Structural basis for protocadherin-1 recognition by New World hantaviruses

Hantaviruses are a large and globally distributed family of viruses associated with severe zoonotic disease in humans. The New World hantaviruses (NWHs) ANDV and Sin Nombre virus (SNV) cause hantavirus cardiopulmonary syndrome (HCPS) in South and North America, respectively, with high case-fatality rates that can approach 40%. Cardiogenic shock and respiratory failure are hallmarks of HCPS. Old World hantaviruses (OWHs) are endemic in Europe and Asia or globally distributed in some cases (Seoul virus (SEOV)). The OWH Hantaan virus (HTNV), Puumala virus (PUUV) and Dobrava-Belgrade virus (DOBV) are associated with severe zoonotic disease and considerable mortality. Although hantaviruses are typically transmitted to humans through inhalation of aerosolized excreta from infected rodents, ANDV appears uniquely capable of human-to-human transmission^2,3, raising its global public health significance and epidemic potential. This risk was underscored by the recent ANDV outbreak on a cruise ship departing Ushuaia, Argentina⁴. No US Food and Drug Administration (FDA)- or European Medicines Agency (EMA)-approved medical countermeasures are available to prevent or treat hantaviral infections.

Hantavirus glycoprotein spikes are formed by two glycoproteins, Gn and Gc, which are organized into a partial lattice at the surface of the virion. Each spike consists of four Gn molecules that form a central stabilizing core surrounded by four Gc molecules at the periphery. Gc mediates the interspike contacts that stabilize the lattice. Gn consists of a head (Gn^H) and a base (Gn^B) domain. Gn^H is exposed at the viral surface and is thought to participate in receptor binding, whereas Gn^B is buried within the spike and mediates most of the intraspike interactions. Gc consists of three domains that drive membrane fusion. At neutral pH, Gn^H interacts with Gc, forming a pre-fusion complex that prevents premature fusion. After exposure to acidic pH, the Gn–Gc complex dissociates, allowing Gc to insert three hydrophobic loops into the host cell membrane and undergo conformational rearrangements that drive fusion of viral and cellular membranes. Structural studies have reported the structures of Gn^H, Gn^B, Gc, the Gn^H–Gc complex and the Gn–Gc spike from several hantaviruses^5,6,7,8,9,10.

Understanding how hantaviruses recognize their cellular receptors is a critical step toward the rational development of therapies and vaccines. Previous studies have identified several cell-surface molecules as candidate entry host factors, including integrins β1–3 (refs. ^11,12,13,14), decay accelerating factor (DAF, encoded by CD55)^15,16, complement factors (gC1qR)¹⁷ and the cadherin-superfamily protein PCDH1 (ref. ¹). However, only PCDH1 has been shown to directly recognize the ANDV Gn–Gc spike^1,18 and to be required for infection of human endothelial cells—major targets of hantavirus infection^1,19. Furthermore, PCDH1 is critical for ANDV infection in vivo and for disease pathogenesis in a lethal hamster model of ANDV challenge^1,18. PCDH1 is a type I membrane protein expressed in human airway epithelial and endothelial cells. It contains seven extracellular cadherin-repeat (EC) domains that extend into the extracellular space. Binding studies have shown that PCDH1 does not recognize soluble monomeric glycoproteins but instead binds to the tetrameric glycoprotein spikes displayed on the viral surface^1,18. Monoclonal antibodies targeting the first EC repeat (EC1) inhibited entry of a recombinant vesicular stomatitis virus (rVSV) pseudotyped with Gn–Gc from multiple NWHs¹, and point mutations at the tip of EC1 prevented PCDH1’s interaction with Gn–Gc of rVSVs bearing the glycoproteins of ANDV (rVSV-ANDV)^1,18. Together, these findings indicate that NWHs infect host cells by engaging the EC1 domain of PCDH1. However, the structural determinants of this interaction remain unclear. Here we sought to determine the structure of PCDH1 bound to the ANDV Gn–Gc spike.

PCDH1 binds to viral particles but not to the soluble Gn^H–Gc heterodimer^1,18, suggesting that the receptor-binding site is presented only on viral spikes. To elucidate the structural basis for PCDH1 recognition, we designed a soluble version of the ANDV spike tetramer by adding a hetero-octamerization domain at the C-terminal end of the Gn and Gc ectodomains (Fig. 1a). The octamerization sequence was derived from HALC4_136, a de novo tetramerization domain comprising three antiparallel helices per protomer²⁰. We split this scaffold by fusing the first two helices to Gn and the third helix to Gc, so that co-expression of Gn and Gc would reconstitute a hetero-oligomeric assembly with the stoichiometry of the native spike (Extended Data Fig. 1a). We cloned both genes into a bicistronic plasmid, purified the complex from the supernatants of stably transfected Drosophila melanogaster S2 cells using affinity and size-exclusion chromatography and confirmed that it interacts with a truncated PCDH1 construct consisting of wild-type EC domains 1 to 4 (sEC1–4^WT) using an enzyme-linked immunosorbent assay (ELISA; Fig. 1b). As a negative control, we used a construct containing the substitutions Y81A and F83A (sEC1–4^MUT), which are known to abrogate virus binding¹⁸. We also confirmed that incubation of the soluble spike with an excess of soluble sEC1^WT, but not sEC1^MUT, prevented the binding to sEC1–4^WT (Fig. 1c). Furthermore, we used biolayer interferometry (BLI) to demonstrate specific spike–EC1 binding. Specifically, sEC1–4^WT or sEC1–4^MUT were immobilized on sensors, which were then dipped into solutions containing increasing concentrations of soluble spike (Fig. 1d). BLI assays performed in the opposite orientation corroborated these findings (Extended Data Fig. 1b).

To verify that the soluble spike proteins fold like the spikes observed on the viral surface, we performed cryo-electron microscopy (cryo-EM). We observed a few two-dimensional (2D) classes compatible with Gn–Gc tetramers (Extended Data Fig. 1c). However, a low-resolution three-dimensional (3D) reconstruction obtained from them was not compatible with the known ANDV spike structure⁵ (Extended Data Fig. 1d), suggesting that, in the absence of lateral contacts between spikes and the transmembrane regions, the intersubunit contacts within the spike are not sufficiently stable to maintain the intact eight-subunit assembly. We next hypothesized that prolonged incubation of the two proteins might help to stabilize the spike in its receptor-binding conformation. To test this hypothesis, we co-transfected S2 cells with the plasmids encoding sEC1–4^WT and the soluble spike and purified the complex using double-affinity and size-exclusion chromatography. We performed cryo-EM studies and observed many new 2D classes showing tetramers (Extended Data Fig. 2 and Extended Data Table 1). Using particles from these classes, we reconstructed a map at a resolution of 2.4 Å, and we used the map to build a model of the spike that closely resembles the tetramers observed on viral particles⁵. The only notable differences between the two models are a slight rotation of the Gn^H domains relative to Gn^B and the poorly resolved density for domain III in the soluble spikes (Extended Data Fig. 3).

The cryo-EM map of the complex revealed an additional density at the centre of the spike, consistent in size with a single EC domain of PCDH1. However, as the map was processed using C₄ symmetry, and the symmetry axis passes through the centre of the spike, we were unable to model this residual density. We reprocessed the map by combining a masked 3D classification with a non-uniform refinement using a pseudo-C₄ symmetry. This approach yielded a new map at a resolution of 2.6 Å (Fig. 1e), in which we modelled PCDH1 EC1 at the centre of the spike (Fig. 1f,g). The final model showed the EC1 domain deeply inserted into a pocket formed by four protomers of Gn, establishing contacts with two neighbouring protomers (660 and 200 Ų, respectively) and sparse contacts with the other two (80 Ų and 35 Ų). This mode of interaction, whereby one receptor molecule interacts with four viral glycoproteins, represents a binding mode that to our knowledge has not been observed in any enveloped virus, and explains why PCDH1 binds to viral particles but not to soluble Gn^H–Gc heterodimers. In the model, EC1 loops BC^EC1 and FG^EC1 interact with the AB and CD loops of Gn^H and the DE loop of Gn^B from two neighbouring protomers (Fig. 1g and Extended Data Fig. 4a). Notably, the densities corresponding to the DE and CD loops are weaker in the protomers that do not engage EC1 than in the receptor-contacting protomers (Extended Data Fig. 3b). This suggests that these loops are flexible and that PCDH1 binding stabilizes specific conformations of these regions. Indeed, comparison of the four protomers reveals that EC1 binding induces a conformational change in the CD and DE loops of one protomer, creating asymmetry within the pocket (Extended Data Fig. 3b). Previous studies have identified PCDH1 mutations that inhibit virus binding¹⁸, and all are in the BC^EC1 loop or nearby regions, probably altering the conformation of the loop. Consistent with the structural data, mutations in the tip of the FG loop had little (D142A, D142R) or no (L143A, V144A, Q145A) effect on virus binding¹⁸, probably because none of them introduce steric restrictions (Fig. 1g).

To validate the model with viral particles, we studied the interaction between PCDH1 and virus-like particles (VLPs). We did not attempt to decorate preformed VLPs with soluble receptor because flow cytometry revealed only weak binding of EC1 to ANDV glycoproteins presented on the cell surface. Instead, reasoning that increasing the local concentration of receptor near the spike could enhance binding, we fused EC1 to the N terminus of the M segment through a flexible linker (Fig. 1a). To increase particle production, we introduced an EABR tag²¹ and the V535K and S1096L mutations, as reported previously^5,22, and transfected the construct into FreeStyle293 cells. We analysed the secreted particles using cryo-EM and obtained a reconstruction of the spike at an overall resolution of 3.9 Å (Extended Data Fig. 5). The map revealed clear densities for Gn and Gc and an additional density at the centre of the spike (Fig. 1h). This central density was compatible in size and position with the PCDH1 EC1 domain observed in the soluble spike complex. However, the quality of this density did not allow unambiguous building of the EC1 domain, preventing a detailed comparison of the VLP–receptor interface with that obtained using the soluble spike (Extended Data Fig. 5c). This work nevertheless confirmed that PCDH1 EC1 occupies the central region of the Gn–Gc tetramer in viral particles.

To begin to investigate the roles of the CD and AB loops of Gn^H and the DE loop of Gn^B in spike–PCDH1 recognition, we produced rVSV-ANDV and rVSVs bearing the glycoproteins of the PCDH1-independent^1,19 OWH HTNV (rVSV-HTNV), as well as rVSV-ANDV variants in which we replaced the Gn^H CD loop or all three loops in contact with the receptor (Methods and Extended Data Fig. 4a) with those from HTNV. After normalizing the amount of incorporated Gn–Gc in the viral particles (Extended Data Fig. 4b), we measured their interaction with immobilized PCDH1 EC1 on ELISA plates. Consistent with the known receptor specificity of hantaviruses and our structural data, PCDH1 EC1 could capture rVSV-ANDV but not rVSV-HTNV or any of the chimeric variants (Fig. 2a).

Next, we used a deep-mutational scanning (DMS) approach to more comprehensively survey the sequence landscape of the spike–PCDH1 interaction. We generated two independent Gn libraries, each introducing all possible amino acid substitutions at 427 of 448 positions in the ectodomain (amino acids 24–471; Methods) and used these to create libraries of alphavirus-based vectors pseudotyped with ANDV Gn–Gc (Sem-Hantavirus (SHV)). We captured SHVs bound to PCDH1 EC1 immobilized on ELISA plates and sequenced the PCDH1-bound SHV fractions to identify variants enriched or depleted relative to the input population. We observed strong selective signals for the Gn^H CD and AB loops and nearby regions in both populations, consistent with roles for these sequences in virus–receptor recognition (Fig. 2b,c and Extended Data Fig. 6). Relatively few variants were recovered from the Gn^B DE loop, consistent with its higher level of conservation (Extended Data Fig. 4a) and suggesting an important structural role for this sequence in the assembly and/or stability of the spike. Notably, none of the Gn ectodomain substitutions observed in the recent Hondius outbreak isolate of ANDV (LS-CH2016/2016-12-23/Switzerland) mapped to the PCDH1-binding pocket or were selected in the DMS experiments, indicating that this virus probably resembles previous isolates of the virus in its ability to recognize and use PCDH1 (Extended Data Figs. 4 and 6).

PCDH1 is a specific receptor for NWHs, and we wondered whether the structure of the complex could reveal insights into receptor use. We analysed the sequence conservation of the PCDH1-binding pocket across hantaviruses and compared the conformation of the loops using the available crystal structures of HTNV and SNV. We identified two amino acid substitutions: one in the Gn^H AB loop of SEOV, DOBV and HTNV (L36F) and one in the Gn^B DE loop of HTNV (G418L) that are predicted to cause steric clashes with PCDH1. By contrast, both the sequence and structure of the Gn^H CD loop vary substantially across hantaviruses—it is an extended loop in NWHs but a compact helix in OWHs (Fig. 3a and Extended Data Fig. 4). These findings suggest that it is not only the residues directly contacting PCDH1 that are important to the spike–receptor interaction but also nearby residues that influence the overall conformation of the loops, as is also indicated by the DMS data (Fig. 2b and Extended Data Fig. 6).

To assess the functional impact of these sequence variations, we compared PCDH1-binding rVSVs displaying the glycoproteins of ANDV (rVSV-ANDV^WT), SNV (rVSV-SNV^WT) and HTNV (rVSV-HTNV^WT), along with variants in which we swapped the PCDH1-binding loops and nearby residues (Fig. 3b,c and Extended Data Fig. 4a). After normalizing the amount of incorporated glycoprotein in the rVSVs (Extended Data Fig. 4b), we measured PCDH1 binding by ELISA, as described above (Fig. 3b). As we showed previously^1,18, rVSV-ANDV^WT bound strongly to PCDH1, rVSV-SNV^WT bound less strongly and rVSV-HTNV^WT did not bind. The phenotypic effects of exchanging the receptor-binding loops between ANDV and SNV were largely consistent with the structural and mutagenesis data. Specifically, ANDV Gn bearing the SNV Gn^H CD loop bound worse than rVSV-ANDV^WT, whereas SNV Gn bearing Gn^H CD or all three receptor-binding loops from ANDV Gn bound better than rVSV-SNV^WT. Notably, a chimeric rVSV-HTNV bearing all three receptor-binding loops from ANDV Gn, but not the Gn^H CD loop alone, could now bind to PCDH1, albeit relatively weakly. This result is consistent with the sequence differences between ANDV and HTNV in the Gn^H AB and Gn^B DE loops and corroborates the proposed roles for these loops in ANDV spike–PCDH1 recognition.

The preceding binding experiments, in which sEC1–2 coated onto ELISA plates was used to capture rVSVs bearing hantavirus spikes, reflect the avid interaction between multiple immobilized PCDH1 molecules and the multiple spikes on each virion, which probably masks differences in the intrinsic affinity of each spike for PCDH1. Accordingly, we next assessed the ability of soluble, monomeric sEC1–2 to inhibit rVSV infection of PCDH1-bearing human umbilical vein endothelial cells (HUVECs) in receptor-competition experiments, as described previously¹⁸ (Fig. 3c). We established the specificity of this assay with a binding-incompetent double-mutant sEC1–2^MUT as a negative control (Fig. 3c). As expected^1,18, sEC1–2 inhibited infection by rVSVs bearing WT hantavirus spikes in a manner that was positively related to their capacity to recognize PCDH1: HTNV^WT << SNV^WT < ANDV^WT. The sensitivities to PCDH1 inhibition of the ANDV-SNV chimeras (SNV^WT ≈ ANDV^SNV-All < ANDV^SNV-CD < ANDV^WT), the SNV-ANDV chimeras (SNV^WT < SNV^ANDV-CD < SNV^ANDV-All < ANDV^WT) and the HTNV-ANDV chimeras (HTNV^WT ≈ HTNV^ANDV-CD < HTNV^ANDV-All < ANDV^WT) were also concordant with our binding results. Taken together, these findings broadly support the conclusion that the sequences shaping the central cavity are the major determinants of hantavirus spike–receptor recognition and of the distinct receptor tropisms of NWHs and OWHs.

The PCDH1-binding cavity is poorly accessible to conventional antibodies. Its dimensions—approximately 2.5 × 2.5 nm at its upper opening, narrowing to around 2.2 × 2.2 nm at the bottom, and around 2.1 nm in depth—are incompatible with those of a Fab fragment (5 × 4 × 7 nm). This raises the question of how some neutralizing antibodies block receptor engagement. To examine their mechanisms of action, we analysed four well-studied human neutralizing antibodies: ANDV-5 (ref. ²³) and MIB22 (refs. ^23,24), which were reported to interfere with the binding of soluble PCDH1 to spike-expressing cells; ANDV-34 (ref. ²³), which does not block this interaction; and ADI-65534 (ref. ¹⁰), which was reported to block the binding of rVSVs to PCDH1 EC1–2 immobilized on ELISA plates. Structures of ANDV-5, ANDV-34 and ADI-65534 are available, but no structure had been reported for MIB22. We therefore determined the crystal structure of an antibody fragment of MIB22 in a complex with ANDV Gn^H at a resolution of 3.1 Å (Extended Data Table 2) and identified an epitope of about 900 Ų that includes the three previously identified escape mutations²³. Notably, MIB22 contacts several residues in the CD loop and recognizes a conformation of the loop different from that adopted after PCDH1 binding (Extended Data Fig. 7a). To assess whether antibody binding occludes the PCDH1-binding pocket, we modelled the variable domains of the four neutralizing antibodies onto the spike–PCDH1 complex (Fig. 4a). In parallel, we produced the four neutralizing antibodies as both full-length antibodies (IgGs) and single-chain variable fragments (scFvs) lacking the antibody constant regions and compared their ability to block rVSV-ANDV–PCDH1 binding by ELISA (Fig. 4b). We also compared the neutralizing activity of the four full-length antibodies against rVSV-ANDV in HUVECs, which express higher levels of PCDH1 than many immortalized cell lines of non-endothelial origin¹ (Fig. 4c).

We observed that ADI-65534 and ANDV-34 do not occlude access to the central pocket, suggesting that they compete with PCDH1 through a steric hindrance mechanism mediated by the antibody constant regions (Fig. 4a). Consistent with this model, ELISA-based competition binding experiments showed that scFvs did not block the interaction of rVSVs with the receptor, whereas the IgGs did (Fig. 4b). Consistent with these observations, ANDV-34 has been reported not to block binding to the soluble receptor in flow-cytometry-based assays²⁵. By contrast, the models with the Fabs of ANDV-5 and MIB22 showed that these antibodies bind much closer to the centre of the spike, directly occluding the PCDH1-binding site. ANDV-5 may additionally make limited clashes with the neighbouring protomer, potentially perturbing the conformation of the tetrameric spike. Concordant with these findings and previous observations²⁵, ANDV-5 and MIB22 interfere with rVSV-ANDV–PCDH1 binding as both full-length antibodies and scFvs (Fig. 4b). In our experiments, the scFv of ANDV-5 blocked receptor binding tenfold more efficiently than its MIB22 counterpart (half-maximum inhibitory concentration (IC₅₀) values of 23 and 240 nM, respectively), which may be explained by ANDV-5’s higher intrinsic affinity for the spike²³. Overall, these results revealed that neutralizing antibodies elicited during viral infection can block receptor binding either directly, by occluding the receptor-binding site, or indirectly, by sterically preventing PCDH1 from accessing the spike.

Here we determined the structure of the ANDV spike bound to human PCDH1. We found that the receptor inserts one membrane-distal EC1 domain into a deep cavity formed by four Gn molecules in the fully assembled spike tetramer.

PCDH1 is a calcium-dependent cell–cell adhesion protein. Biochemical studies with soluble ectodomain fragments showed that PCDH1 can exist in a monomer–dimer equilibrium in solution^18,26; however, the oligomeric distribution of full-length PCDH1 at the plasma membrane is not known. Structural studies showed that PCDH1 can form antiparallel homodimers through an EC1–EC4 interface²⁶—a configuration that is incompatible with the receptor-binding mode described here because EC4 would not fit into the pocket. Previous studies with recombinant PCDH1 mutants engineered to remain monomeric did not reduce their interaction with rVSV-ANDV, whereas stabilization of the dimer decreased binding¹⁸. Overall, these data suggest that the virus engages monomeric PCDH1 at the cell surface, although dynamic PCDH1 dimers may also contribute if they transiently expose the EC1 domain.

The binding mode reported here explains why PCDH1 recognizes spikes on viral particles but not soluble glycoproteins and shows that receptor binding depends on both glycoprotein sequence and spike organization. The concept that receptors can recognize quaternary surfaces formed only in assembled viral glycoprotein complexes is not unique to hantaviruses, as the receptor-binding pocket for multiple alphaviruses is accessible only in the complete spike. For example, western equine encephalitis virus (WEEV) can accommodate multiple receptors, including protocadherin-10 (PCDH10)^27,28,29,30, the very-low-density lipoprotein receptor (VLDLR)^29,30 and MXRA8 (refs. ^27,31), using the same pocket formed by adjacent E2–E1 heterodimers. It is interesting to compare the interactions between ANDV and PCDH1 with those between WEEV and PCDH10 because both receptors are non-clustered δ-cadherins³² and use their membrane-distal EC1 domain for viral engagement. However, the two viruses use distinct binding modes and protocadherin-binding surfaces. PCDH10 EC1 engages the alphavirus spike laterally through its A, G, F and C β-strands, whereas PCDH1 EC1 inserts axially into the central cavity of the hantavirus spike using its BC and FG loops. By analogy with WEEV, we propose that the pockets analogous to the PCDH1-binding pocket of ANDV and other NWHs may serve as binding sites for the distinct receptors used by OWHs and other non-PCDH1-using hantaviruses. Consistent with this idea, we found that variations in the receptor-binding loops of HTNV Gn, which normally cannot recognize PCDH1, enable a rVSV displaying the HTNV spike to bind PCDH1. Structural, sequence and mutagenesis studies further showed that the Gn^H CD loop is flexible, variable and tolerant to multiple mutations, suggesting that it could adapt to recognize alternative receptors. DMS data indicated that substitutions in the Gn^H CD and AB loops can alter affinity for PCDH1. Furthermore, the fact that the pocket is formed by four subunits may also facilitate viral adaptation to different receptors through a limited number of amino acid substitutions.

Another interesting parallel with alphaviruses concerns fusion activation. In alphaviruses, the quaternary interactions that form the receptor-binding site must be disrupted to allow membrane fusion, and MXRA8 binding has been proposed to modulate the pH threshold for fusion activation³³. By contrast, the PCDH1-binding site does not involve the hantavirus fusion protein Gc, suggesting that receptor binding and membrane fusion by hantaviruses might be independent events. However, it has been reported that Gn adopts a different tetrameric arrangement at acidic pH⁷, raising the possibility that PCDH1 engagement may also influence fusion activation.

The neutralizing antibody response is essential for controlling hantavirus infections, as neutralizing titres correlate with HCPS severity³⁴ and administration of human ANDV immune plasma reduced the case fatality rate in a small trial³⁵. Antibodies may neutralize enveloped viruses by several non-mutually exclusive mechanisms, including by blocking viral attachment, membrane fusion or egress, inducing virion aggregation and, in vivo, by recruiting Fc-dependent effector functions^36,37. Several neutralizing monoclonal antibodies targeting the ANDV spike have been described^23,24,38. Some block viral membrane fusion by preventing dissociation of the pre-fusion heterodimer or the formation of the post-fusion trimer. Others interfere with PCDH1 engagement, but the lack of structural information has prevented a mechanistic understanding of how they do so. The structural data presented here show that the PCDH1-binding pocket is too narrow to readily accommodate a conventional antibody. We therefore analysed a panel of four human neutralizing antibodies to understand how they block receptor binding. We observed that ADI-65534 and ANDV-34 do not directly compete with PCDH1 but instead inhibit receptor binding through a steric interference mechanism mediated by the antibody constant region. This model of neutralization, in which antibody binding to accessible spike surfaces can prevent virus attachment without directly targeting the receptor-binding sites, has been reported for other viruses^37,39,40,41, and suggests that strategies to increase antibody occupancy at the virion surface, such as combining non-competing antibodies targeting accessible epitopes, or focusing the response on exposed loops in the most apical regions of the tetramer, may enhance neutralizing activity against hantaviruses.

In contrast to ADI-65534 and ANDV-34, ANDV-5 and MIB22 directly occlude the PCDH1-binding pocket by targeting epitopes adjacent to, but not within, the central cavity. This recessed receptor-binding site is reminiscent of the canyon-like clefts described in picornaviruses, which inspired Rossmann’s canyon hypothesis. In its original form, this hypothesis proposed that receptor-binding residues located deep within a narrow cleft could remain accessible to the receptor but not to conventional antibodies, thereby reducing the immune pressure on the most conserved residues^42,43. However, the subsequent description⁴³ of HRV14-specific neutralizing antibodies of which the Fabs can penetrate the receptor-binding cleft showed that virus–host arms races could indeed produce antiviral solutions to this strategy. The preceding observations with HRV14 raise an interesting alternative possibility also noted by these authors⁴³: that the recessed receptor-binding site affords the virus a second line of defence by allowing it to escape neutralization through amino acid substitutions that disrupt a more extensive and surface-exposed virus–antibody interface while sparing a smaller and more buried virus–receptor interface. Indeed, ANDV escape mutations to MIB22 are at residues distinct from, and more peripheral to, those that are critical for ANDV–PCDH1 recognition. Our work suggests that access to conserved sequences at the base of the receptor-binding cavity in the hantavirus spike is likely to require antibodies with extended CDR loops, single-domain antibodies or biologics designed to fit into the cavity. The identification of such molecules could yield exceptionally potent and broad antiviral inhibitors. Finally, because PCDH1 binds to a pocket that is present only in the fully assembled spike and hantavirus glycoproteins are structurally dynamic⁴⁴, it may also be possible to identify antibodies that recognize epitopes partially buried in the intact spike and thereby prevent formation of the receptor-binding site.

Protein purification

Soluble spike proteins

To produce the soluble spikes of ANDV, we inserted synthetic genes for Gn and Gc of ANDV (strain Chile-9717869, UniProt: Q9E006) codon-optimized for expression in Drosophila cells into a bicistronic plasmid to produce the plasmid p-sSpike. We designed the hetero-octamerization domain starting from HALC4_136, a synthetic homo-tetramerization domain formed by three antiparallel helices, which has the appropriate dimensions (PDB: 8D09)²⁰. In the first cistron of the plasmid, we cloned the ectodomain of Gn (amino acids 21–483) followed by the first two helices of HALC4_136, which we refer to as HALC4_136^N. The second cistron contains the complete ectodomain of Gc (amino acids 652–1107) followed by the third α-helix of HALC4_136, named HALC4_136^C. To avoid aggregation, we introduced four mutations in Gc: F1089A, F1093A, L1100A and L1104A. To further stabilize the hetero-oligomerization domain, we designed a disulfide bond joining the C terminus of HALC4_136^N with the N terminus of HALC4_136^C. For purification purposes, we introduced a double strep-tag at the C terminus of HALC4_136^C. For the structural studies, we cloned the extracellular cadherin domains EC1–4 of human PCDH1 (amino acids 58–503, UniProt: Q08174) into a pMT/BiP plasmid in frame with a His and a double-strep tag to produce the plasmid p-sEC1–4. We used these plasmids to obtain stable transfectants of Drosophila S2 cells (Thermo Fisher Scientific, R69007, authenticated by Thermo Fisher Scientific by karyotyping and isozyme profiling), together with the pCoPuro plasmid (ratio 1:20) for puromycin selection⁴⁶. We generated stable cell lines with the p-sSpike (soluble spike) plasmid alone and cell lines co-transfecting p-sSpike and p-sEC1–4 (EC1–4). Stable cell lines were selected and maintained in LM-SFM4 medium (Cytiva) containing 8 μg ml⁻¹ puromycin. Cultures of 1–3 l were grown in spinner flasks in medium supplemented with 1% penicillin–streptomycin antibiotics to approximately 10⁷ cells per ml, and protein expression was induced with 4 μM CdCl₂. After 5 days, the S2 medium supernatant was concentrated to 40 ml using a 10 kDa cut-off VivaFlow concentrator (Sartorius) and supplemented with Biolock (IBA) and 0.1 M Tris-HCl (pH 8.0), centrifuged for 30 min at 20,000g, and purified by Strep-Tactin affinity chromatography using a StrepTrap HP column (Cytiva). For the co-expression of soluble ANDV spikes and PCDH1 EC1–4, Strep-Tactin purification was followed by His affinity chromatography using the TALON column (Cytiva). Proteins were then purified by size-exclusion chromatography (SEC) on a 10/300 Superose6 Increase column (Cytiva) in 10 mM Tris pH 8.0, 100 mM NaCl. The protein purity and the presence of (Gn–Gc)₄ spikes and PCDH1 in different SEC fractions after the co-expression were assessed by SDS–PAGE under reducing and non-reducing conditions.

ANDV-VLPs

To obtain the complex of PCDH1–EC1 with ANDV-VLPs, the nucleic acid sequence of the first extracellular domain (amino acids 58–165) of human protocadherin 1 plus a C-terminal 5 × GGGS linker were codon-optimized, synthesized by GenScript, and inserted into the pCAG-ANDV-DM-VLP plasmid⁵ to make the pCAG-PCDH1-EC1-GS5-ANDV-DM-VLP plasmid. The insertion was between the ANDV native signal peptide sequence (1–18), predicted by the SignalP 5.0 server⁴⁷, and the N-terminal residue of Gn. For protein expression, the pCAG-PCDH1-EC1-GS5-ANDV-DM-VLP plasmid was mixed with 25 kDa polyethyleneimine (PEI; Polysciences) and transfected into FreeStyle 293-F cells (Thermo Fisher Scientific) at a density of 10⁶ cells per ml. Transfected cells were incubated in a 37 °C shaking incubator with 8% CO₂ and 80% humidity for 72 h. After transfection, the supernatants were collected by centrifugation, filtered with 0.22 μm Rapid-Flow filters (Nalgene), and concentrated to about 40 ml using a Centricon Plus-70 filter unit (MilliporeSigma). The concentrated supernatant was subsequently purified through discontinuous sucrose density gradient (20%, 50%, 70% sucrose in PBS) ultracentrifugation at 4°C, 135,000g for 2 h. Fractions corresponding to VLPs were collected, pooled and further concentrated to around 60 μl in PBS (pH 7.4) for cryo-EM grid preparation.

PCDH1 fragments

To obtain soluble His-tagged versions of human PCDH1 domains EC1–4 (sEC1–4), EC1–2 (sEC1–2) and EC1 (sEC1) for ELISA and BLI analyses, codon-optimized human PCDH1 fragments (amino acids 58–503 for sEC1–4, 58–284 for sEC1-2, and 58–160 for sEC1) were cloned into a pCAGGS (sEC1–4 and sEC1) or pcDNA3.1 (sEC1–2) plasmid in frame with a histidine tag. The sEC1–4 and sEC1 plasmids were used to transfect Expi293 (Thermo Fisher Scientific, A14527, not authenticated) at a density of 3 × 10⁶ cells per ml using FectoPRO DNA transfection reagent (Polyplus). After 5 days of incubation at 37 °C, the supernatant was collected and centrifuged for 30 min at 3,100g. Proteins were purified by TALON affinity chromatography followed by SEC on the HiLoad 16/600 Superdex 75 column (Cytiva) in 10 mM Tris pH 8.0 and 100 mM NaCl. The sEC1–2 plasmid, which also included a FLAG tag, was used to transfect FreeStyle 293-FS cells (Thermo Fisher Scientific) at a density of 1 × 10⁶ cells per ml using linear PEI (Polysciences). After incubation for 6 days at 37 °C, the supernatant was collected and clarified by centrifugation at 4,000 rpm for 20 min. Proteins were purified by nickel-chelation chromatography with Nickel-NTA resin (Qiagen). The protein was eluted in PBS containing 250 mM imidazole, and the buffer was exchanged with PBS.

Antibodies

To obtain full-length IgG versions of the monoclonal antibodies used in this study, synthetic variable regions of the light and heavy chains for each antibody were cloned into pCAGGS plasmids upstream of the corresponding human IgG1 constant regions. All antibodies were expressed in Expi293 cells as described above. Antibodies were purified by protein G affinity chromatography, eluted with 0.1 M glycine pH 2.7, neutralized with 1 M Tris-HCl pH 8.0 and further purified by SEC on the HiLoad 16/600 Superdex 200 column (Cytiva) in PBS. To obtain scFv versions of the monoclonal antibodies used in this study, we cloned the variable regions of the heavy and light chains joined by a linker (GTGGSGGGGSGGGGSGGGAS) into a pMT/BiP plasmid in frame with an enterokinase cleavage site and a double-Strep tag. These plasmids were used to generate stable transfectants of Drosophila S2 cells, as described above. Antibody fragments were purified by Strep-Tactin affinity chromatography using the StrepTrap HP column (Cytiva), followed by SEC on a HiLoad 16/600 Superdex 75 column (Cytiva) in PBS.

Cryo-EM grid preparation and data collection

Soluble spike proteins

To perform the cryo-EM studies, we diluted the purified soluble spikes alone or in a complex with PCDH1 EC1–4 to a concentration of 0.34 mg ml⁻¹ in 10 mM Tris pH 8.0, 100 mM NaCl and deposited 3 μl on UltrAufoil R1.2/1.3 300 mesh copper grids that had been glow-discharged twice for 25 s in the PELCO easyglow cleaning system (15 mA, 0.3 mBar). After 5 s of wait time, excess protein was blotted away for 3 s using Whatman No.1 filter paper with a blot force of −1 at 15 °C in 100% humidity before being plunge frozen in liquid ethane using Vitrobot Mark IV (Thermo Fisher Scientific). Frozen grids were imaged on the Titan Krios POLARIS (Synchrotron Soleil) system equipped with a FALCON 4i detector (Thermo Fisher Scientific), operating in super-resolution mode at a magnification of ×165,000, corresponding to a calibrated pixel size of 0.7455 Å per pixel. A full description of the cryo-EM data-collection parameters is provided in Supplementary Table 1.

ANDV-VLPs

For the VLPs, the R1.2/R1.3 300 mesh ultrAuFoil grids were glow-discharged with a setting of 20 mA, 1 min. Then, 3 μl concentrated PCDH1-EC1–GS5–ANDV-DM-VLP sample was added onto glow-discharged grids, blotted with a blot force of 4 and a blot time of 20 s and plunge-frozen into liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). The prepared grids were screened for optimal ice thickness on a 200 kV Glacios (Thermo Fisher Scientific) at The University of Texas at Austin. Screened grids were transferred to the cryo-EM facilities at UT Southwestern Medical Center (UTSW) for high-resolution data collection. Data collection was performed on the Titan Krios G4 (Thermo Fisher Scientific) system equipped with an X-FEG, K3 Summit camera (Gatan) and BioContinuum Imaging Filter (Gatan). A total of 18,507 videos was acquired at a pixel size of 0.413 Å, with a total dose of 56 e⁻ Å⁻² and a 15 eV energy filter slit width.

EM data processing and model building

Soluble spikes

Motion correction, CTF estimation, particle picking, 2D classification, 3D classification and non-uniform 3D refinement were performed using cryoSPARC (v.4.6.2)⁴⁸. For model building, the map was sharpened using DeepEMhancer⁴⁹. The initial model was generated using ChimeraX⁵⁰ by fitting the high-resolution ANDV tetramer structure (PDB: 9P3I)⁵ and the crystallographic structure of PCDH1 EC1 (PDB: 6BX7)²⁶ in the electron density map. The model was further built manually in Coot⁵¹ and refined iteratively using phenix.refine⁵² using a refinement strategy that includes minimization_global, local_grid_search and adp, with rotamer and Ramachandran restraints. The full cryo-EM data processing workflow is shown in Supplementary Fig. 1 and the model refinement statistics are provided in Supplementary Table 1.

ANDV-VLPs

For the VLPs, the data processing follows a similar pipeline reported for ANDV glycoprotein single tetramer⁵. In brief, the dataset was imported into CryoSPARC (v.4.6.2)⁴⁸ for patch-based motion correction, contrast transfer function (CTF) estimation and exposure curation. A model manually trained on 200 motion-corrected, dose-weighted micrographs was then used to predict particle coordinates within crYOLO (v.1.9.9)⁵³. Picked particles were then imported into CryoSPARC for particle extraction and 2D classification. The ANDV glycoprotein tetramer EM map was used as the reference for the homogeneous refinement to get the initial model. Then, recentring particles and particle re-extraction were performed, followed by rounds of homogeneous refinement with C₄ symmetry, 3D classification, reference-based motion correction and non-uniform refinement. In the final round of 3D classification, to extract potentially different binding poses of PCDH1–EC1 to the ANDV tetramer, the particles were first symmetry expanded, and a 3D classification was performed on the expanded particle datasets with a sphere mask created in ChimeraX centred on the PCDH1 EC1 density. The classified particles were then locally refined, and the refined EM maps were sharpened by EMReady (v.2.0)⁵⁴ to discriminate different binding poses.

ELISA

ELISA plates (Thermo Fisher Scientific, 439454) were coated with 100 µl of recombinant His tagged PCDH1 EC1–4 (WT or double mutant) diluted to 10 µg ml⁻¹ in PBS (pH 7.4). After overnight incubation at 4 °C, plates were washed three times with PBS containing 0.05% Tween-20 (PBS-T) and blocked with 250 µl per well of blocking buffer (PBS supplemented with 5% non-fat dry milk and 0.05% Tween-20). For the binding experiment with ANDV (Gn–Gc)₄ spikes (Fig. 1b), serial dilutions of the spike were prepared in blocking buffer and 100 µl per well were added to the wells and incubated for 4 h at room temperature with agitation at 300 rpm. The plates were then washed as described above, and bound spike protein was detected by adding 100 µl per well of anti-Strep-tag monoclonal antibody (C23.21, Thermo Fisher Scientific, 910STRL) and incubating for 90 min at room temperature with agitation at 300 rpm. After washing, 100 µl per well of anti-mouse HRP-conjugated antibody (Jackson ImmunoResearch, 115-035-146, 1:10,000) was added and incubated for 60 min at room temperature. The plates were washed five times, and binding was detected using TMB substrate (ELISA 1-Step Ultra TMB, Thermo Fisher Scientific, 34028). The reaction was stopped with 50 µl per well of 2 M H₂SO₄, and the absorbance at 450 nm was measured using the Tecan Infinite M100 multimode microplate reader. Reported values correspond to the mean and s.d. from independent experiments, each performed in triplicate.

For the competition experiments with soluble sEC1 WT or the double mutant (Fig. 1c) or with human IgGs or scFvs (Fig. 4b), purified protein or antibodies were either diluted to final concentrations of 10, 2, 0.4 and 0.08 µM in blocking buffer supplemented with 500 nM ANDV tetrameric spike protein or serially diluted (starting at 1.2 µM) and supplemented with 0.36 µl of FSL-biotin labelled rVSV-ANDV per dilution. After pre-incubation for 1 h at room temperature, the spike- or virus-competitor mixtures were added to the wells of WT PCDH1 EC1–4- or EC1–2-coated plates and the experiment proceeded as described in the previous paragraph.

For rVSV chimera ELISAs, 96-well ELISA plates (Corning) were coated with purified sEC1–2 (100 ng per well) overnight at 4 °C, washed briefly with PBS and blocked with 5% non-fat dry milk in PBS (2 h at room temperature). rVSV chimeras were normalized based on Gc expression by western blotting (see specifics below) and membrane-labelled with a short-chain phospholipid probe, functional-component spacer diacyl lipid conjugated to biotin (FSL-biotin (5 µg ml⁻¹); Sigma-Aldrich) for 1 h at 37 °C. The rVSVs were then serial diluted twofold and added to sEC1–2-coated wells (1 h at 37 °C). Bound rVSVs were detected by incubation with Pierce high-sensitivity Streptavidin–horseradish peroxidase (HRP) conjugate (1:10,000 dilution, Thermo Fisher Scientific). The ELISA signal was developed using 1-Step Ultra TMB-ELISA substrate solution (Thermo Fisher Scientific) and measured at an absorbance at 450 nm on the Cytation5 cell imaging multi-mode reader (Agilent BioTek Gen5 Microplate Reader and Imager software, v.3.2). Exact P values for Fig. 3b were as follows: ANDV versus SNV: 4.7 × 10⁻¹³, ANDV versus ANDV^SNV-all: 0.62, ANDV versus ANDV^SNV-CDLoop: 9.1 × 10⁻⁵ (left); SNV versus ANDV: 4.7 × 10⁻¹³, SNV versus SNV^ANDV-all: 4.7 × 10⁻¹³, SNV versus SNV^ANDV-CDLoop: 4.7 × 10⁻¹³ (middle); and HTNV versus ANDV: 1.1 × 10⁻¹², HTNV versus HTNV^ANDV-all: 8.9 × 10⁻⁷, HTNV versus HTNV^ANDV-CDLoop: 0.99 (right).

BLI measurements

BLI data shown in this Article were produced using the Octet R8 (Sartorius). The BLI data shown in Fig. 1d were generated as follows. First, Ni-NTA sensors (Sartorius) were activated in 10 mM NiCl₂ and then dipped into a solution of His-tagged PCDH1 EC1–4 at 2 μM in PBS pH 7.4 for 600 s. Next, sensors were quenched in solutions of PBS pH 7.4 with 500 nM His-tagged human TNF and 1 mg ml⁻¹ BSA for 600 s. The functionalized sensors were then dipped into solution containing soluble spikes diluted at different concentrations (900, 300, 112, 100 and 33 nM) in PBS pH 7.4 supplemented with 1 mg ml⁻¹ BSA (PBS-BSA) and finally moved to PBS-BSA for dissociation. Association and dissociation rates were recorded for 300 s. Non-specific interactions were monitored using a sensor coated with unrelated His-tagged protein (TNF) and signals were subtracted. The curves were processed using the Octet Analysis Studio software (v.13.0; Sartorius), and the K_D values were obtained after fitting to a 1:1 global model.

The BLI data shown in Extended Data Fig. 1b were generated as follows. First, Strep sensors (Sartorius, 18-0063) were dipped into a solution of strep-tagged (Gn–Gc)₄ ANDV soluble spike at 20 nM in PBS pH 7.4 for 2,400 s. Next, sensors were quenched in solutions of PBS pH 7.4 with 100 nM strep-tagged Vaccinia virus D8 protein and 1 mg ml⁻¹ BSA for 1,200 s. The functionalized biosensors were then dipped into solutions containing soluble PCDH1 EC1–4 at different concentrations (2,000, 1,000, 500 and 250 nM) in PBS pH 7.4 supplemented with 1 mg ml⁻¹ BSA and finally moved to PBS-BSA for dissociation. Association and dissociation rates were recorded for 3,600 s, and non-specific interactions were monitored using a sensor coated with unrelated strep-tagged vaccinia D8 protein.

Crystallization and structure determination of the complex Gn^H–MIB22

For crystallization of the complex Gn^H–scFv-MIB22, we produced Gn^H (amino acids 21–374) of ANDV and the scFv of MIB22 in D. melanogaster S2 cells, as described above. We removed the purification tags using enterokinase before running SEC to isolate the complex, which was concentrated in 10 mM Tris-HCl pH 8.0 and 150 mM NaCl using Vivaspin centricons. Crystallization screening trials were performed using the vapour-diffusion method on the Mosquito nanodispensing system (STPLabtech) according to established protocols⁵⁵. Crystals were grown in 0.2 M CaCl₂, 0.1 M Tris-HCl pH 8.5, 20% (w/v) PEG 4000 and were cryo-protected with 20% (v/v) glycerol, followed by rapid transfer into liquid nitrogen. Diffraction images were integrated with XDS (v.10 January 2022)⁵⁶, and crystallography calculations were carried out with programs from the CCP4 program suite (v.9)⁵⁷. The structure was solved by molecular replacement with Phaser (v.2.8.2) from the suite PHENIX (v.1.19.2-4158)⁵⁸ using PUUV Gn^H (PDB: 5FXU) and the scFv C10 of dengue virus (PDB: 4UT9) as search models. The final model was built by combining real space model building in Coot (v.0.9.8.92)⁵¹ with reciprocal space refinement with phenix.refine (v.1.14rc3). The final model was validated with Molprobity⁵⁹. The crystallography statistics and PDB codes are provided in Supplementary Table 1. The structural analysis software was installed using SBGRID⁶⁰.

Cloning of pSHV-ANDV Gn libraries

The development of the SHV system, a Semliki Forest virus (SFV) replicon expressing an adaptively mutated ANDV Gn–Gc (Chilean field strain R123, GenBank: NP_604472.1) in place of its structural proteins in the pCMV-SFV4 plasmid⁶¹ to study hantavirus entry, will be described elsewhere (L.K. et al., manuscript submitted). Transfection of a single (pSHV-ANDV) plasmid encoding the SHV genome into Huh7.5.1 cells produces self-replicating viruses, enabling the genotype–phenotype link required for deep mutagenesis scanning of Gn–Gc proteins. To generate SHV Gn mutants encompassing amino acids 24–471 of its ectodomain, the Gn sequence was divided into six sections based on the availability of unique restriction sites: lib1 (amino acids 24–92), lib2 (amino acids 101–166), lib3 (amino acids 170–230), lib4 (amino acids 235–312), lib5 (amino acids 316–386) and lib6 (amino acids 390–471). The lib1 sequence was flanked by NruI (5′ end) and AclI (3′ end); lib2 by AclI (5′ end) and PspXI (3′ end); lib3 by PspXI (5′ end) and XmnI (3′ end); lib4 by XmnI (5′ end) and BbvCI (3′ end); lib5 by BbvCI (5′ end) and XmaI (3′ end); and lib6 by XmaI (5′ end) and BlpI (3′ end). Site-saturation variant libraries of oligo pools carrying single amino acid changes, with one codon per amino acid for all 20 amino acids, including the parental sequence, were synthesized by Twist Biosciences and cloned into the pSHV vector by digesting the vector at the corresponding restriction sites and assembling the fragments using NEB HiFi DNA assembly. The parental Gn sequence was also included in each library. The selected codons for each amino acid were optimal for human cell expression: Ala, GCC; Cys, TGC; Asp, GAC; Glu, GAG; Phe, TTC; Gly, GGC; His, CAC; Ile, ATC; Lys, AAG; Leu, CTG; Met, ATG; Asn, AAC; Pro, CCC; Gln, CAG; Arg, CGC; Ser, AGC; Thr, ACC; Val, GTG; Trp, TGG; and Tyr, TAC. The sequences of the assembled plasmid libraries were verified by Illumina MiSeq amplicon sequencing (see below for details).

Generation and amplification of SHV libraries

Individual pSHV-ANDV sublibrary (lib1–6) plasmid pools were transfected using Lipofectamine-3000 according to the manufacturer’s instructions, at 10 μg per 100 mm plate of Huh7.5.1 cells, and SHVs were collected at 72 h after transfection to generate passage 0 stocks. To eliminate trans-complementation of dysfunctional Gn–Gc variants by parental SHV or other functional variants, the library was passaged twice at a multiplicity of infection (MOI) of 0.1 in Huh7.5.1 cells (a gift from C. M. Rice⁶² and F. V. Chisari⁶³). Viral supernatants collected at passage 2 were clarified by centrifugation at 4,500 rpm at 4 °C for 15 min, repeated twice. Viruses in the clarified supernatant were concentrated by ultracentrifugation at 25,000 rpm at 4 °C for 1:45 h. The virus pellet was resuspended in 1× serum-free DMEM and incubated overnight at 4 °C. The resuspended pellet was then overlaid on a 10% sucrose cushion and ultracentrifuged at 25,000 rpm at 4 °C for 1:45 h. Subsequently, viral pellets were resuspended in 1× NT Buffer (10 mM Tris-Cl, 135 mM NaCl, pH 7.5) and stored at −80 °C before further downstream analysis. The SHV titres were estimated by immunostaining naive Huh7.5.1 cells infected with limiting dilutions of SHVs for expression of SFV non-structural protein 1 (nsP1) at 12 h after infection, using rabbit antisera (1:300) provided by A. Merits at the University of Tartu⁶⁴, followed by an Alexa-Fluor-488-conjugated anti-rabbit secondary antibody (1:500).

Pull-down of PCDH1-bound SHV particles

To enrich PCDH1-binding SHV particles, a sEC1–2 capture ELISA was performed as described previously^1,18. First, the amounts of SHV particles across the sublibraries and parental SHVs were normalized by labelling the pelleted SHV particle membranes with FSL-biotin (functional spacer di-acyl lipid probe conjugated to biotin) (Sigma-Aldrich). Equal amounts of FSL-biotin (10 μg ml⁻¹) and pelleted SHVs were mixed and incubated at 37 °C for 1 h for labelling. Serial twofold dilutions of labelled SHVs were coated on a high-binding ELISA plate (Corning) at 37 °C for 1 h. After removing unbound virus and washing with 1× PBS three times for 5 min each, the wells were blocked with 5% non-fat dry milk (Bio-Rad) prepared in 1× PBS for 1 h at 37 °C. The FSL-labelled SHVs were detected using a high-sensitivity streptavidin-HRP conjugated antibody (BioLegend) at a final dilution of 1:3,000 after incubation at 37 °C for 1 h. The ELISA signal was developed using Ultra-TMB ELISA substrate (Thermo Fisher Scientific) and measured at 450 nm on the Cytation-5 multi-mode plate reader (BioTek, Agilent). SHV input amounts for the sEC1–2 capture ELISA for each sublibrary were calculated from half-maximum effective concentration values that were obtained by fitting the data to a sigmoidal four-parameter logistic model, with the logarithm of the virus amount on the x axis, using Prism GraphPad v.11.

For the sEC1–2 capture ELISA, high-affinity 96-well ELISA plates (Corning) were coated with sEC1-2 (100 ng per well) at 4 °C overnight. The next day, after a single wash with 1× PBS, the plates were blocked with 5% blotting-grade non-fat dry milk (Bio-Rad) in 1× PBS for 1 h at 37 °C. Normalized amounts of pelleted parental and sublibrary SHVs labelled with FSL-biotin, as described above, were applied to the sEC1–2-coated wells for 1 h at 37 °C. After three thorough washes with 1× PBS to remove unbound SHVs, the bound viruses were processed for RNA isolation and PCR with reverse transcription (RT–PCR) as described below. Based on infectious units, sufficient SHV particles were included to ensure 1,000× coverage in the input of each sublibrary for the capture ELISA, and 10–12 wells per sublibrary were pooled for RNA isolation. For each capture ELISA, a parallel set of wells was processed for streptavidin–HRP-based detection of bound SHV particles, as described above for the normalization ELISA.

Illumina sequencing of SHV genomes

PCDH1-bound SHVs from independent biological duplicates for each sublibrary were processed by adding 50 μl of Zymogen RNA lysis buffer per well, incubating at room temperature for 8 min, and then isolating RNA using the Quick RNA Viral Kit according to the manufacturer’s instructions (Zymo Research). Input SHVs for each sublibrary were also sequenced to establish the baseline. cDNA was synthesized using the SuperScript III First-Strand Synthesis System (Invitrogen) according to the manufacturer’s instructions, with a reverse primer (5′-ATTGCCTACCCTAACACTAGTTGCAG-3′) located downstream of the Gn–Gc sequence. Sufficient viral RNA was used to ensure 1,000× coverage of each sublibrary. cDNA synthesis was followed by PCR with ANDV Gn-specific primers tagged at their 5′-ends with Illumina MiSeq adapters for the forward (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-3′) and reverse (5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-3′) sequences for each sublibrary. The sequences of the Gn-specific part of these primers were as follows: lib1-F, 5′-GTGCTATACTCTGACTCTCGCGATG-3′ and lib1-R, 5′- CGGTCTTTGTCTGTGCCTCAAAC-3′; lib2-F, 5′-CCACCAACGCAGCTTCCAC-3′ and lib2-R 5′-TTCTCATATATCACCTGTATCCTGTATCCTCGAGGT-3′; lib3-F, 5′-CTTGCATGGCCTCTGTGTTCAC-3′ and lib3-R, 5′-TGTGCACCCTGTCTTGGTGA-3′; lib4-F, 5′-AGCTGAAGGTGATCAAAACATTTGAAGG-3′ and lib4-R 5′-ACCCCGGCGAAAGCTGTC-3′; lib5-F, 5′-GTTCCTTCAACCTCTTCTACAGATACGC-3′ and lib5-R, 5′-TGTAGGCCTCGCAAGAAGCA-3′; and lib6-F, 5′-TGTTCTGTACTCTGGCCGGC-3′ and lib6-R, 5′-AGGGACACAGTTCGACGG-3′. Cycling conditions for the PCR were initial denaturation at 98 °C for 1 min; followed by 30 cycles of denaturation at 98 °C for 15 s, annealing at 65 °C for 30 s, extension at 72 °C for 30 s; and then a final extension step at 72 °C for 5 min. After verification of amplified PCR products on a 1% agarose gel, the products were purified using Ampure XP beads and indexed by a second round of PCR, followed by another cleanup with Ampure XP beads. Completed libraries were assessed by gel electrophoresis using the automated Agilent TapeStation D1000 Kit (Agilent Technologies) and quantified using the Qubit dsDNA BR Assay Kit (Thermo Fisher Scientific). Libraries were normalized to 40 nM, pooled and diluted to 4 nM. The 4 nM pool was denatured and diluted to 12.5 pM, then spiked with 10% Phi X DNA to increase base diversity. Paired-end 2 × 250 sequencing was performed on the Illumina MiSeq system using the Illumina MiSeq Reagent v2 kit (Illumina). Primary data analysis, base calling and quality scoring were performed using the onboard MiSeq Control Software v.4.1.0.480 and RTA v.1.18.54.4. Samples were demultiplexed, adapter sequences were trimmed and FASTQ files were generated for downstream analysis.

Analysis of the SHV sequencing data

The FASTQ files were aligned to the parental reference sequence using the Burrows–Wheeler Alignment Tool (BWA, v.0.7.17). Sequencing reads were processed using a custom Python-based DMS pipeline. Paired-end reads were first filtered to retain only high-quality sequences that (1) matched expected primer regions within defined mismatches; (2) had limited ambiguous base calls (N); and (3) showed sufficient overlap between R1 and R2 reads. Trimmed reads were merged, and coding regions were translated to protein sequences using a codon dictionary described above. To identify valid single amino acid variants, translated protein sequences were compared to the parental sequence. Only sequences differing by one amino acid and free of stop codons were retained. For each variant, the mutation position (adjusted by an offset to match full-length protein numbering) and substituted residue were recorded. Mutation counts were aggregated per position and normalized to compute the relative frequency of each variant, generating a position-wise matrix of single amino acid substitutions. These results were saved in a CSV file and used for downstream analysis.

For each set (input and bound replicates), CSVs containing Mut_type, Mutation_count and Position_count were parsed to extract a numeric position from the leading digits of Mut_type (for example, 24C → 24) and define Variant = Mut_type. Variant-level enrichment in each bound replicate versus input was quantified using a smoothed within-position frequency with pseudocount pc = 0.5: freq = (mutation_count + pc)/(position_count + 2 pc), and enrichment was reported as log₂[FC]= log₂[freq_bound/freq_input]. For each variant, a two-sided Fisher’s exact test compared the variant versus all other reads at that position between bound and input using a 2 × 2 table (mut, nonmut=position_count−mutation_count), followed by Benjamini–Hochberg FDR correction per replicate. Reproducibility across bound replicates was assessed by Pearson/Spearman correlation of log₂[FC] values and by calling variants reproducibly enriched/depleted if they had the same sign of log₂[FC] in both replicates and FDR < 0.05 in both. Finally, an input-only filter was applied: only variants with input mutation_count ≥ 10, an arbitrary limit, were retained for the filtered master workbook (bound counts were not filtered). A custom Python code used to process the raw FASTQ files to variant counts per position is available at GitHub (https://github.com/Jwang30/DMS_Calling).

rVSVs

Replication-competent rVSVs expressing an eGFP reporter and bearing hantavirus Gn–Gc of ANDV (NP_604472.1), SNV (NP_941974.1) or HTNV (NP_941978.1), with previously described modifications²², or chimeric variants of the three listed, were generated using a plasmid-based rescue system in HEK293T cells (ATCC, authenticated by short-tandem-repeat profiling) and propagated on Vero (ATCC, authenticated by karyotyping) or HUH7.5.1 (J. Carette, not authenticated) cells as described previously^65,66. The constructs encoding the chimeric Gn–Gcs were cloned into the pVSV plasmid using standard molecular techniques. The chimera rVSVs contained residue swaps between ANDV and SNV, and ANDV and HTNV within the receptor-binding site, encompassing the three EC1-interacting loops (AB, CD and DE) or the CD-loop alone. Additional mutations were included to accommodate space for the loop swaps between the NWHs and OWHs. The comprehensive list of Gn amino acid changes in the ‘CD loop’ and ‘all’ chimeras is as follows (using ANDV numbering). CD Loop: ANDV residues 118–127 and adjacent residues 40 and 251. All: residues 35–38, 40, 118–127, 153, 156–157, 180, 195, 251, 257 and 413–419. A visual summary of all of the swaps within Gn is provided in Extended Data Fig. 4. After viral plaque purification, the sequences of all rVSV glycoproteins were amplified from viral genomic RNA by RT–PCR and analysed using Sanger sequencing.

Infection inhibition and neutralization assays

To evaluate the ability of PCDH1 to block infection of the rVSV chimeras, rVSVs bearing either WT or swapped PCDH1-interacting loops (all three loops: AB, CD, DE or CD alone) of ANDV/SNV and ANDV/HTNV were diluted to an MOI of 0.1 and incubated with threefold serial dilutions of soluble PCDH1 EC1–2 (WT or mutant containing the substitutions F83A and D85R) at room temperature for 1 h before being added to HUVEC (Thermo Fisher Scientific, authenticated by Invitrogen by cell morphology and cell-type-specific markers) monolayers in 96-well plates. rVSVs bearing ANDV Gn–Gc with HTNV-all residue swap did not readily infect HUVECs and therefore could not be evaluated. Exact P values for Fig. 3c (sEC1–2^WT) were as follows: ANDV versus SNV: 1.0 × 10⁻¹⁵, ANDV versus ANDV^SNV-all: 1.0 × 10⁻¹⁵, ANDV versus ANDV^SNV-CDLoop: 1.6 × 10⁻⁷ (left); SNV versus ANDV: 1.0 × 10⁻¹⁵, SNV versus SNV^ANDV-all: 6.1 × 10⁻⁵, SNV versus SNV^ANDV-CDLoop: 8.9 × 10⁻³ (middle); and HTNV versus ANDV: 1.0 × 10⁻¹⁵, HTNV versus HTNV^ANDV-all: 5.4 × 10⁻⁹, HTNV versus HTNV^ANDV-CDLoop: 0.99 (right). Exact P values for Fig. 3c, bottom row (sEC1-2^MUT) were as follows: ANDV versus SNV: 1.00, ANDV versus ANDV^SNV-all: 0.69, ANDV versus ANDV^SNV-CDLoop: 1.00 (left); SNV versus ANDV: 1.00, SNV versus SNV^ANDV-all: 0.68, SNV versus SNV^ANDV-CDLoop: 1.00 (middle); and HTNV versus ANDV: 0.02, HTNV versus HTNV^ANDV-all: 0.37, HTNV versus HTNV^ANDV-CDLoop: 0.71 (right). To compare the neutralizing activity of four full-length antibodies (ADI-65534, ANDV-34, ANDV-5, and MIB22), rVSV-ANDV (MOI 0.1) was incubated with threefold serial dilutions of each antibody, starting at 33 nM, for 1 h before being added to HUVECs cultured in 96-well plates. For all infections, cells were fixed 14–16 h after infection with 4% formaldehyde (Sigma-Aldrich) and counterstained with Hoechst nuclear stain (Invitrogen). The number of infected cells (GFP-positive cell counts) and viral infectivity (the percentage of GFP-positive cells) were quantified by automated enumeration of eGFP-expressing cells from images acquired using the Cytation 5 Cell Imaging Multi-Mode Reader with Agilent BioTek Gen5 Microplate Reader and Imager software (BioTek, v.3.2).

Immunoblotting

To normalize the amount of each virus used for rVSV–sEC1–2 ELISAs, the level of glycoprotein (Gc) incorporation was quantified by immunoblotting. rVSV particles bearing ANDV or SNV Gc were detected using an anti-hantavirus Gc rabbit pAb 16281 made from immunized rabbits with the peptide containing ANDV Gc amino acids 1067–1087 (Pacific Immunology). The rVSVs bearing ANDV or HTNV Gc were detected using a mouse anti-hantavirus Gc clone 2H4/F6 monoclonal antibody⁶⁷. Both antibodies were diluted in 1% NFDM and blots were stained for 1 h at room temperature, followed by washing with PBS with 0.01% Tween-20 (PBST), and incubation with an anti-rabbit IgG Alexa Fluor Plus 680 pAb (Thermo Fisher Scientific), an anti-mouse IgG Alexa Fluor 680 pAb (Thermo Fisher Scientific) or an anti-mouse IgG IRDye-800CW pAb (LI-COR) for 1 h at room temperature. After thoroughly washing with PBST, blots were imaged on the iBright FL1500 Imaging System and analysed using the iBright Analysis Software. The chimera rVSVs were normalized to their WT parent based on the amount of incorporated glycoprotein. Uncropped and unprocessed scans of the blots are provided in Supplementary Fig. 1.

Detection of soluble PCDH1 bound to Gn–Gc-expressing cells by flow cytometry

To evaluate the binding of sEC1 to ANDV glycoproteins displayed within a membrane, Vero cells seeded in six-well plates were infected with rVSVs expressing eGFP, bearing either ANDV or HTNV Gn–Gc with an MOI of 0.3 IUs per cell for 1 h. Infection was stopped by the addition of NH₄Cl (20 mM). At 14 h after infection, cells were collected with a cell scraper and blocked with chilled 10% FBS in PBS for 30 min at 4 °C. Cells were then incubated for 1 h at 4 °C with soluble PCDH1 EC1–2 (500 nM or 1000 nM) or buffer only (10% FBS in PBS) as a negative control. After multiple washes with PBS, cells were fixed with 4% formaldehyde (Sigma-Aldrich) for 5 min at 4 °C and washed once with PBS. To detect bound sEC1–2, cells were incubated with a mouse anti-DDK(FLAG)-DyLight-594 monoclonal antibody (Thermo Fisher Scientific) for 1 h at 4 °C, followed by multiple washes with PBS. Cells were then resuspended in 2% FBS in PBS, passed through a 41 µm nylon net filter (Millipore) and analysed using the LSR II flow cytometer and FlowJo v.10 software. The gating strategy is provided in Supplementary Fig. 2.

Figure preparation and visualization

Molecular representations were prepared using PyMol v.3.0.3 (Schrödinger) and ChimeraX (v.1.6.1)⁵⁰. The panels in Fig. 1a were generated using the web server IBS (v.2.0)⁶⁸ and the Extended Data Fig. 4a using the server ESPRIPT^69,70,71,72. Extended Data Fig. 6a was generated using Seaborn (v.0.13.2)⁷³. Graphs were generated using GraphPad Prism (v.11). Figures were assembled and edited using Adobe Illustrator (v.28.5).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

A custom Python code used to process the raw FASTQ files to variant counts per position is available at GitHub (https://github.com/Jwang30/DMS_Calling) and has been archived at Zenodo⁷⁴ (https://doi.org/10.5281/zenodo.22751343).

1. Jangra, R. K. et al. Protocadherin-1 is essential for cell entry by New World hantaviruses. Nature 563, 559–563 https://doi.org/10.1038/s41586-018-0702-1 (2018).

   Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

2. Martinez, V. P. et al. “Super-spreaders” and person-to-person transmission of Andes virus in Argentina. N. Engl. J. Med. 383, 2230–2241 https://doi.org/10.1056/NEJMoa2009040 (2020).

   Article  CAS  PubMed  Google Scholar 

3. Padula, P. J. et al. Hantavirus pulmonary syndrome outbreak in Argentina: molecular evidence for person-to-person transmission of Andes virus. Virology 241, 323–330 https://doi.org/10.1006/viro.1997.8976 (1998).

   Article  CAS  PubMed  Google Scholar 

4. Andes Virus Outbreak Working Group. Andes hantavirus outbreak on a cruise ship, 2026. N. Engl. J. Med. 394, 2477–2479 (2026).

5. Guo, L. et al. High-resolution in situ structures of hantavirus glycoprotein tetramers. Cell 189, 2731–2747 https://doi.org/10.1016/j.cell.2026.01.030 (2026).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

6. Li, S. et al. A molecular-level account of the antigenic hantaviral surface. Cell Rep. 15, 959–967 https://doi.org/10.1016/j.celrep.2016.03.082 (2016).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

7. Rissanen, I. et al. Structural transitions of the conserved and metastable hantaviral glycoprotein envelope. J. Virol. 91, 21 https://doi.org/10.1128/JVI.00378-17 (2017).

8. Guardado-Calvo, P. et al. Mechanistic insight into bunyavirus-induced membrane fusion from structure-function analyses of the hantavirus envelope glycoprotein Gc. PLoS Pathog. 12, e1005813 https://doi.org/10.1371/journal.ppat.1005813 (2016).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

9. Serris, A. et al. The hantavirus surface glycoprotein lattice and its fusion control mechanism. Cell 183, 442–456 https://doi.org/10.1016/j.cell.2020.08.023 (2020).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

10. Mittler, E. et al. Structural and mechanistic basis of neutralization by a pan-hantavirus protective antibody. Sci. Transl. Med. 15, eadg1855 https://doi.org/10.1126/scitranslmed.adg1855 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

11. Matthys, V. S., Gorbunova, E. E., Gavrilovskaya, I. N. & Mackow, E. R. Andes virus recognition of human and Syrian hamster β3 integrins is determined by an L33P substitution in the PSI domain. J. Virol. 84, 352–360 https://doi.org/10.1128/JVI.01013-09 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

12. Gavrilovskaya, I. N., Shepley, M., Shaw, R., Ginsberg, M. H. & Mackow, E. R. beta3 Integrins mediate the cellular entry of hantaviruses that cause respiratory failure. Proc. Natl Acad. Sci. USA 95, 7074–7079 https://doi.org/10.1073/pnas.95.12.7074 (1998).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

13. Gavrilovskaya, I. N., Brown, E. J., Ginsberg, M. H. & Mackow, E. R. Cellular entry of hantaviruses which cause hemorrhagic fever with renal syndrome is mediated by β3 integrins. J. Virol. 73, 3951–3959 https://doi.org/10.1128/JVI.73.5.3951-3959.1999 (1999).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

14. Raftery, M. J. et al. β2 integrin mediates hantavirus-induced release of neutrophil extracellular traps. J. Exp. Med. 211, 1485–1497 https://doi.org/10.1084/jem.20131092 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

15. Krautkramer, E. & Zeier, M. Hantavirus causing hemorrhagic fever with renal syndrome enters from the apical surface and requires decay-accelerating factor (DAF/CD55). J. Virol. 82, 4257–4264 https://doi.org/10.1128/JVI.02210-07 (2008).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

16. Buranda, T. et al. Equilibrium and kinetics of Sin Nombre hantavirus binding at DAF/CD55 functionalized bead surfaces. Viruses 6, 1091–1111 https://doi.org/10.3390/v6031091 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

17. Choi, Y. et al. A hantavirus causing hemorrhagic fever with renal syndrome requires gC1qR/p32 for efficient cell binding and infection. Virology 381, 178–183 https://doi.org/10.1016/j.virol.2008.08.035 (2008).

    Article  CAS  PubMed  Google Scholar 

18. Slough, M. M. et al. Two point mutations in protocadherin-1 disrupt hantavirus recognition and afford protection against lethal infection. Nat. Commun. 14, 4454 https://doi.org/10.1038/s41467-023-40126-y (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

19. Dieterle, M. E. et al. Genetic depletion studies inform receptor usage by virulent hantaviruses in human endothelial cells. eLife 10, e69708 https://doi.org/10.7554/eLife.69708 (2021).

20. Wicky, B. I. M. et al. Hallucinating symmetric protein assemblies. Science 378, 56–61 https://doi.org/10.1126/science.add1964 (2022).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

21. Hoffmann, M. A. G. et al. ESCRT recruitment to SARS-CoV-2 spike induces virus-like particles that improve mRNA vaccines. Cell 186, 2380–2391 https://doi.org/10.1016/j.cell.2023.04.024 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

22. Slough, M. M., Chandran, K. & Jangra, R. K. Two point mutations in Old World hantavirus glycoproteins afford the generation of highly infectious recombinant vesicular stomatitis virus vectors. mBio 10, e02372-18 https://doi.org/10.1128/mBio.02372-18 (2019).

23. Engdahl, T. B. et al. Antigenic mapping and functional characterization of human New World hantavirus neutralizing antibodies. eLife 12, e81743 https://doi.org/10.7554/eLife.81743 (2023).

24. Garrido, J. L. et al. Two recombinant human monoclonal antibodies that protect against lethal Andes hantavirus infection in vivo. Sci. Transl. Med. 10, eaat6420 https://doi.org/10.1126/scitranslmed.aat6420 (2018).

25. Engdahl, T. B. et al. Broad and potently neutralizing monoclonal antibodies isolated from human survivors of New World hantavirus infection. Cell Rep. 35, 109086 https://doi.org/10.1016/j.celrep.2021.109086 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

26. Modak, D. & Sotomayor, M. Identification of an adhesive interface for the non-clustered delta1 protocadherin-1 involved in respiratory diseases. Commun. Biol. 2, 354 https://doi.org/10.1038/s42003-019-0586-0 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

27. Raju, S. et al. Structural basis for plasticity in receptor engagement by an encephalitic alphavirus. Cell 188, 2943–2956 https://doi.org/10.1016/j.cell.2025.02.036 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

28. Liang, S. et al. Structural basis for engagement of western equine encephalitis virus with the PCDH10 receptor. Nat. Commun. 16, 6290 https://doi.org/10.1038/s41467-025-61659-4 (2025).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

29. Fan, X. et al. Molecular basis for shifted receptor recognition by an encephalitic arbovirus. Cell 188, 2957–2973 https://doi.org/10.1016/j.cell.2025.03.029 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

30. Ma, B. et al. Structural basis for the recognition of two different types of receptors by Western equine encephalitis virus. Cell Rep. 44, 115724 https://doi.org/10.1016/j.celrep.2025.115724 (2025).

    Article  CAS  PubMed  Google Scholar 

31. Zimmerman, O. et al. Vertebrate-class-specific binding modes of the alphavirus receptor MXRA8. Cell 186, 4818–4833 https://doi.org/10.1016/j.cell.2023.09.007 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

32. Harrison, O. J. et al. Family-wide structural and biophysical analysis of binding interactions among non-clustered delta-protocadherins. Cell Rep. 30, 2655–2671 https://doi.org/10.1016/j.celrep.2020.02.003 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

33. Basore, K. et al. Cryo-EM structure of Chikungunya virus in complex with the Mxra8 receptor. Cell 177, 1725–1737 https://doi.org/10.1016/j.cell.2019.04.006 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

34. Bharadwaj, M., Nofchissey, R., Goade, D., Koster, F. & Hjelle, B. Humoral immune responses in the hantavirus cardiopulmonary syndrome. J. Infect. Dis. 182, 43–48 https://doi.org/10.1086/315657 (2000).

    Article  CAS  PubMed  Google Scholar 

35. Vial, P. A. et al. A non-randomized multicentre trial of human immune plasma for treatment of hantavirus cardiopulmonary syndrome caused by Andes virus. Antivir. Ther. 20, 377–386 https://doi.org/10.3851/IMP2875 (2015).

    Article  CAS  PubMed  Google Scholar 

36. Burton, D. R. Antiviral neutralizing antibodies: from in vitro to in vivo activity. Nat. Rev. Immunol. 23, 720–734 https://doi.org/10.1038/s41577-023-00858-w (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

37. Parren, P. W. & Burton, D. R. The antiviral activity of antibodies in vitro and in vivo. Adv. Immunol. 77, 195–262 https://doi.org/10.1016/s0065-2776(01)77018-6 (2001).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

38. Mittler, E. et al. Human antibody recognizing a quaternary epitope in the Puumala virus glycoprotein provides broad protection against orthohantaviruses. Sci. Transl. Med. 14, eabl5399 https://doi.org/10.1126/scitranslmed.abl5399 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

39. Yang, X., Lipchina, I., Cocklin, S., Chaiken, I. & Sodroski, J. Antibody binding is a dominant determinant of the efficiency of human immunodeficiency virus type 1 neutralization. J. Virol. 80, 11404–11408 https://doi.org/10.1128/JVI.01102-06 (2006).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

40. Ren, X., Sodroski, J. & Yang, X. An unrelated monoclonal antibody neutralizes human immunodeficiency virus type 1 by binding to an artificial epitope engineered in a functionally neutral region of the viral envelope glycoproteins. J. Virol. 79, 5616–5624 https://doi.org/10.1128/JVI.79.9.5616-5624.2005 (2005).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

41. Laird, M. E. & Desrosiers, R. C. Infectivity and neutralization of simian immunodeficiency virus with FLAG epitope insertion in gp120 variable loops. J. Virol. 81, 10838–10848 https://doi.org/10.1128/JVI.00831-07 (2007).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

42. Rossmann, M. G. The canyon hypothesis. Hiding the host cell receptor attachment site on a viral surface from immune surveillance. J. Biol. Chem. 264, 14587–14590 (1989).

    Article  CAS  PubMed  Google Scholar 

43. Smith, T. J., Chase, E. S., Schmidt, T. J., Olson, N. H. & Baker, T. S. Neutralizing antibody to human rhinovirus 14 penetrates the receptor-binding canyon. Nature 383, 350–354 https://doi.org/10.1038/383350a0 (1996).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

44. Bignon, E. A., Albornoz, A., Guardado-Calvo, P., Rey, F. A. & Tischler, N. D. Molecular organization and dynamics of the fusion protein Gc at the hantavirus surface. elife 8, e46028 https://doi.org/10.7554/eLife.46028 (2019).

45. Stass, R. et al. Mechanistic basis for potent neutralization of Sin Nombre hantavirus by a human monoclonal antibody. Nat. Microbiol. 8, 1293–1303 https://doi.org/10.1038/s41564-023-01413-y (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

46. Meola, A. & Guardado-Calvo, P. Production and purification of hantavirus glycoproteins in Drosophila melanogaster S2 cells. Methods Mol. Biol. 2762, 3–16 https://doi.org/10.1007/978-1-0716-3666-4_1 (2024).

    Article  CAS  PubMed  Google Scholar 

47. Almagro Armenteros, J. J. et al. SignalP 5.0 improves signal peptide predictions using deep neural networks. Nat. Biotechnol. 37, 420–423 https://doi.org/10.1038/s41587-019-0036-z (2019).

    Article  CAS  PubMed  Google Scholar 

48. Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 https://doi.org/10.1038/nmeth.4169 (2017).

    Article  CAS  PubMed  Google Scholar 

49. Sanchez-Garcia, R. et al. DeepEMhancer: a deep learning solution for cryo-EM volume post-processing. Commun. Biol. 4, 874 https://doi.org/10.1038/s42003-021-02399-1 (2021).

    Article  PubMed  PubMed Central  Google Scholar 

50. Meng, E. C. et al. UCSF ChimeraX: tools for structure building and analysis. Protein Sci. 32, e4792 https://doi.org/10.1002/pro.4792 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

51. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D 66, 486–501 https://doi.org/10.1107/S0907444910007493 (2010).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

52. Afonine, P. V. et al. Towards automated crystallographic structure refinement with phenix.refine. Acta Crystallogr. D 68, 352–367 https://doi.org/10.1107/S0907444912001308 (2012).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

53. Wagner, T. et al. SPHIRE-crYOLO is a fast and accurate fully automated particle picker for cryo-EM. Commun. Biol. 2, 218 https://doi.org/10.1038/s42003-019-0437-z (2019).

    Article  PubMed  PubMed Central  Google Scholar 

54. He, J., Li, T. & Huang, S. Y. Improvement of cryo-EM maps by simultaneous local and non-local deep learning. Nat. Commun. 14, 3217 https://doi.org/10.1038/s41467-023-39031-1 (2023).

    Article  ADS  CAS  PubMed  Google Scholar 

55. Weber, P. et al. High-throughput crystallization pipeline at the crystallography core facility of the institut pasteur. Molecules 24, 4451 https://doi.org/10.3390/molecules24244451 (2019).

56. Kabsch, W. Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr. D 66, 133–144 https://doi.org/10.1107/S0907444909047374 (2010).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

57. Collaborative Computational Project, Number 4 The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D 50, 760–763 https://doi.org/10.1107/S0907444994003112 (1994).

    Article  ADS  Google Scholar 

58. Liebschner, D. et al. Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix. Acta Crystallogr. D 75, 861–877 https://doi.org/10.1107/S2059798319011471 (2019).

    Article  ADS  CAS  Google Scholar 

59. Williams, C. J. et al. MolProbity: more and better reference data for improved all-atom structure validation. Protein Sci. 27, 293–315 https://doi.org/10.1002/pro.3330 (2018).

    Article  CAS  PubMed  Google Scholar 

60. Morin, A. et al. Collaboration gets the most out of software. elife 2, e01456 https://doi.org/10.7554/eLife.01456 (2013).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

61. Ulper, L., Sarand, I., Rausalu, K. & Merits, A. Construction, properties, and potential application of infectious plasmids containing Semliki Forest virus full-length cDNA with an inserted intron. J. Virol. Methods 148, 265–270 https://doi.org/10.1016/j.jviromet.2007.10.007 (2008).

    Article  CAS  PubMed  Google Scholar 

62. Blight, K. J., McKeating, J. A. & Rice, C. M. Highly permissive cell lines for subgenomic and genomic hepatitis C virus RNA replication. J. Virol. 76, 13001–13014 https://doi.org/10.1128/jvi.76.24.13001-13014.2002 (2002).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

63. Zhong, J. et al. Robust hepatitis C virus infection in vitro. Proc. Natl Acad. Sci. USA 102, 9294–9299 https://doi.org/10.1073/pnas.0503596102 (2005).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

64. Lulla, V. et al. Timeliness of proteolytic events is prerequisite for efficient functioning of the alphaviral replicase. J. Virol. 92, 14 https://doi.org/10.1128/JVI.00151-18 (2018).

65. Whelan, S. P., Ball, L. A., Barr, J. N. & Wertz, G. T. Efficient recovery of infectious vesicular stomatitis virus entirely from cDNA clones. Proc. Natl Acad. Sci. USA 92, 8388–8392 https://doi.org/10.1073/pnas.92.18.8388 (1995).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

66. Kleinfelter, L. M. et al. Haploid genetic screen reveals a profound and direct dependence on cholesterol for hantavirus membrane fusion. mBio 6, e00801 https://doi.org/10.1128/mBio.00801-15 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

67. Acuna, R. et al. Hantavirus Gn and Gc glycoproteins self-assemble into virus-like particles. J. Virol. 88, 2344–2348 https://doi.org/10.1128/JVI.03118-13 (2014).

    Article  CAS  PubMed  Google Scholar 

68. Xie, Y. et al. IBS 2.0: an upgraded illustrator for the visualization of biological sequences. Nucleic Acids Res. 50, W420–W426 https://doi.org/10.1093/nar/gkac373 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

69. Gouet, P., Courcelle, E., Stuart, D. I. & Metoz, F. ESPript: analysis of multiple sequence alignments in PostScript. Bioinformatics 15, 305–308 https://doi.org/10.1093/bioinformatics/15.4.305 (1999).

    Article  CAS  PubMed  Google Scholar 

70. Gouet, P. & Courcelle, E. ENDscript: a workflow to display sequence and structure information. Bioinformatics 18, 767–768 https://doi.org/10.1093/bioinformatics/18.5.767 (2002).

    Article  CAS  PubMed  Google Scholar 

71. Gouet, P., Robert, X. & Courcelle, E. ESPript/ENDscript: Extracting and rendering sequence and 3D information from atomic structures of proteins. Nucleic Acids Res. 31, 3320–3323 https://doi.org/10.1093/nar/gkg556 (2003).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

72. Robert, X., Guillon, C. & Gouet, P. FoldScript: a web server for the efficient analysis of AI-generated 3D protein models. Nucleic Acids Res. 53, W277–W282 https://doi.org/10.1093/nar/gkaf326 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

73. Waskom, M. seaborn: statistical data visualization. J. Open Source Softw. 6, 3021 https://doi.org/10.21105/joss.03021 (2021).

74. Wang, J. Code for ‘Structural basis for protocadherin-1 recognition by New World hantaviruses’. Zenodo https://doi.org/10.5281/zenodo.22751343 (2026).

Download references

Author notes

1. Catalina Florez & Nicole M. Josleyn

   Present address: The Henry M. Jackson Foundation, Bethesda, MD, USA

2. Lohit Khera

   Present address: Norgen Biotek, Thorold, Ontario, Canada

3. These authors contributed equally: Leandro Battini, Megan M. Slough, Ramandeep Kaur

Authors and Affiliations

1. Structural Biology of Infectious Diseases G5+ Unit, Institut Pasteur, Université Paris Cité, Paris, France

   Leandro Battini, Annalisa Meola & Pablo Guardado-Calvo

2. Department of Microbiology and Immunology, Albert Einstein College of Medicine, New York, NY, USA

   Megan M. Slough, Gorka Lasso & Kartik Chandran

3. Department of Microbiology and Immunology, Center for Applied Immunology and Pathological Processes, Center for Cardiovascular Diseases and Sciences, Louisiana State University Health Sciences Center-Shreveport, Shreveport, LA, USA

   Ramandeep Kaur, Upendra P. Lambe, Lohit Khera, Jian Wang, Rona S. Scott & Rohit K. Jangra

4. Department of Molecular Biosciences, The University of Texas at Austin, Austin, TX, USA

   Luqiang Guo & Jason S. McLellan

5. U.S. Army Medical Research Institute of Infectious Diseases, Fort Detrick, MD, USA

   Catalina Florez, Nicole M. Josleyn & Andrew S. Herbert

6. Institut Pasteur, Université Paris Cité, CNRS UMR 3569, Unité de Virologie Structurale, Paris, France

   Alexandra Serris & Félix A. Rey

7. Institute of Infection, Immunity and Transplantation, University College London, London, UK

   Gorka Lasso

8. Laboratorio de Virología Molecular, Centro Ciencia & Vida, Fundación Ciencia & Vida, Santiago, Chile

   Nicole D. Tischler

9. Escuela de Bioquímica, Facultad de Ciencias, Universidad San Sebastián, Santiago, Chile

   Nicole D. Tischler

10. Ichor Biologics, New York, NY, USA

    Raymond Alvarez, Jose L. Garrido & Maria I. Barria

11. Facultad de Medicina, Universidad San Sebastián, Puerto Montt, Chile

    Jose L. Garrido

12. Montefiore-Einstein Comprehensive Cancer Center, Albert Einstein College of Medicine, New York, NY, USA

    Kartik Chandran

Contributions

L.B., M.M.S., J.S.M., K.C. and P.G.-C. conceptualized the project. L.B., M.M.S., R.K., L.G., C.F., G.L., J.S.M., R.K.J., K.C. and P.G.-C. developed the methodology and performed formal analyses. L.B., M.M.S., R.K., L.G., C.F., A.S., A.M., U.P.L., L.K., J.W., R.S.S., J.L.G., M.I.B., N.M.J., F.A.R., A.S.H., J.S.M., R.K.J., K.C. and P.G.-C. performed the investigation. N.D.T., R.A., J.S.M., K.C. and P.G.-C. provided resources. K.C. and P.G.-C. wrote the manuscript. L.B., M.M.S., L.G., C.F., N.D.T., A.S.H., J.S.M., F.A.R., R.K.J., K.C. and P.G.-C. reviewed and edited the manuscript. J.S.M., R.K.J., K.C. and P.G.-C. supervised the project and acquired funding. All of the authors provided feedback on the manuscript draft.

Corresponding authors

Correspondence to Rohit K. Jangra, Kartik Chandran or Pablo Guardado-Calvo.

Competing interests

N.D.T., P.G.-C. and F.A.R. are named inventors on a patent describing orthohantavirus Gn–Gc spike stabilization (PCT/US19/22134); K.C. holds equity in Integrum Scientific and Eitr Biologics. R.K.J., L.K., R.K. and K.C. are listed as inventors of the SHV platform.

Peer review information

Nature thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions