Molecular rationale for antibody-mediated targeting of the hantavirus fusion glycoprotein

The intricate lattice of Gn and Gc glycoprotein spike complexes on the hantavirus envelope facilitates host-cell entry and is the primary target of the neutralizing antibody-mediated immune response. Through study of a neutralizing monoclonal antibody termed mAb P-4G2, which neutralizes the zoonotic pathogen Puumala virus (PUUV), we provide a molecular-level basis for antibody-mediated targeting of the hantaviral glycoprotein lattice. Crystallographic analysis demonstrates that P-4G2 binds to a multi-domain site on PUUV Gc and may preclude fusogenic rearrangements of the glycoprotein that are required for host-cell entry. Furthermore, cryo-electron microscopy of PUUV-like particles in the presence of P-4G2 reveals a lattice-independent configuration of the Gc, demonstrating that P-4G2 perturbs the (Gn-Gc)4 lattice. This work provides a structure-based blueprint for rationalizing antibody-mediated targeting of hantaviruses.


Introduction
Rodent-borne hantaviruses (genus Orthohantavirus, family Hantaviridae) are enveloped, negativesense RNA viruses found worldwide in small mammals (Jonsson et al., 2010;Watson et al., 2014). Cross-species transmission into humans typically results from inhalation of aerosolized excreta from chronically infected rodents and has two primary clinical outcomes. Hantaviral species native to Europe and Asia cause hemorrhagic fever with renal syndrome (HFRS), while those in the Americas cause hantavirus cardiopulmonary syndrome (HCPS) (Lee and van der Groen, 1989;Watson et al., 2014;Zaki et al., 1995). The case mortality rate may exceed 30% for HCPS and ranges from <1% to 15% for HFRS (Alonso et al., 2019;Heyman et al., 2011;Jonsson et al., 2010). Despite variable clinical characteristics, both syndromes arise from excessive proinflammatory and cellular immune responses to infection (Maes et al., 2004;Terajima and Ennis, 2011). Due to their potential to cause severe disease and to be transmitted by aerosols, hantaviruses have been identified as potential bioterrorism agents by the Centers for Disease Control and Prevention (Moran, 2002).
Entry of a hantavirus into a host cell is negotiated by two membrane-anchored glycoproteins, Gn and Gc, which form a lattice decorating the lipid bilayer envelope of the mature virion (Battisti et al., 2011;Hepojoki et al., 2010;Huiskonen et al., 2010). Gn and Gc are translated as a single polypeptide and cleaved at a 'WAASA' signal sequence during protein folding and assembly (Lö ber et al., 2001;Schmaljohn et al., 1987). Recent studies have reported crystal structures of both Gn (Li et al., 2016;Rissanen et al., 2017) and Gc (Guardado-Calvo et al., 2016;Willensky et al., 2016) ectodomains, and their higher-order, tetrameric organization has been postulated based on biochemical characterization  and cryo-electron microscopy reconstructions of purified virions (Battisti et al., 2011;Huiskonen et al., 2010;Li et al., 2016). The N-terminal region of the Gn ectodomain exhibits a mixed a/b fold, which locates to the membranedistal region of the tetrameric spike and is linked to the virion envelope by a C-terminal stalk region (Huiskonen et al., 2010;Li et al., 2016;Rissanen et al., 2017). The Gc adopts a three-domain (domains IÀIII) class II fusion protein architecture and is postulated to associate closely with the Gn, linking adjacent Gn tetramers (Guardado-Calvo et al., 2016;Hepojoki et al., 2010;Li et al., 2016;Willensky et al., 2016). The Gn has been shown to shield the fusion loop resident in domain II of the Gc, in a manner similar to that proposed in phleboviruses and alphaviruses (Allen et al., 2018;Guardado-Calvo et al., 2016;Guardado-Calvo and Rey, 2017;Halldorsson et al., 2018;Li et al., 2016;Serris et al., 2020;Voss et al., 2010). Following host cell binding and endocytotic uptake of the virus, acidification in endosomal compartments results in at least partial dissolution of the GnÀGc lattice (Acuña et al., 2015;Hepojoki et al., 2010;Rissanen et al., 2017). Endosomal escape is mediated by the Gc in a process that involves engagement of the target membrane by the Gc-encoded fusion loop, followed by upward 'zippering action' conformational rearrangements of the transmembrane-anchored Gc domain III, which bring together host and viral membranes (Guardado-Calvo et al., 2016;Willensky et al., 2016).
In contrast to the chronic, asymptomatic infection observed in the native rodent reservoir of a given hantaviral species, human infections are often acute and characterized by a strong immune response (Schö nrich et al., 2008). Indeed, the generation of robust antibody-mediated immune response against Gn and Gc has been shown to be key in limiting disease progression in humans, where high titers of neutralizing antibodies (nAbs) in the acute phase correlate with a more benign course of disease (Bharadwaj et al., 2000;Dantas et al., 1986;Pettersson et al., 2014). Furthermore, a neutralizing humoral immune response also conveys long-lasting immunity against human hantaviral infection (Lundkvist et al., 1993b;Schmaljohn et al., 1990;Settergren et al., 1991).
Although elicitation of a productive nAb response constitutes an essential component of the host immune response to hantavirus infection (Arikawa et al., 1989;Heiskanen et al., 1999;Levanov et al., 2019;Lundkvist and Niklasson, 1992), there is a paucity of information with regard to the molecular determinants of antibody-mediated neutralization. Here we structurally characterize the epitope of mAb P-4G2, a potently neutralizing antibody derived following experimental infection of bank voles with Puumala virus (PUUV), a hantavirus responsible for one of the most frequently occurring hantaviral diseases, a mild form of HFRS known as nephropathia epidemica (NE) (Heyman et al., 2011;Linderholm and Elgh, 2001). Through a combined crystallographic and electron cryo-tomography (cryo-ET) analysis, we report the structure of mAb P-4G2 in complex with PUUV Gc and define the P-4G2 epitope on the Gc subcomponent of the mature GnÀGc spike. This work provides the first molecular-level insights into antibody-mediated targeting of the antigenic hantaviral surface.

Results
Recombinantly derived and Gc-specific bank vole mAb P-4G2 potently neutralizes PUUV The hybridoma cell line producing the Gc-specific neutralizing mAb P-4G2 was reported in 1992 following experimental infection of the natural reservoir species, bank voles (Myodes glareolus), with PUUV (Lundkvist and Niklasson, 1992). While mAb P-4G2 has been shown to be potently neutralizing against PUUV, with a mAb concentration of~4 nM (0.6 mg/ml) (Lundkvist and Niklasson, 1992) resulting in 80% inhibition of PUUV infection in focus reduction neutralization tests (Heiskanen et al., 1997;Lundkvist and Niklasson, 1992;Lundkvist et al., 1993c), there has been a paucity of information with regard to the structural basis for neutralization by this or any other Gcspecific antibody. We rescued the sequence of the P-4G2 variable heavy (V H ) and kappa (V K ) chain regions (see Figure 1-figure supplement 1) from hybridoma cells by PCR with gene-specific primers originally designed for mouse V H and V K chains (von Boehmer et al., 2016). The resulting recombinantly produced mAb P-4G2 comprised bank vole variable and mouse constant regions and, consistent with previously reported in vitro studies of hybridoma-derived mAb P-4G2 Lundkvist and Niklasson, 1992), potently neutralizes PUUV in a hantaviruspseudotyped VSV-DG RFP (Red Fluorescent Protein) neutralization assay at an IC 50 value of 0.088 mg/ml (Figure 1).
Crystal structure of the Fab P-4G2 in complex with PUUV Gc reveals the P-4G2 epitope on pre-fusion Gc To ascertain the molecular basis for P-4G2-mediated recognition of PUUV, we determined the crystal structure of the PUUV Gc ectodomain in complex with the antigen-binding fragment (Fab) of P-4G2 to 3.50 Å resolution ( Consistent with previous studies of hantaviral Gc glycoproteins (Guardado-Calvo et al., 2016;Willensky et al., 2016), PUUV Gc assumes a class II fusion glycoprotein architecture composed of three domains (domains IÀIII), with domain I and domain II forming an elongated structure and domain III forming inter-domain contacts at the side of domain I. Fab P-4G2 recognizes a proteinspecific epitope at the junction of domain I and domain II of PUUV Gc, distal from the hydrophobic fusion loop at the tip of domain II (Figure 2). The epitope comprises~830 Å 2 of buried surface area and all complementarity-determining region (CDR) loops participate in the interaction (Figure 2figure supplement 2). Arg100 (CDRH3) is the most centrally located paratope residue, and is predicted to form multiple hydrogen bonds and a putative salt bridge with Gc residue Glu725 and across the antibodyÀantigen interface ( Figure 2B and Figure 2-figure supplement 2; interacting residues were identified using the PDBePISA server [Krissinel and Henrick, 2007]). Site-directed mutagenesis of Arg100 to Ala considerably reduced the potency of P-4G2 to neutralize PUUV Figure 1. Composition and neutralization potency of recombinantly-derived bank vole mAb P-4G2. (A) Composition of the complementaritydetermining regions (CDRs) of the mAb P-4G2 antigen-binding fragment (Fab) heavy (V H ) and kappa (V K ) chains. For the full sequence of Fab P-4G2 variable regions, please see Figure 1-figure supplement 1. (B) A hantavirus-pseudotyped VSV-DG RFP neutralization assay shows that recombinantly produced mAb P-4G2 neutralizes Puumala virus-and Andes virus-pseudotyped VSV (black and blue traces, respectively), but not Hantaan virus-pseudotyped VSV (gray trace). Each neutralization assay was carried out three times in duplicate. A representative experiment is shown. Error bars represent the range of the value for the experiment performed in duplicate. The online version of this article includes the following figure supplement(s) for figure 1:  Figure 2C). One Fab P-4G2ÀPUUV Gc complex was observed in the crystallographic asymmetric unit, and there was no evidence for the formation of higher-order Gc dimers or trimers, such as those observed in pre-and post-fusion class II fusion glycoprotein structures (Modis, 2013).
Consistent with the hypothesis that mAb P-4G2 targets a pre-fusion conformation of PUUV Gc representative of that displayed on the native virion surface, structure overlay analysis reveals that PUUV Gc from the Gc-P-4G2 complex (PUUV Gc Gc-P-4G2 ) is highly similar to the recently reported pre-fusion state of Andes virus (ANDV) Gc (Serris et al., 2020), and is distinct from the known PUUV Figure 2. Crystal structure of neutralizing antibody P-4G2 in complex with Puumala virus (PUUV) Gc. (A) Crystal structure of Fab P-4G2ÀPUUV Gc complex at 3.5 Å resolution. PUUV Gc, a class II fusion protein, comprises domains IÀIII (colored red, yellow, and blue, respectively), a Gc C-terminal tail (light blue) and the viral fusion loop at Ser771-Thr785 (orange). Fab P-4G2, comprised of a heavy chain (dark gray) and a light chain (white), is observed bound at the junction of domains I and II on PUUV Gc. A domain schematic of the PUUV glycoprotein precursor with the signal peptide (SP), transmembrane domains (TM), intra-virion domain (IV), and WAASA signal peptidase cleavage site is shown alongside the construct used in crystallization (schematic was produced using the DOG software [Ren et al., 2009]). N-linked glycosylation sequons are shown in the domain schematic as pins and the N-linked glycan observed at Asn937 is rendered as sticks. There was no evidence of glycosylation at Asn898. (B) A close-up view of the central portion of the Fab P-4G2-PUUV Gc interface around paratope residue Arg100 of the Fab P-4G2 CDRH3. Residues contributing to a putative hydrogen bonding network (indicated by dashed lines) are rendered as sticks and were identified using the PDBePISA server (Krissinel and Henrick, 2007). (C) Site-directed mutagenesis of Arg100 to Ala reduces the neutralizing potency of mAb P-4G2, confirming the importance of this residue in the interaction interface. We compared the neutralization potency of mAb P-4G2 R100A mutant (magenta trace) to that of the wild-type mAb P-4G2 (black trace) against PUUV pseudovirus in a hantaviruspseudotyped VSV-DG RFP neutralization assay. The virus neutralization assay presented in panel C was conducted as a part of the same experiment as the neutralization assay presented in Figure 1B. Each neutralization assay was carried out three times in duplicate. A representative experiment is shown. Error bars represent the range of the value for the experiment performed in duplicate. The online version of this article includes the following figure supplement(s) for figure 2:   Notably, the conformations of domain III from both the recently determined ANDV Gc (Serris et al., 2020) and our PUUV Gc contrast those observed across hantavirus Gc structures (Figure 3-figure supplement 1), suggesting that this region is likely flexible in the absence of the stabilizing environment of the glycoprotein lattice or crystal environments. Additionally, the conformation of PUUV Gc Gc-P-4G2 closely resembles class II fusion protein pre-fusion states, such as those observed for RVFV Gc (PDB 4HJ1) and ZIKV E (PDB 5IRE) (Figure 3-figure supplement 1).
Investigation into the residues that comprise the P-4G2 epitope revealed that many of these residues are buried in the post-fusion trimeric Gc ( Figure 3D), indicating that mAb P-4G2 is specific to the pre-fusion state of PUUV Gc. This observation is consistent with previous experimental findings, which demonstrated that mAb P-4G2 did not recognize PUUV under low pH conditions but remained bound to the virus and protected the epitope if introduced prior to low-pH exposure . Taken together, these results support a model whereby P-4G2 specifically targets pre-fusion Gc and may sterically preclude the formation of a fusogenic configuration of the Gc.
The epitope of P-4G2 is targeted for neutralization across hantaviral species Hantaviruses are often divided into two groups, termed Old World and New World hantaviruses (Jonsson et al., 2010), reflecting their distribution and pathobiological features. Interestingly, while PUUV is an Old World hantavirus prevalent in Northeastern Europe, the residues comprising the P-4G2 epitope on PUUV Gc exhibit a relatively high level of sequence conservation with Gc proteins  (Willensky et al., 2016). (C) Fab P-4G2 binding precludes the rearrangement of domain III (blue arrow) to the post-fusion conformation. PUUV Gc is represented as a surface, with the epitope of Fab P-4G2 outlined dark gray. Fab P-4G2 is shown in ribbon representation, with heavy and light chains colored gray and white, respectively. (D) Top view of the P-4G2 epitope shows that it overlaps with the neutralization evasion mutation sites reported for PUUV-neutralizing human antibody 1C9 (Hö rling and Lundkvist et al., 1993a) and HTNV-neutralizing mouse antibodies HC02 and 16E6 (Arikawa et al., 1989;Wang et al., 1993). Furthermore, the epitope overlaps with the binding site of domain III in the post-fusion conformation (medium blue). Interfacing residues were identified using the PDBePISA server (Krissinel and Henrick, 2007). The online version of this article includes the following figure supplement(s) for figure 3: from New World hantaviruses, such as ANDV (72% sequence identity across the P-4G2 epitope), and a lower level of conservation with Hantaan virus (HTNV), an Old World hantavirus (38% sequence identity across the epitope) ( Figure 2-figure supplement 3). The epitope-based sequence analysis provides a rational basis for understanding the cross-reactivity of P-4G2 with ANDV but not HTNV, where we identify a greater level of sequence variation on HTNV Gc at the epitope and fewer sequence differences on ANDV Gc ( Figure 2-figure supplement 3). The observed similarity at the epitope between PUUV and New World hantaviruses is in line with the evolutionary history of these viruses and their hosts (Plyusnin and Sironen, 2014), and provides a structural basis for the ability of P-4G2 to neutralize ANDV, albeit at a lower IC 50 value (6.48 mg/ml) ( Figure 1B). Furthermore, despite the sequence variation at the P-4G2 epitope between the two Old World hantaviruses PUUV and HTNV ( Figure 2-figure supplement 3), and the lack of HTNV cross-neutralization ( Figure 1B), we note that the observed P-4G2 binding site overlaps with the putatively assigned binding sites of HTNV-neutralizing mAbs HCO2 and 16E6 (Arikawa et al., 1989;Wang et al., 1993; Figure 3), indicating that this region of the Gc is likely to be immunologically accessible across hantaviral species.
Mab P-4G2 is specific to a lattice-free state of virion-displayed PUUV Gc Previous investigations of the non-pathogenic viral orthologue, Tula virus (TULV), have revealed the ultrastructure of GnÀGc assemblies, as displayed on the hantaviral envelope (Huiskonen et al., 2010;Li et al., 2016). These studies indicate that the Gc forms elongated structures that extend from the virion membrane and are shielded by the cognate Gn (Li et al., 2016).
To ascertain the position of the P-4G2 epitope in the context of the hantaviral envelope, we generated PUUV virus-like particles (VLPs) by transient expression of the PUUV genomic M-segment in mammalian cells (Acuña et al., 2014). Cryo-ET of purified PUUV VLPs, in tandem with sub-tomogram averaging, revealed that the VLP surface displays the expected GnÀGc spike architecture that can form locally ordered lattices, as observed in native hantavirions (Figure 4-figure supplement 1). Consistent with previous studies of hantavirus ultrastructure and Gc oligomerization (Bignon et al., 2019;Li et al., 2016), breakpoints exist in the (GnÀGc) 4 lattices (Huiskonen et al., 2010;Li et al., 2016;Serris et al., 2020). Previous studies have suggested that contacts between (GnÀGc) 4 spikes are mediated by Gc homo-dimers (Bignon et al., 2019;Huiskonen et al., 2010;Li et al., 2016), while lattice breaks are expected to display Gc ('lattice-free Gc') molecules that are not integrated within the lattice assembly through Gc homo-dimer interactions.
Following the validation that PUUV VLP presents a glycoprotein architecture that accurately resembles the native virion, we applied an identical cryo-ET approach to VLPs treated with Fab P-4G2. The data were split in two sets: (i) spikes that were part of a lattice, and (ii) spikes in regions of incomplete lattice. The former yielded a reconstruction similar to the VLP sample prepared without the Fab ( Figure 4A and Figure 4-figure supplement 1), while the reconstruction focused on lattice-free spikes yielded a discrete (GnÀGc) 4 spike with additional density ( Figure 4B). These results suggest that mAb P-4G2 binding is incompatible with glycoprotein lattice formation, and further, that Fab P-4G2 binding may affect the higher order lattice assembly. To test this hypothesis, we quantified the number of neighbors for each spike on VLPs both in the presence and absence of Fab P-4G2. The frequency of lattice-free spikes (with zero neighbors) was higher in the presence of Fab P-4G2, while the frequency of lattice-bound spikes (with four or more neighbors) was higher in the particles not treated with Fab P-4G2 ( Figure 4C and D and Figure 4-figure supplement 2). We note that treatment with Fab P-4G2 had a heterogeneous effect on the lattice of different VLPs, as can be observed in Figure 4-figure supplement 2. These results suggest that Fab P-4G2 influences the presentation of the GnÀGc lattice by dislodging the (GnÀGc) 4 spikes from each other.
Fitting of the crystal structure of Fab P-4G2ÀPUUV Gc complex and PUUV Gn (Li et al., 2016) into the reconstruction of the Fab-treated (GnÀGc) 4 spike confirms that the additional density is comprised of Fab P-4G2 ( Figure 5). PUUV Gc GcÀP-4G2 fits well within the cryo-ET reconstruction, where the crystallographically observed conformation of domain III accurately matches a feature of the glycoprotein spike ( Figure 5C), further supporting that PUUV Gc has crystallized in the prefusion conformation presented on the mature virion. Additionally, this fitting confirms the location of Gc in the density spanning from the membrane-distal globular lobes to the viral membrane, as we and others have proposed previously (Li et al., 2016;Serris et al., 2020). The fitting locates domain III of the Gc adjacent to the membrane, allowing the transmembrane region of the C-terminus to be linked to the virion envelope ( Figures 5A and 2A), and places the fusion loop from Gc domain II into membrane-distal lobe density in close contact with Gn ( Figure 5A and B), in a manner similar to that observed for other class II fusion proteins and their cognate accessory proteins Voss et al., 2010).
Drawing upon the fitting generated above, we used a 13.9 Å cryo-ET reconstruction of the lattice-incorporated (non-Fab P-4G2 treated) PUUV VLP surface to create a model of the higher-order (GnÀGc) 4 spike lattice. While the overall assembly of (GnÀGc) 4 spikes remain similar to that obtained from fitting into the cryo-ET reconstruction of an individual, Fab-treated glycoprotein spike ( Figure 5) fitting into the latticed reconstruction provides a structural basis for the formation of   . Fitting of the Fab P-4G2ÀPuumala virus (PUUV) Gc crystal structure into the cryo-ET reconstruction confirms the P-4G2 epitope in the context of the viral surface and supports the hypothesis that the observed Gc conformation constitutes a pre-fusion state. (A) Side view and (B) top view of the Fab P-4G2-treated PUUV virus-like particle (VLP) spike at 13.4 Å resolution. Crystal structures of Fab P-4G2ÀPUUV, along with PUUV Gn (PDB id 5FXU), were fitted in the cryo-ET reconstruction as rigid bodies and display excellent conformity with the cryo-ET derived envelope. (C) A zoom-in of the fit of domain III of Gc into the cryo-ET reconstruction presented from two points of view, which are related by a 180˚rotation. The goodness of fit supports the hypothesis that domain III has crystallized in a conformation that closely resembles the pre-fusion state presented in the mature hantaviral spike. (D) A close-up of Fab P-4G2 similarly presents a high-correlation fit. A simplified look at the composition of the hantaviral spike, where fitted Gc and Gn are represented as surfaces, is shown from side view (E) and top view (F).
PUUV Gc homo-dimers at contact points between neighboring spikes (Figure 6), which closely resemble a crystallographically observed homo-dimer interface reported previously for HTNV Gc (Bignon et al., 2019;Guardado-Calvo et al., 2016; Figure 6-figure supplement 1). Furthermore, modeling of Fab P-4G2 binding onto the Gc homodimer demonstrates that neighboring epitopes are not mutually accessible for Fab binding in their lattice-integrated form, providing clues to how Fab P-4G2 recognition may be accompanied by dissociation of the Gc homodimer assembly (Figure 6-figure supplement 2). Altogether, our combined crystallographic and cryo-EM investigation reveals, in molecular detail, that P-4G2 is specific to the lattice-free, pre-fusion state of PUUV Gc presented on the mature viral envelope.

Discussion
Despite the importance of neutralizing antibodies in mitigating the progression of hantaviral disease (Bharadwaj et al., 2000;Dantas et al., 1986;Pettersson et al., 2014) and conveying long-lasting protection against infection (Settergren et al., 1991;Valdivieso et al., 2006;Ye et al., 2004), little is known about how they target virion-displayed glycoproteins, Gn and Gc. Here we provide a molecular-level description of the epitope targeted by mAb P-4G2, a host-derived antibody that potently neutralizes PUUV, a zoonotic hantavirus prevalent in Northern Europe and Russia. We find that mAb P-4G2 targets an epitope at the junction of domains I and II of PUUV Gc (Figure 2). When combined with a cryo-ET reconstruction of Fab P-4G2 complexed with PUUV (GnÀGc) 4 spike, our analysis reveals that mAb P-4G2 targets a pre-fusion, lattice-free conformation of PUUV Gc.
Our structural data provides several molecular-level insights into hantavirus Gc functionality and antibody-mediated neutralization. First, comparison of our PUUV Gc pre-fusion state with that of the previously reported PUUV Gc post-fusion state clarifies the conformational changes that PUUV Gc undergoes to facilitate membrane fusion (Figure 3 and    (Dessau and Modis, 2013) and in the context of the virion , which is consistent with domain III flexibility being an intrinsic feature required for Gc glycoprotein functionality.
Second, examination of the antibodyÀantigen interface provides a structure-based model for understanding the mechanism of P-4G2-mediated neutralization, where overlay analysis reveals that Fab P-4G2 binding is incompatible with the pH-dependent conformational transitions of PUUV Gc required to negotiate fusion of the viral and cellular membranes. While it is also possible that mAb P-4G2 interferes with other steps of the host-cell entry pathway (e.g. receptor recognition), our modeling indicates that the antibody targets residues crucial for the formation of a post-fusion Gc configuration (Figure 3). Indeed, binding to this site likely impedes the upward-occurring 'zippering action' of domain III during Gc-mediated viral fusion (Guardado-Calvo et al., 2016;Willensky et al., 2016). This proposed structure-based mechanism of neutralization is supported by previously reported in vitro studies, which demonstrate that mAb P-4G2 protects the epitope upon low-pH exposure . Further support for this hypothesis can be gleaned from previous reports, which have shown that hindering such movements of domain III, via either mutagenesis or the introduction of competing peptides (Barriga et al., 2016), drastically reduces viral infectivity. Given that small molecule ligands have also been shown to prevent fusogenic rearrangements on the analogous DIÀDII interface region of flavivirus E glycoproteins (Chao et al., 2018;Modis et al., 2003), which also bear a class II fusion architecture, it will be interesting to assess whether small molecule inhibitor development may be similarly successful against this region on the hantavirus Gc.
In line with the hypothesis that the identified epitope on Gc domains I and II may constitute a site of vulnerability across hantaviruses, more generally, structure-based mapping revealed that the P-4G2 binding site overlaps with epitopes putatively targeted by neutralizing antibodies specific to HTNV and PUUV. These antibodies, derived following hantaviral infection, include mouse mAbs HC02 and 16E6 (Arikawa et al., 1989;Wang et al., 1993) that neutralize HTNV, and human mAb 1C9 (Hö rling and Lundkvist et al., 1993a), which neutralizes PUUV (Figure 3). Thus, while our identified epitope appears to be a common target of the neutralizing antibody response, future investigations into the molecular basis of immune recognition will undoubtedly be required to fully define the antigenic topography of Gn and Gc. Such investigations will also likely clarify whether the neutralizing epitope defined here is similarly targeted upon immunization, and whether the Ab-immune response is differentially focused against Gn and Gc between reservoir species (e.g. bank vole) and accidental (i.e. humans) hosts.
Lattice integrated and lattice-free Gc are expected to transiently exist on the hantavirus envelope: contacts between (GnÀGc) 4 spikes are mediated by Gc homo-dimers, while lattice breaks display Gc that is not engaged in discernible homotypic interactions (Bignon et al., 2019;Huiskonen et al., 2010;Li et al., 2016). In our cryo-ET reconstruction, Fab P-4G2 was found to bind PUUV Gc only in the lattice-free state, and the frequency of lattice-free spikes was higher in the presence of Fab P-4G2. This observation demonstrates that P-4G2 reduces the frequency of Gc-mediated lattice formation. While the bivalent binding mode of full-length mAb P-4G2, with respect to the structurally characterized Fab P-4G2, may result in a different extent of lattice deformation, it is likely that both the mAb and Fab function by either active destabilization of Gc-Gc interfaces through competition with adjacent epitopes (Figure 6-figure supplement 2), or by sequestration of naturally occurring free spikes in a process that drives the equilibrium between homo-dimeric and lattice-free Gc toward the lattice-free state.
Characterization of the Fab P-4G2 epitope has important implications for rational therapeutic design efforts and indicates that lattice-free conformation of the Gc can be targeted by the antibody-mediated immune response that arises during infection. Indeed, although the development of recombinant immunogens that recreate the higher-order hantaviral (GnÀGc) 4 lattice is desirable and would likely elicit a robust immune response, our data also indicate that a reverse vaccinology approach (Burton, 2017) focusing on recombinant and more minimal (monomeric) hantaviral Gn or Gc may also elicit anti-hantaviral nAbs. Indeed, such an approach has been successful in eliciting neutralizing antibodies against the glycoproteins of other bunyaviruses, including Rift Valley fever virus (phlebovirus) (Allen et al., 2018) and Schmallenberg virus (orthobunyavirus) (Hellert et al., 2019).
At present, treatment and prevention options for hantaviral disease remain extremely limited due to the absence of approved therapeutics. However, strides are being made toward the development of efficacious treatments, as shown by the recent development of novel mAbs against ANDV (Duehr et al., 2020;Garrido et al., 2018), and a polyclonal antibody treatment against HTNV (Hooper et al., 2014), both of which demonstrate efficacy in animal models. In toto, our integrated structural and functional analysis provides a first structure-based blueprint for targeting the glycoprotein surface of this group of important emerging pathogens.

Materials and methods
Key resources Sequencing mAb P-4G2 variable regions from hybridoma cell line RNA from the P-4G2 hybridoma cell line was converted into cDNA (SuperScript III reverse transcriptase, Life Technologies) using random hexamers following the manufacturer's protocol. The bank vole antibody variable regions of heavy and kappa chains were PCR amplified using previously described mouse primers and PCR conditions (von Boehmer et al., 2016). PCR products were purified and cloned into mouse IgG expression plasmids (von Boehmer et al., 2016) using sequence and ligation independent cloning (SLIC) under ampicillin selection. Antibody variable regions were sequenced by Sanger sequencing, and CDRs were determined from the sequences with IMGT/V-QUEST (Brochet et al., 2008) server using Mus musculus as comparison group.
Antibody heavy and light plasmids were co-transfected at a 1:1 ratio into HEK 293F cells (Thermo Fisher Scientific) using PEI Max 40K (linear polyethylenimine hydrochloride, Polysciences, Inc), as previously described . Antibody supernatants were harvested 7 days following transfection and purified using protein G affinity chromatography following the manufacturer's protocol (GE healthcare). In order to validate the functional relevance of Arg100, the R100A mutation was introduced into the P-4G2 heavy chain plasmid using site directed mutagenesis (forward primer: GTATTACTGTACAAGAGATGCATTAGGCCCTTTTGA, reverse primer: TCAAAAGG GCCTAATGCA TCTCTTGTACAGTAATAC) and the mutation was verified using Sanger sequencing. Mab P-4G2 R100A mutant was expressed and purified as described above for WT mAb P-4G2.
Two additional rounds of transfection and virus production were carried out. 15 Â 10 6 HEK 293T cells were transfected with a plasmid encoding PUUV GPC, ANDV GPC, or HTNV GPC (100 mg, in pCAGGS vector) and Tim-1 (6 mg) using PEI (3:1 PEI: DNA, Polysciences, Inc). After 24 hr, cells were infected with 400 ml of 'first stock' virus. Infection was monitored using RFP expression and virus was harvested (second stock) when cells showed rounding and began to detach (48-72 hr). This process was repeated to generate 'third stock' virus which was used in subsequent neutralization assays.

Hantavirus-pseudotyped VSV neutralization assay
Neutralizing activity was assessed using a single-round replication pseudovirus assay with HEK 293T target cells. Briefly, the antibody was serially diluted in a 96-well black, flat-bottom plate and preincubated with virus for 1 hr at 37˚C. Cells at a concentration of 30,000/well were added to the virusantibody mixture, and RFP quantified 72 hr following infection (Envision plate reader, Perkin Elmer). Dose-response curves were fitted using nonlinear regression (GraphPad Prism) to determine 50% inhibitory concentration (IC 50 ).

Recombinant protein expression and purification
Codon-optimized synthetic cDNA (GeneArt, Life Technologies) coding for Fab P-4G2 light chain and Fab P-4G2 heavy chain were individually cloned into the pHLsec mammalian expression vector and co-transfected into HEK 293T cells (ATCC CRL-3216) for transient expression, as previously described (Avanzato et al., 2019). Fab P-4G2-containing supernatant was collected, clarified by centrifugation, and diafiltrated using the Ä KTA Flux tangential flow filtration system. Diafiltrated cell supernatant was purified by immobilized nickel affinity chromatography (5 ml fast flow crude column and Ä KTA fast protein liquid chromatography [FPLC] system; GE Healthcare) at room temperature, using 250 mM imidazole for elution. Finally, the sample was purified by size exclusion chromatography (SEC) using a Superdex 200 10/300 Increase column (GE Healthcare), in 10 mM Tris (pH 8.0)-150 mM NaCl buffer.
In order to produce PUUV Gc, a HEK 293T cell line stably expressing the protein was generated. PUUV Gc residues 659-1105 (GenBank accession no. CAB43026.1), modified to contain a 3C protease cleavable SUMO tag and hexahistidine tag at the N-terminus and a 1D4 tag in the C-terminus (Chang et al., 2015), were cloned into the pURD vector (Zhao et al., 2014). The stable cell line was generated following the method published by Seiradake et al., 2015; in short, HEK 293T cells were transfected with the pURD vector containing the Gc insert, along with an integrase expression vector (pgk-jC31/pCB92) (Chen et al., 2011), followed by consecutive rounds of cultivation under puromycin selection. For crystallization, PUUV Gc-producing cells were cultivated in the presence of kifunensine (Chang et al., 2007), and purification was performed via successive diafiltration, nickel affinity, and SEC, as described above for Fab P-4G2. Oligomannose glycans derived from expression in the presence of kifunensine were retained in the purified Gc sample.
Prior to crystallization, the N-terminal SUMO and hexahistidine tags were cleaved from the PUUV Gc (1:10 molar ratio of protein to 3C protease added to purified sample, followed by incubation at 21˚C for 12 hr). PUUV Gc sample was then mixed with pure Fab P-4G2 in 1:1.2 molar ratio, and the complex was purified by SEC using a Superdex 200 10/300 Increase column (GE Healthcare), in 10 mM Tris (pH 8.0)-150 mM NaCl buffer.

Crystallization and structure determination
A solution containing a complex of PUUV Gc ectodomain and Fab P-4G2 was crystallized by sitting drop vapor diffusion using 100 nl protein (1.3 mg/ml) with 100 nl precipitant (20% vol/vol PEG 6000, 0.1 M MES pH 6, and 1 M lithium chloride) and 100 nl additive (6% 2-methyl-2,4-pentanediol) at room temperature. Crystals formed over 2 weeks and were flash frozen by immersion into a cryoprotectant containing 25% (vol/vol) glycerol followed by rapid transfer into liquid nitrogen. X-ray diffraction data were recorded at beamline I24 (l = 0.9686 Å ), Diamond Light Source, United Kingdom.
Data were indexed and integrated with XIA2/XDS (Winter, 2010) and scaled using CCP4/SCALE-PACK2MTZ. Processing statistics are presented in Supplementary file 1. The structure of Fab P-4G2ÀPUUV Gc was solved by molecular replacement in Phenix-MR (Adams et al., 2002) using PUUV Gc (PDB 5J81) and a homology model of the Fab P-4G2, generated using the SWISS-MODEL server (Waterhouse et al., 2018) with PDB 6BPB as the template structure, as search models. Structure refinement was performed by iterative refinement using REFMAC and Phenix (Adams et al., 2002). Coot (Emsley and Cowtan, 2004) was used for manual rebuilding and MolProbity  was used to validate the model.

Cell lines
Authenticated cell lines were obtained from ATCC (HEK 293T cells) or, in the case of FreeStyle 293F cell line, from the manufacturer Thermo Fisher, and were tested for mycoplasma contamination and found negative.

Molecular graphics and protein interface analysis
Molecular graphics images were generated using PyMOL (The PyMOL Molecular Graphics System, Version 1.7.0.3, Schrö dinger, LLC) and UCSF Chimera (Pettersen et al., 2004). Interfacing and interacting residues were identified using the PDBePISA server (Krissinel and Henrick, 2007) run with default parameters, where the maximum allowable distance between the heavy atoms of a hydrogen bond donor and acceptor was limited to 3.89 Å , and the angle between acceptor-hydrogen-donor atoms was limited to between 90 and 270˚. The maximal allowable distance between heavy atoms of a salt bridge was 4.0 Å . LigpPlot+ (Laskowski and Swindells, 2011) was used to visualize the Fab P-4G2ÀGc interface (Figure 2-figure supplement 2).

Preparation of PUUV VLPs
PUUV VLPs were produced by transient expression of the complete PUUV M-segment (GenBank CCH22848.1) cloned into the pHLsec vector (Aricescu et al., 2006) in HEK 293T cells. Six five-layer 875 cm 2 flasks (Falcon) were used to produce 750 ml of VLP-containing media that was clarified at 3000 Â g for 20 min to remove cell debris and filtered through a 0.45 mm filter. The virus containing medium was concentrated down to approximately 30 ml using a pump-powered filter (100 kDa cutoff; Vivaflow, Sartorius) and then dialyzed into an excess of buffer (10 mM Tris pH 8.0-150 mM NaCl) through a 1-MDa cut-off dialysis membrane (Biotech CE Tubing, Spectrum Chemical) for a few days. The media was further concentrated to~3 ml with a 100 kDa cut-off centrifugal concentrator filter (Amicon Ultra, Merck Millipore) and layered onto a 20-60% w/v sucrose density gradient in PBS buffer. The gradient was prepared using a Gradient Master (BioComp Instruments, Canada) in a SW32 Beckman tube, and the VLP banded by ultracentrifugation at 4˚C for 4 hr at 25,000 rpm. The diffuse band (volume of~3-4 ml) was collected manually, diluted to 20 ml with PBS and pelleted through a cushion of 10% sucrose in PBS to further clean and concentrate the sample (SW32 Beckman centrifuge tube at 25,000 rpm at 4˚C for 2 hr). Finally, the pellet was resuspended in 60 ml of PBS and stored at 4˚C.
Cryo-EM grid preparation, data acquisition, and data processing A 3 ml aliquot of VLP sample supplemented with 3 ml of 6 nm gold fiducial markers (Aurion) was applied on a holey carbon grid (2 mm hole diameter, C-flat, Protochips) that had been glow discharged in a plasma cleaner (Harrick) for 15 s. The grids were blotted for 3-4 s and plunged into an ethane/propane mixture using a vitrification apparatus (Vitrobot, Thermo Fisher Scientific). For Fab P-4G2-treated VLP EM sample preparation, a suspension of purified VLPs was incubated with 2.71 mM Fab P-4G2 for 1 hr at room temperature prior to grid preparation.
Data were collected using a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operated at 300 kV and at liquid nitrogen temperature. Tomo4 software was used to acquire tomographic tilt data on a direct electron detector (K2 Summit; Gatan) mounted behind an energy filter (0-20 eV; QIF Quantum LS, Gatan). Tilt series were collected from -30 to 60˚in 3˚increments with a dose of 4.5 e À /Å 2 per tilt and six frames in each movie (Supplementary file 1).
Movie frames were aligned and averaged using Motioncor2 to correct for beam induced motion (Zheng et al., 2017). Contrast transfer function (CTF) parameters were estimated using CTFFIND4 (Rohou and Grigorieff, 2015) and a dose-weighting filter was applied to the images according to their accumulated electron dose as described previously (Grant and Grigorieff, 2015). These preprocessing steps were carried out using a custom script named tomo_preprocess  available at https://github.com/OPIC-Oxford/TomoPreprocess). Tilt images were then aligned using the fiducial markers, corrected for the effects of CTF by phase flipping and used to reconstruct 3D tomograms in IMOD (Mastronarde and Held, 2017). The amplitudes of the subvolumes cropped from the tomograms were weighted to correct for the low-pass filtering function resulting from dose-weighting of the original images using a custom script (available at https://github.com/OPIC-Oxford/TomoPreprocess).
Subvolume averaging was performed in Dynamo following previously established procedures (Huiskonen et al., 2014;Li et al., 2016). Briefly, initial particle locations ('seeds') were created by modeling the VLP membrane surface using the Dynamo's tomoview program. Subvolumes extracted from these locations were iteratively refined against their rotational average, calculated around the membrane model surface normal at each position, to accurately align their locations onto the membrane. Subsequent refinements were carried out using a map of the TULV GP spike (EMDB-4867) as an initial template. The template was low-pass filtered to 50 Å frequency to avoid model bias. Overlapping particles were removed based on a distance filter (106 Å ) and cross-correlation threshold after each iteration.
A custom script ; available at https://github.com/OPIC-Oxford/PatchFinder) was used to divide the spikes into those that were part of a lattice and those that were not. The script locates the eight closest spikes for each particle and iterates through each with two checks. First, the position of the neighbor must be within a given distance (49 Å ) of the ideal values determined from the positions of neighbors in the previously reported hantaviral reconstruction (EMD code; EMD-4867) and the orientation of the neighbor must match the parent particle within a set tolerance (25˚). Spikes were defined as being part of a lattice if they had at least three interacting neighbors. This classification approach allowed reconstructing the structure of the lattice-bound spike for both VLP datasets (with and without Fab P-4G2) in addition to a lattice-free, P-4G2-bound spike in the case of the Fab P-4G2 sample (Supplementary file 1). In the latter case, the refinement was repeated using P-4G2-bound spike as a starting model and by excluding weakly correlating particles (below 0.15). This led to a reconstruction with a lattice-free spike fully decorated with Fab P-4G2 around its perimeter (Supplementary file 1). The final maps were filtered to the resolution determined by FSC (0.143 threshold) and rendered as isosurfaces. The surface threshold value was determined according to the molecular weight of the viral ectodomain proteins (and additional Fab) assuming an average protein density of 0.81 Da/Å 3 .
Quantification of the frequency of lattice formation on VLP surface in the presence and absence of Fab P-4G2 The positions of each spike and the number of interacting neighbors were analyzed in PUUV VLPs in the presence and absence of Fab P-4G2 from eight VLPs from each cryo-ET dataset. To facilitate the analysis, only VLPs displaying good glycoprotein coverage and roughly spherical morphology were included. Both datasets were processed using the workflow described above, and the PatchFinder script was applied to locate those spikes that were part of a lattice. The particles on the top and bottom of the virus were then removed from further analysis by restricting the second Euler angle (tilt) to 45-135˚. This was done in order to exclude those spikes that are facing the air-water interface, which can in itself result in a loss of visible lattice due to its denaturing effect. The number of neighbors for each spike was quantified and the 3D locations of the spikes were plotted in 2D using Mercator projections (Figure 4-figure supplement 2).

Fitting of the crystal structure into the cryo-EM map
The crystal structure of Fab P-4G2ÀPUUV Gc was docked into the reconstruction of the Fab-decorated PUUV Gn-Gc spike using the fitmap function in Chimera (Pettersen et al., 2004). A map was simulated for the crystal structure to the resolution of 13.4 Å , to match the resolution of the spike reconstruction, and fitting was performed using an initial global search followed by a local fitting that accounts for the C4 symmetry of the spike. The top solution (presented in Figure 5) yielded a local map-to-map correlation score of 0.91 (Supplementary file 1). In order to complete our model of the hantaviral spike, we then added PUUV Gn to the fit based on the recently reported ANDV GnÀGc complex crystal structure (PDB 6Y5F) that describes the native GnÀGc assembly. In short, MatchMaker function in Chimera was used to superpose ANDV GnÀGc crystal structure on the docked Fab P-4G2ÀPUUV Gc structure, and PUUV Gn (PDB 5FXU) was then similarly superposed on the ANDV GnÀGc structure. Models of PUUV Gc (docked) and PUUV Gn (placed according to the known GnÀGc assembly from PDB 6Y5F) were then combined in order to generate a GnÀGc model that could be reliably fitted into a reconstruction of lattice-incorporated PUUV VLP surface, where no Fab P-4G2 density is present. Local fitting of this PUUV GnÀPUUV Gc assembly into a reconstruction of lattice-incorporated PUUV VLP surface, using the fitmap function in Chimera, yielded a correlation score of 0.89 (PUUV VLP without Fab; map for structure simulated to 13.9 Å resolution). C4 symmetry was then applied to generate (Gn-Gc) 4 spike assemblies, which were docked to the densities of the central spike and the neighboring spikes in Chimera. Finally, clashes at spike interfaces were resolved by sequential fitting of the interfacing components. The resulting model of the Gn-Gc assemblies that comprise the hantaviral lattice is presented in Figure 6.

Accession codes
Atomic coordinates and structure factors of the PUUV GcÀFab P-4G2 complex have been deposited in the PDB (accession code 6Z06). Cryo-EM reconstructions of the PUUV VLP surface alone and in the presence of Fab P-4G2 from areas of continuous and discontinuous lattice have been deposited in the EMDB at the EBI under accession codes EMD-11966, EMD-11965, and EMD-11964, respectively. Coordinates of protein structures fitted into these cryo-EM reconstructions have been deposited in the PDB database (accession code 7B09 for the Fab P-4G2-PUUV Gc and PUUV Gn fitted into EMD-11964, and accession code 7B0A for PUUV Gc and PUUV Gn fitted into EMD-11966, respectively). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Additional files

Supplementary files
. Supplementary file 1. Table S1: Crystallographic data collection and refinement statistics for Fab P-4G2ÀPuumala virus Gc. Table S2. Cryo-EM tomography data collection, sub-tomogram reconstruction, and fitting statistics.
. Transparent reporting form

Data availability
Atomic coordinates and structure factors of the PUUV Gc-Fab P-4G2 complex crystal structure have been deposited in the PDB under accession code 6Z06. Cryo-EM reconstructions of the PUUV VLP surface alone and in the presence of Fab P-4G2 from areas of continuous and discontinuous lattice have been deposited in the EMDB at the EBI under accession codes EMD-11966, EMD-11965 and EMD-11964, respectively. Coordinates of protein structures fitted into these cryo-EM reconstructions have been deposited in the PDB database (accession code 7B09 for the Fab P-4G2-PUUV Gc and PUUV Gn fitted into EMD-11964, and accession code 7B0A for PUUV Gc and PUUV Gn fitted into EMD-11966, respectively). The following authors authored the data deposited to PDB and EMDB, as described above.
The following datasets were generated: