Mechanistic Insight into Bunyavirus-Induced Membrane Fusion from Structure-Function Analyses of the Hantavirus Envelope Glycoprotein Gc

Abstract

Hantaviruses are zoonotic viruses transmitted to humans by persistently infected rodents, giving rise to serious outbreaks of hemorrhagic fever with renal syndrome (HFRS) or of hantavirus pulmonary syndrome (HPS), depending on the virus, which are associated with high case fatality rates. There is only limited knowledge about the organization of the viral particles and in particular, about the hantavirus membrane fusion glycoprotein Gc, the function of which is essential for virus entry. We describe here the X-ray structures of Gc from Hantaan virus, the type species hantavirus and responsible for HFRS, both in its neutral pH, monomeric pre-fusion conformation, and in its acidic pH, trimeric post-fusion form. The structures confirm the prediction that Gc is a class II fusion protein, containing the characteristic β-sheet rich domains termed I, II and III as initially identified in the fusion proteins of arboviruses such as alpha- and flaviviruses. The structures also show a number of features of Gc that are distinct from arbovirus class II proteins. In particular, hantavirus Gc inserts residues from three different loops into the target membrane to drive fusion, as confirmed functionally by structure-guided mutagenesis on the HPS-inducing Andes virus, instead of having a single "fusion loop". We further show that the membrane interacting region of Gc becomes structured only at acidic pH via a set of polar and electrostatic interactions. Furthermore, the structure reveals that hantavirus Gc has an additional N-terminal "tail" that is crucial in stabilizing the post-fusion trimer, accompanying the swapping of domain III in the quaternary arrangement of the trimer as compared to the standard class II fusion proteins. The mechanistic understandings derived from these data are likely to provide a unique handle for devising treatments against these human pathogens.

Fig 1. Structure of Hantaan virus Gc.

A) Organization of the hantavirus M genomic RNA segment. The black striped regions represent the non-coding 3’ and 5’ RNA ends, with the single open-reading frame in between. Regions corresponding to the signal sequences for Gn and Gc are in cyan, and other trans-membrane regions are in dark grey. The black line below indicates the region spanned by the recombinant Gc ectodomain (rGc). B) The rGc / scFv A5 complex. Gc is colored according to domains with domains I, II, and III in red, yellow, and blue, respectively, with secondary structure elements labeled. The linker between domains I and III is shown in cyan, and the C-terminal end in magenta (beginning of the “stem” region). The disulfide bonds are displayed as sticks with sulfur atoms in green, and numbered in green. The scFvA5 is shown with the heavy chain dark green and the light chain in grey. The lower panel shows an orthogonal view of Gc, with the scFvA5 removed for clarity. The single N-linked glycan Asn280 is labeled. Dotted lines indicate the disordered region, corresponding to the cd loop (orange) and the distant end of the bc loop (yellow). C) The Gc surface colored according to domains, with the epitopes of—and sites of escape mutation to—neutralizing antibodies mapped in grey or green in different shades. The protein is shown in the same two orthogonal views of B. D) Multiple sequence alignment of Gc from seven representative hantaviruses. Strictly conserved and highly similar residues are highlighted in red or brown background, respectively. All the cysteines are shown in a green background, and asparagine residues within N-linked glycosylation motifs in a blue background. The corresponding host is noted as MUR (Murinae), SIG (Sigmodontinae), ARV (Arvicolinae) corresponding to rodent subfamilies and as INS (Insectivores), and BAT (Chiroptera). The secondary structure elements are displayed above the sequences, on a background colored according to the tertiary structure, as in panel B. The missing regions are in grey or black (within the colored tertiary structure bar, a top grey half bar denotes residues disordered in the neutral pH structure, and a bottom black half bar those not visible in the post-fusion structure). A hashed grey bar denotes the TM segment. Disulfide bonds are numbered in green below the sequences. The domain III residues involved in inter-subunit interactions with the Asn280 glycan in the post-fusion trimer are marked with a green star under the alignment.

Fig 2. Gc interactions with lipids and post-fusion structure.

A) scFv A5 specifically inhibits liposome binding by rGc as measured by SPR. The liposomes were immobilized on an SPR chip (see Methods) and Gc alone or together with scFv A5 at increased concentrations was injected to measure binding (top panel). The same experiment with an unrelated, non-binding scFv is provided as control in the lower panel. B) Left panel: Electron micrographs of negative stained samples of rGc show long thin rods with dimensions comparable to the neutral pH structure shown in Fig 1B and 1C. Right Panel: Incubation of rGc with liposomes (see Materials and Methods for lipid composition) at low pH results in rGc insertion into the membranes, with a concomitant change in aspect to resemble class II viral fusion proteins inserted into liposomes in their post-fusion, trimeric conformation (as shown previously for alphaviruses and flaviviruses). C) The structure of the Gc trimer crystallized at pH 6.5 (W115H mutant) displays the typical class II post-fusion conformation. The protomer in the foreground is shown in ribbons colored by domains, with the single glycosylation site at Asn280 in sticks, as well as key residues at the tip of domain II. The two protomers in the background are shown as surfaces, one colored in grey and the other by domains, with the glycan surface in green. The structure determination used a W115H mutant, but the representation shows the location of the mutated tryptophan (labeled). The trimer axis is drawn in black. D) Conformational transition of rGc. The neutral pH monomeric form colored gray and the low pH trimer subunit colored according to domains were superposed on domain I and are shown in two views to highlight the overall reorganization of the molecule. The left panel shows the axis (nearly normal to the plane of the Figure) about which domain III pivots by more than 120 degrees. The right panel shows the 26 degrees hinge of domain II about the domain I/II junction. E) Close up view (from the top) of the tip of domain II showing the structuring role of Asn118. The polypeptide chain is shown in yellow except for the cd loop (orange), with non-carbon atoms in different color (nitrogen, oxygen and sulfur in blue, red and green, respectively). Asn118 is highlighted in white sticks with an orange outline. Its side chain carboxamide group donates two hydrogen bonds (yellow dots) to the main chain carbonyls of Cys122 and Gly124, and accepts a hydrogen bond from the main chain amide of Ser114. In addition, the main chain carbonyl of Asn118 accepts two hydrogen bonds from the ij loop, from the main chain amide groups of Ala251 and Thr252. Note that Asn118 is framed by disulfides 4 (immediately upstream) and 1 (downstream), which connect the cd loop to the ij and bc loops, respectively. Trp115 and Phe250, discussed in the paper, are displayed in sticks, as a guide. F) Cartoon to show the swapping of β-strand A₀ during trimerization. The left panel shows domain I in the pre-fusion conformation, with inner and outer β-sheets in yellow and green, respectively. The short A₀ strand runs antiparallel to C₀ at one edge of the outer β-sheet (indicated in yellow to highlight that it becomes part of the inner sheet in the post-fusion form). In the trimer, the A₀ strand swaps to the adjacent domain I (depicted with inner and outer β sheets in pink and red, respectively), to run antiparallel to the B₀ strand in the inner β-sheet. For clarity, the front subunit in the trimer (which would have inner and outer sheets in pale and dark blue) was removed, leaving only its A₀ strand, which inserts into the β-sandwich of the yellow/green domain (on which the monomer was superposed for this view). The “pink” A₀ strand inserts into the blue subunit (not displayed for clarity). Downstream strand I₀, the domain I-III linker runs alongside the inserted A₀ strand in the trimer, swapping also domain III, which is further downstream (see panel C). The limits of strand A₀ are labeled, to show what residues rearrange (see also the secondary structure diagram in Fig 1D).

Fig 3. A carboxylate-carboxylic acid hydrogen bond structures the domain II tip.

A) Left panel: the Glu106 side chain is at the center of a string of hydrogen bonds linking the side chains of Asp108 and Trp98 through that of Glu106. The overall network also involves the phenol group of Tyr96 making a hydrogen bond with the main chain carbonyl of Cys91 in the bc loop (involved in disulfide bond 2, connecting the bc loop to the d strand) and via a water molecule to the imidazole ring of His86 (located at the end of the b strand, not displayed for clarity). See the corresponding electron density in the matching S1 Movie. Middle panel: the shift in position of Tyr105 to chelate the K⁺ ion affects Glu106, such that its pK drops and becomes deprotonated. It now accepts a bifurcated hydrogen bond from the imidazole ring of His104, while the phenol group of Tyr96 now donates a hydrogen bond to Asp108, releasing a set of interactions that lead to de-structuring of the cd loop and the N-terminal part of the bc loop (dotted lines). In particular, the side chain of Cys91 becomes disordered, and only the side chain of its partner in disulfide bond 2 remains partially visible (Cys129). The corresponding electron density in shown in the S2 Movie. Right panel: the tip of domain II at neutral pH from the structure of the complex with scFv A5 (which was removed, for clarity). Although the resolution of this structure is lower (3 Å instead of 1.6 Å and 1.4 Å) and the hydrogen bond interactions are less well defined than in the structures of the Gc trimer, it is clear that at neutral pH Glu106 accepts a hydrogen bond from His99. The corresponding electron density is shown in the S3 Movie. Asp108 is the last residue with a visible side chain, and is too far to interact with E106. At this pH, such an interaction is not expected, since both side chains will be negatively charged and repel each other. B) Close view of the K⁺ binding site, near the 3-fold axis on the Gc trimer when crystallized in the presence of KCl at concentrations above 200 mM. The reference protomer is colored yellow and an adjacent one dark grey. The third protomer is not displayed, for clarity, and the 3-fold axis is represented in light blue. β-strands are labeled. The K⁺ ion is shown as a small green sphere, and the residues involved in its coordination are depicted in sticks, as well as Glu106, which forms the carboxylate-carboxylic acid hydrogen bond at low pH in the absence of KCl. Except for Tyr105, the K⁺ coordinating atoms are all from the main chain. The various crystal structures at different KCl concentrations are summarized in the S4 Fig and in the S4 Movie. C) and D) Andes virus Gc mutants in the residues involved in the side chain Asp108-Glu106-Trp98 hydrogen bond network are impaired in membrane fusion activity. The expression level and plasma membrane localization of these mutants (which concern residues strictly conserved in the hantavirus genus, Fig 1D) is provided in the S5 Fig, which shows that they reach the cell surface in similar amounts as wild type. C) Syncytia formation of cells expressing Gn and wild type or mutant Gc when exposed at the indicated pH. Fluorescence microscopy employing three different stains was used for the quantitation of cells and nuclei and syncytia formation results from at least n = 2 experiments were averaged (see the S6 Fig). D) Entry of Vero E6 by SIV particles pseudotyped with Andes virus spikes with wild type Gn and either Gc wild type or the indicated mutant (Gn/Gc pp). Reporter gene (GFP) expressing cells were quantified by cell cytometry (S7 Fig).

Fig 4. Domain III is “swapped” in the post-fusion Gc trimer with respect to the standard arbovirus class II fusion proteins.

A) Comparison of the post-fusion Gc trimer with the other class II fusion proteins of known structure. A cartoon representation of Hantaan virus Gc next to the rubella virus E1 (labeled RV; PDB entry 4B3V [36]) to its left, and Semliki Forest Virus E1 (alphavirus, labeled SFV; PDB entry 1RER [51]), the Dengue virus serotype 1 E protein (flavivirus, labeled DV1; 4GSX [49]), and the C. elegans cellular fusogen EFF-1 (4OJC [59]). A « fused membrane » is diagramed above, with aliphatic and hydrophilic layers in dark and light gray, respectively. In each trimer, a “reference” subunit is shown in bright standard class II colors (with the stem region in dark pink), the anticlockwise subunit when looking from the membrane in pale colors, and the third in light grey. Disulfide bonds are green, and glycan chains are shown in gold sticks with a red outline. The trimer axes are drawn in each case. Note that the non-arthropod-borne rubiviruses and hantaviruses have domain III swapped with respect to the others (i.e, the dark blue domain III is to the right and not to the left, as in the others). Note also two features specific of hantavirus Gc: the N tail (boxed in red) interacting with the domain I-III linker and with domain III, and the Asn280 glycan interacting with a neighboring protomer (boxed in blue). B) Close up view showing the inter-subunit contacts of the sugar residues within the trimer. The domain III amino acids interacting with the glycan are also marked with green stars in the alignment of Fig 1D. C) Close up view of the polar contacts of conserved residues from the N tail. Hydrogen bonds are indicated as dotted lines, including those between main chain atoms, in which case the interacting atoms are not drawn but only the ribbon, for clarity. D-E) Mutation of conserved N tail residues render Gc non-functional for fusion. D) Fusion activity of cells expressing Gn and N tail mutants of Andes virus Gc at neutral or acid pH. Syncytia formation was quantified by counting cells and nuclei using three-color fluorescence microscopy (S6 Fig) and results from at least n = 2 experiments were averaged. E) Entry of Vero E6 by SIV particles pseudotyped with Andes virus spikes (Gn/Gc pp) with wild type Gn and Gc wild type or the indicated N tail mutant. Reporter gene (GFP) expressing cells were quantified by cell cytometry (S7 Fig). F-G) trimer formation and stability of selected N tail mutants of Andes virus Gc. F) Acid-induced trimer formation of wild type or mutant Gc. Sucrose sedimentation of Andes virus VLPs after treatment at the indicated pH and subsequent extraction by Triton X-100. The presence of Gc in different fraction was detected by western blot analysis and the molecular mass of each fraction determined by a molecular marker. G) Trimer stability of wild type and mutant Andes virus Gc assayed by trypsin. VLPs including Gn and wild type or mutant Gc were treated at the indicated pH for 30 min, back-neutralized, and incubated with trypsin for 30 min. Gc resistance to trypsin was assessed by western blot analysis and results were quantified by densitometry from n = 3 independent experiments. As a control, mutation of residues in the membrane interaction region, which are not active in fusion (see below, Fig 5) but are not involved in trimer contacts (Y88A and F250A), behave like wild type in this assay. The statistical evaluation of each data point was performed in relation to the wild type Gc treated at pH 5.5. ***, P < 0.00025; **, P < 0.0025; *, P < 0.025; ns, not significant.

Fig 5. The Gc fusion loops.

A) Surface representation of the tip of HNTV domain II viewed down the three-fold axis from the membrane. Two of the subunits are shown in solid surface representation and the third was rendered semi-transparent to visualize the polypeptide chain and selected side chains. B) Orthogonal view of one protomer, with the same residues and loops labeled. C) For comparison, the same representation of flavivirus E (Dengue 2, 1OK8; [53]) and D) phlebovirus Gc (RVFV, 4HJ1; [20]) fusion loop region. Note that in hantavirus Gc, the ij loop is longer and projects Phe250 to the membrane, whereas in flaviviruses and phleboviruses, it is not the case (the same holds true for alphaviruses, not depicted here). Similarly, the bc loop projects Tyr88 towards the membrane. E) Fusion activity of cells expressing Andes virus wild type Gn and wild type or mutant Gc at neutral or acid pH. Syncytia formation was quantified by counting cells and nuclei using three-color fluorescence microscopy (S6 Fig) and results from at least n = 2 experiments were averaged. F) Entry into Vero E6 cells viral particles pseudotyped with Andes virus spikes with wild type Gn and wild type or indicated mutant Gc (Gn/Gc pp; as in Figs 3D and 4E; see the S7 Fig for cell cytometry quantification of reporter GFP production). G) Liposome co-flotation assay to visualize acid-induced membrane interaction of wild type and mutant Andes virus VLPs. VLPs were incubated with liposomes labeled with a fluorescent dye (DPH, see Materials and Methods) at pH 7.4 or 5.5. Fractions of the step gradient sedimentation were examined for the presence of Gc by western blot and liposomes by fluorescence. H) Trimer formation by wild type and mutant Gc. Sucrose sedimentation of VLPs incubated at pH 7.4 or 5.5 and detection of Gc by western blot analysis. I) Acid-induced lipid mixing kinetics with liposomes of VLPs including Gn and wild type or mutant Gc. The decrease of pyrene-excimer fluorescence intensity was detected upon low pH incubation of pyrene-labeled VLPs with liposomes as a function of time. Results are representative for at least n = 2 independent experiments.

Fig 6. A common tertiary structure motif in Gc from four bunyavirus genera.

A) Multiple Gc amino acid sequence alignment from one representative each of the Hantavirus (HNTV, P08668.1), Tospovirus (TSWV, NP_049359.1), Orthobunyavirus (BUNV, NP_047212.1), and Nairovirus (CCHFV, AAK52743.1) genera. The secondary structural elements of Hantaan virus Gc are shown in the upper line (as in Fig 1D), the hantavirus disulfide bonds are labeled, and cysteines not conserved across genera are in an orange background. The residues exposed at the membrane-interacting end of domain II in hantavirus Gc are marked with a blue star under the alignment. The four boxes mark clusters of high conservation, which are very informative when analyzed by pairs, Hanta-Nairovirus and Tospo-Orthobunyavirus genera, which are closer to each other than to the others (see the S8 Fig). A cyan background within boxes 2 and 4 (marked also by a small black arrow underneath) highlights amino acids with complementary electrostatic charges predicted to come into contact at the membrane interacting surface, between disulfides 4 and 5, which connect cd and ij loops. B) The motif obtained from the alignment in A (top line) was run in Prosite to scan the SwissProt and trEMBL databases, and the taxonomic distribution of the matched entries are provided in the colored pie chart. The 1057 entries found encompass all of the known viruses within the four genera, plus some that are “unclassified” by the server. Searching the literature shows that the 7 viruses given as “unclassified Bunyaviridae” have been assigned to one of the genera (three orthobunyaviruses: Buffalo Creek, Mapputa and Marpik viruses [62]; one tospovirus [63]; and three nairoviruses: Sanxia Water Strider, Shayang Spider and Xinzhou Spider viruses [64]. The two “unclassified ssRNA negative-polarity viruses” are Wuhan Louse Fly virus and Wuhan insect virus 2, which were reported in a recent metagenomic analysis [64] and remain unclassified.