 6.5, 20% (wt/vol) PEG 8000 at 277 K. Good diffraction crystals of the complex protein were finally obtained in 0.3 M potassium/sodium tartrate, 20% (wt/vol) PEG 3350, and 0.1 M Bis Tris propane, pH 7.5, with protein concentration of 15 mg mL⁻¹. Crystals were frozen in liquid nitrogen in reservoir solution supplemented with 20% glycerol (vol/vol) as a cryoprotectant. Data were collected at the Shanghai Synchrotron Radiation Facility BL17U. All data were processed with HKL2000.

Structure Determination and Refinement.

The structure of the SFTSV Gn head-MAb 4–5 complex was determined by molecular replacement using an antibody Fab fragment (PDB ID code 4RIR) as search probe. The molecular model was rebuilt using COOT (54) and refined with REFMAC (55). Subsequently, the automatic model extension of the missing Gn domain was carried out with AUTOBUILD in PHENIX (34). The method used for determination of the unknown structure by getting the phasing information from antibody is called “antibody walking.” The SFTSV Gn head structure was solved by molecular replacement using the refined coordinates obtained from the complex and the structure was refined using REFMAC. The RVFV Gn head structure was solved by single-wavelength anomalous diffraction, with a gold derivative (NaAuCl₄·2H₂O). Glycans were added at the final stages of model building and refinement according to the density map. Structure validation was performed with PROCHECK (56). Data collection and refinement statistics are summarized in Table S1. Figures were prepared with PyMOL (www.pymol.org).

Binding Assays.

Protein interactions were tested using both SPR analysis and the Octet RED96 biosensor method. For the SPR assays, a BIAcore 3000 spectrometer was used to measure kinetic constants at room temperature (298 K). All proteins were exchanged into a buffer of 20 mM Hepes, pH 7.4, 150 mM NaCl and 0.005% (vol/vol) Tween 20. MAb 4–5 proteins were immobilized on CM5 chips (GE Healthcare) at approximately 1,000 response units and analyzed for real-time binding by flowing through gradient concentrations (ranging from 7.8 to 1,000 nM) of Gn head proteins. For the Octet RED96 biosensor method, samples of buffer were dispensed into polypropylene 96-well black flat-bottom plates (Greiner Bio-One) at a volume of 200 μL per well, and all measurements were performed at 298 K in HBS-EP buffer with the plate shaking at the speed of 1,000 rpm. Anti-human IgG Fc (AMC)-coated biosensor tips (Pall ForteBio) were used to capture antibody MAb 4–5 from 10 ng/μL stock buffer. The buffer was 20 mM Hepes, pH 7.4, 150 mM NaCl and 0.005% (vol/vol) Tween 20. Kinetic measurements for antigen binding were performed by exposing biosensors to a series of analyte concentrations (15.6−250 nM for Gn wild-type, and 15.6–500 nM for Gn mutants) and background subtraction was used to correct for sensor drifting. All sensors were generated with 10 mM Glycine-HCl, (pH 1.7, GE Healthcare). The data were processed by ForteBio’s data analysis software and plotted by Origin software.

Neutralization Assay.

Fifty-microliters of twofold serial-diluted MAb 4–5 or control antibody (Ebola monoclonal antibody 13C6) (57) was mixed with equal volume of 100 TCID₅₀ SFTSV (SDYY007), or RVFV (BJ01) at 310 K for 1 h. The virus–antibody mixture was then transferred to 80% confluent Vero cells in a 96-well plate and incubated at 310 K for 1 h. The plate was washed with DMEM three times, and incubated at 310 K in a 5% CO₂ incubator. Cytopathic effects were observed every 24 h for 6 d. The antibody was considered as having neutralizing capacity if 100% of viral cytopathic effects was inhibited.

Identification of Protein Disulfide Bonds by MS Analysis.

Purified SFTSV Gn-C was denatured in a buffer containing 8 M Urea, 1 mM Tris, pH 6.5, 2 mM N-ethylmaleimide before it was subjected to SDS/PAGE. The monomer and dimer bands of Gn-C were excised and digested in gel according to a previously published protocol (58), with the following modifications: (i) reduction and alkylation were omitted, (ii) 0.5 mM N-ethylmaleimide was added to all of the destaining solutions and wash buffers, and (iii) dehydrated gel slices were rehydrated with 100 mM Tris, pH 6.5, containing Lys-C and Asp-N at 5 ng/µL each for overnight digestion. Digested peptides were analyzed on a Q Exactive mass spectrometer coupled to an Easy Nano-LC 1000 liquid chromatography system (Thermo Fisher Scientific).

Peptides were desalted on a 75-μm × 6-cm precolumn that was packed with 10 μm, 120 Å ODS-AQ C18 resin (YMC Co., Ltd.) and connected to a 75-μm × 10-cm analytical column packed with 1.8 μm, 120 Å UHPLC-XB-C18 resin (Welch Materials). The peptides were separated over a 34-min linear gradient from 5% buffer B (100% acetonitrile, 0.1% formic acid), 95% buffer A (0.1% formic acid) to 30% buffer B, followed by a 3-min gradient from 30 to 80% buffer B, then maintaining at 80% buffer B for 7 min. The flow rate was 200 nL/min. The MS parameters were as follows: the top 20 most-intense ions were selected for HCD dissociation; R = 140,000 in full scan, R = 17,500 in HCD scan; AGC targets were 1e6 for FTMS full scan, 5e4 for MS2; minimal signal threshold for MS2 = 4e4; precursors having a charge state of +1, >+8, or unassigned were excluded; normalized collision energy, 30; peptide match, preferred.

The raw data of the 10-protein sample were preprocessed using pParse (59), which was set to exclude coeluting precursor ions and precursors of +1, +2 charge. To identify disulfide-linked peptides, the MS2spectra was searched using pLink (60) against a protein database containing the sequences of the analyte protein and all of the proteases used. The pLink parameters search were maximum number of missed cleavages = 5; minimum peptide length = 4 amino acids; fixed modification of −1.007285 Da on cysteine and the disulfide mass was set to zero to allow the identification of more than one disulfide bond in a peptide pair; candidate pairs satisfying Mα+ Mβ+ Mlinker < M ± 5 Da were scored for spectrum-peptide matching (Mα, Mβ, Mlinker, and M denote the masses of the candidate α-peptide, candidate β-peptide, the linker, and the observed precursor, respectively). pLink search results were filtered by requiring E-value < 0.001, false-discovery rate < 0.05, spectra count ≥ 20, and no more than 10-ppm deviation between the monoisotopic masses of the observed precursor and the matched disulfide-linked peptide (or peptide pair).

Western Blot Analysis.

The SFTSV and RVFV Gn head proteins were fractionated by 12% SDS/PAGE. The separate proteins were electro-transferred to a nitrocellulose membrane and incubated with full-length MAb 4–5 antibody and a horseradish peroxidase-conjugated secondary goat anti-human IgG antibody (SC-2453; Santa Cruz). SuperSignal West Pico Chemiluminescent Substrate (Pierce 34080) was used for detection according to the manufacturer’s instructions.

Fitting of the RVFV Gn Structure into the Low-Resolution EM Structure.

The crystal structure of RVFV Gn was manually fitted into the capsomers in the 22-Å resolution cyro-EM map of RVFV (EMDataBank ID code EMD-1550) with UCSF Chimera (61). The positions of Gn with the five- and six-coordinated capsomers were adjusted according to the fivefold and sixfold rotational symmetry, respectively, without major steric clashes.