Mechanistic and antigenic boundaries of Henipavirus and Parahenipavirus glycoproteins Authors: Aaron J. May, Muralikrishna Lella, Jared Lindenberger, Alex Berkman, Ujjwal Kumar, Kejun Liu, Moumita Dutta, Maggie Barr, Rob Parks, Xiaozhi Lu, Madison Berry, Amanda Powell, Amelia G. Thompson, Sravya Sowdamini Nakka, Camila T. França, Xiao Huang, Arpita Mrigwani, Kijun Song, Victor Ilevbare, Salam Sammour, Chan Soo Park, Radha Devkota Adhikari, Priyanka Devkota, Katarzyna Janowska, Yanshun Liu, Garrett Scapellato, Taylor N. Spence, Katayoun Mansouri, Kevin Wiehe, Nancy J. Sullivan, Rosemarie Mason, Robert J. Edwards, Kevin O. Saunders, Barton F. Haynes, Priyamvada Acharya Published: 2026 ABSTRACT Henipaviruses, in the Paramyxoviridae family, includes the highly virulent Nipah virus that causes reoccurring outbreaks of deadly disease. Recent discoveries of Henipavirus-like species, including the zoonotic Langya virus, have revealed much higher antigenic diversity than currently characterized and prompted the reorganization of these viruses into the Henipavirus and Parahenipavirus genera. Here, to explore the limits of structural and antigenic variation in both genera, collectively referred to as HNVs, we construct an expanded, diverse panel of HNV fusion and attachment glycoproteins from non-redundant HNV strains that better reflect global HNV diversity. We express and purify the fusion protein ectodomains and the attachment protein head domains and study their biochemical and biophysical properties. We perform immunization experiments in mice, eliciting antibodies reactive to multiple HNV fusion proteins. Cryo-electron microscopy structures elucidate molecular determinants of differential pre-fusion state stability and higher order contacts. A crystal structure of the Gamak virus attachment head domain reveals an additional domain appended to the conserved 6-bladed, β-propeller fold. Taken together, these studies expand the known structural and antigenic limits of the HNVs, reveal cross-reactive epitopes within both genera and provide foundational data for the development of broadly reactive countermeasures. Introduction Henipaviruses (HNVs) are a genus of single-stranded RNA viruses that includes the highly virulent Nipah virus (NiV) responsible for reoccurring outbreaks of deadly disease. The Henipavirus genus is a part of the Paramyxoviridae family, and therefore related to other notable human pathogens such as the parainfluenza viruses, measles, and mumps1. The combined possibility within the family for both rapid transmission, as with measles, and high lethality2, as with NiV, further highlights pandemic risk and necessitates urgent research to establish preparedness that will enable a rapid and effective response to a future emergent threat. For HNVs specifically, their identification in diverse animal reservoirs in different geographical locations, their high risk of zoonotic transmission3,4, and current lack of approved vaccines or therapies5,6 to treat HNV infection in humans highlights the high risk and pandemic potential of this genus. HNVs have been studied since the discovery of Hendra (HeV) and Nipah viruses in 1994 and 1998, respectively7–9. Additional members of the genus have been identified since then, but it was not until 2022, with the discovery of Langya virus (LayV), that another species with pathogenicity in humans was confirmed10. Like HeV and NiV, LayV was of zoonotic origin. However, unlike HeV, NiV, and most of the other known HNVs at the time, LayV was found to have likely originated from a shrew reservoir, rather than from fruit bats. Several concurrent studies identified more shrew-borne HNVs over a wide geographic range, indicating the existence of a largely uncharacterized clade within the Henipavirus genus11–15. Among these studies were some untargeted surveillance efforts, revealing HNV sequences among dozens of other sequences across multiple virus families16–18. While these shrew-borne species had initially been classified as members of the Henipavirus genus, they have recently been reclassified into the Parahenipavirus genus19. In this study, we use the abbreviation HNV to refer collectively to species of both genera. The HNV attachment (G) and fusion (F) surface glycoproteins together facilitate virus entry into host cells, with G responsible for receptor binding and F mediating viral and host membrane fusion. During this process, both G and F undergo important conformational changes, including destabilization of the prefusion conformation of F by G. The specifics of these conformational steps, including how G destabilizes the pre-fusion F conformation, triggering the pre-fusion to post-fusion transition, remain elusive. Although a class I fusion protein20, the HNV-F protein, along with all Paramyxoviridae F proteins, differs from the canonical class I fusion mechanism in that there is a separation of the receptor attachment function into a separate protein21. Whereas other viral fusion proteins proceed through the fusion conversion process by repositioning or removal of their attachment subunits, HNV-F proteins retain inherent metastability through F protein architecture alone. In addition to their role in virus entry, F and G are the sole targets for neutralizing antibodies on the surface of HNVs, and thus are important from the perspective of preventative and therapeutic countermeasure development22. The antigenicity of the F and G proteins, outside those of the more well-studied Nipah and Hendra viruses, and the possibility of cross-reactivity across the genus remains largely unexplored. While there are several known monoclonal antibodies (mAbs) that are able to neutralize both NiV and HeV23–27, so far none of these have been found to have any reactivity with LayV28,29. Of the few mAbs known to bind to LayV, some have been noted to be cross-reactive with Mojiang virus (MojV), but not to NiV or HeV29. Whether there are any broadly reactive anti-HNV antibodies or where the limits of cross-reactivity are within the genera considering recently-discovered species remains unknown. To address these genus-wide mechanistic and antigenic questions, here we assemble a curated set of HNV-G and -F sequences that samples the diversity of the genus as broadly as possible. We purify and characterize these proteins biochemically, biophysically, and structurally. We elicit and isolate antibodies with broad reactivity and identify a strategy for pre-fusion conformation stabilization of F proteins across the HNV genera. Results Identification and classification of HNV species To begin categorizing this diverse set of HNV strains, we identified as many non-redundant HNV-G and -F ectodomains as could be found and assessed the phylogeny of these sequences. We used NCBI BLAST30 to search for non-redundant HNV-like sequences related to the input sequence, NC_00272831. We also carried out an extensive literature search to include other sequences that may not have been discoverable through BLAST, such as Gamak virus11. Next, we narrowed the list to include only sequences from mostly complete genomes, leaving out those where only partial DNA sequences of G or F proteins were available. Almost all the strains that were selected had genome lengths close to the canonical 18.2 kbp for HNVs32, with some of the species selected having slightly longer genomes reported. No sequence in our panel had a deposited genome shorter than 11.2 kbp. The several sequences that were shorter than the rest were either Nipah-Malaysia or Hendra strains that we included since they have been well-characterized in the literature. Next, Clustal Omega33 was used to align the G and F amino acid sequences. To generate the soluble ectodomain constructs, we truncated the sequences, using trends in the sequence alignments to decide on exact truncation points between species. All HNV-F ectodomains were of similar length, allowing all to be truncated at the C-terminal end of the ectodomain at the site equivalent to NiV (Malaysia strain NC_002728) residue 488, consistent with previous F ectodomain constructs28 (Supplementary Data 1 and 2). For the G proteins, the domain boundaries did not align as consistently across species as they do with the overall less variable F protein, though the N-terminal portion of the sequence up to the end of the transmembrane domain was relatively consistent across all species, allowing for the truncation position for the ectodomain to be set at the analogous site to NiV residue 71. While the full G ectodomain and its elusive fusion promotion mechanism is an important area for research, in this study, we focus on the G head domain that harbors the receptor binding site and is targeted by a myriad of neutralizing antibodies23,25,34,35. Therefore, while our sequence and phylogenetic analysis considers the whole genome or full G amino acid sequence, only the head domains of G proteins were purified. For these head domains, sequence variability resulted in the transition between the neck domain and head domain being less clear from sequence (Supplementary Data 3). In classical Henipavirus sequences, there are frequent glycine and proline residues leading up to a conserved cysteine in the head domain. The Parahenipaviruses also contain a proline-rich region before the head domain, with as many as five consecutive prolines in some sequences. This proline-rich region is often followed in Parahenipaviruses by many aspartic and glutamic acid residues. Ultimately, for each strain, we sought to select the site analogous to NiV residue 177 as a suitable start point based on sequence analysis and available structural data for Nipah and Langya G proteins23,36 (Supplementary Data 1 and 3). After removing redundant sequences where differences were noted only in the signal sequence, transmembrane domains, cytosolic domains, or other regions not included in the final F ectodomain or G head domain constructs, the panel was assembled into a phylogenetic tree based on complete genomes (Fig. 1 and Supplementary Fig. 1A). From this list, a subset of non-redundant F ectodomains and G head domains was purified that would broadly sample the diversity of species in the phylogenetic tree, resulting in 35 purified F ectodomains and 32 purified G head domains. A listing of the sequence source for each purified protein can be found in Supplementary Data 1 and exact amino acid sequences in Supplementary Data 2 and 3.Fig. 1Selection and classification of henipavirus species.A phylogenetic tree based on full-genome nucleotide sequences of the selected strains. Horizontal branch lengths are to scale to represent sequence divergence. Hollow boxes are drawn to represent broad groupings within the tree, such as genus, with Classical/Bat Henipaviruses (green) and Shrew/Parahenipaviruses (orange), as well as potential species groupings, such as Nipah viruses (dark red) and Hendra viruses (dark blue). Solid highlight boxes are drawn to represent more narrow groupings of specific species or subspecies groupings, such as Nipah-Malaysia (yellow), Nipah-Bangladesh (purple), and Langya-China (tan). The tips of each branch are labeled with the strain names and associated gen bank accession number. The abbreviations used are described in more detail in Supplementary Note 1. Each tip is colored based on which non-redundant glycoproteins are available from that strain: non-redundant F ectodomain only, blue, non-redundant G ectodomain only, yellow, no non-redundant G head domain, dashed red border, both non-redundant F and G ectodomain, green. Selection and classification of henipavirus species. A phylogenetic tree based on full-genome nucleotide sequences of the selected strains. Horizontal branch lengths are to scale to represent sequence divergence. Hollow boxes are drawn to represent broad groupings within the tree, such as genus, with Classical/Bat Henipaviruses (green) and Shrew/Parahenipaviruses (orange), as well as potential species groupings, such as Nipah viruses (dark red) and Hendra viruses (dark blue). Solid highlight boxes are drawn to represent more narrow groupings of specific species or subspecies groupings, such as Nipah-Malaysia (yellow), Nipah-Bangladesh (purple), and Langya-China (tan). The tips of each branch are labeled with the strain names and associated gen bank accession number. The abbreviations used are described in more detail in Supplementary Note 1. Each tip is colored based on which non-redundant glycoproteins are available from that strain: non-redundant F ectodomain only, blue, non-redundant G ectodomain only, yellow, no non-redundant G head domain, dashed red border, both non-redundant F and G ectodomain, green. The International Committee on Taxonomy of Viruses (ICTV) has recently updated terminology related to the HNVs, establishing the Henipavirus and Parahenipavirus genera, and specifying select species within these groups19. Our effort to compare the properties of dozens of non-redundant F and G sequences required finer granularity than the species level, and as a result, for this study, we adopted an abbreviation and numbering scheme to aid in distinguishing specific sequences. The basis for this terminology is expanded upon in Supplementary Note 1, which builds upon many individual viral surveillance efforts10–12,16,37–45. The corresponding accession code and current ICTV name for each sequence is listed in Supplementary Data 1. For F proteins, the sequence identity relative to NiV-M1-F is high for both NiV and HeV strains, with all NiV sequences over 98% and HeV strains over 97% (Supplementary Fig. 1B). The sequence identity of most non-NiV/HeV strains to NiV-M1-F is ~40%, with sequence similarity around 55–60%, revealing that the F protein is relatively well conserved across genera. A phylogenetic tree of HNV-F proteins based on amino acid sequences was overall similar to the whole genome tree with several small differences. Specifically, the distinction between NiV-C-F and NiV-M strains is more apparent in the F-based phylogenetic tree, the F proteins of SHNV3 and SHNV11 appear to be very closely related, and AngV-F appeared distinct from all other HNV strains, rather than part of the bat clade as seen through full-genome phylogeny (Supplementary Fig. 2, Top). Compared to F proteins, HNV-G proteins demonstrate much greater sequence variability. Sequence identity values are only ~78% between NiV and HeV, and drop to below 20% for Parahenipavirus G proteins, compared to NiV-M1-G (Supplementary Fig. 1C). The sequence similarity scores are also rather low, with most Parahenipavirus strains at around 35% similarity to NiV (Supplementary Fig. 1C). When considering amino acid-based phylogeny, the relationships between strains remain similar as we had observed for the F ectodomains (Supplementary Fig. 2, Top). One notable difference is that SHNV3-G and SHNV11-G, as with their F counterparts, appear more closely related than their full genome would suggest (Supplementary Fig. 2, Bottom). Overall, the limited sequence similarity observed for the G head domains suggests that there could be greater antigenic, structural, and mechanistic differences between different HNV-G proteins than among F proteins. Preparation of diverse HNV-F ectodomains We expressed and purified 35 HNV-F ectodomain constructs using a C-terminal Twin-Strep tag, followed by size exclusion chromatography (SEC). We obtained variable yields of F proteins, with some constructs expressing over 2 mg/L and others less than 0.1 mg/L (Supplementary Fig. 1). For eight F proteins, the yields were too low to perform any downstream studies. We found that yields could vary dramatically between strains that only differed by a few amino acids. For example, NiV-B1.1-F and NiV-B1.2-F differ by only three residues in the ectodomain, but while NiV-B1.2-F produced ~1 mg/L, NiV-B1.1-F had no yield across two purification attempts. HNV-F ectodomains typically eluted from SEC with three peaks: one that eluted immediately after the column void volume, a broad middle peak, and the main peak eluting last, which corresponds to the molecular weight of an F trimer (Fig. 2A and Supplementary Fig. 3). The ratio of the peaks differed from strain to strain, with some strains showing a shoulder for the middle peak, suggesting multiple sized components eluting within this peak (Supplementary Fig. 1B). Sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) of the middle peak fractions showed a major band at the molecular weight expected for the trimeric ectodomain unit, indicating that the presence of the higher molecular weight peak was not a result of covalent linkage or differential glycosylation that would change the molecular weight enough to be distinguishable on SDS-PAGE (Supplementary Fig. 4). This conclusion is supported by mass photometry, where both the middle and main peaks, individually or mixed together, generated an essentially identical molecular weight profile, with each being dominated by the expected molecular weight of the trimer, and both showing a very small presence of species ~2–3 times the mass of the trimer (Fig. 2A, Insert and Supplementary Fig. 5). SDS-PAGE also indicated that all F proteins are being expressed in their F0 state, without any proteolytic processing to F1/F2 (Supplementary Fig. 4). Negative stain electron microscopy (NSEM) of each middle or main peak reveals comparable sets of two-dimensional class averages. The typical pre- and post-fusion shapes identified previously28 can be seen throughout the panel for both middle and main peak components (Supplementary Fig. 6). Taken together, these data suggest transient, non-covalent association of the F ectodomains, reminiscent of previous studies of Paramyxoviridae F proteins where oligomeric states were reported for the F protein trimer46–48. In summary, though the F protein expression yields varied greatly, similar profiles were observed through SEC, SDS-PAGE, and NSEM analysis across all HNV species tested.Fig. 2Purification and antigenic characterization of henipavirus fusion proteins.A Representative SEC profiles of HNV-F Proteins. The oligomer-of-trimer peak (middle peak) and the main trimer peaks are collected separately. NiVop08 is a pre-fusion stabilized construct49 not part of the panel but purified for a immunization study and for comparison to wild-type proteins. An inset to the right shows mass photometry analysis of NiV-M1-F SEC components. Individual measurements are binned in 10 kDa bin sizes. The graph reports the count of events for each molecular weight bin with the main, middle, and mixed peak data overlayed. B Heatmap of antibody binding to HNV-F proteins, on a spectrum of blue for low binding to red for high binding. The left-side graph is based on ELISA data with immobilized F proteins, antibody analytes, and anti-Human Fab-conjugated HRP as a reporter. The values for each combination are based on the log of the area under the curve for the plotted absorbance over increasing concentrations. The right-side graph is based on BLI with immobilized F proteins and direct measurement of analyte to the BLI sensors, either mouse Fabs or mouse IgG (22F5 and 1C8). The values for each combination are based on the maximum wavelength shift measured during the association phase of the BLI experiment. HNV-F antigens colored as in Fig. 1, antibody analytes colored as follows: green, previously published anti-NiV-F binding; salmon, anti-LayV-F binding; gray, anti-influenza HA negative control antibody (ELISA only). C Heatmap of neutralizing potency in units of EC50 (µg/mL) of antibodies against a series of Nipah and Hendra pseudoviruses. The strain used for F and G proteins is listed with the corresponding protein from the panel below. In cases where the corresponding ectodomain was not purified in this study but is identical to one that was, the available protein is shown in brackets below. In cases where no identical ectodomain was purified, the name is marked with an asterisk. Colors for proteins and antibodies as in (B). EC50 values based on the average of two independent experiments (with the exception of one experiment for 22F5), that are in turn comprised of two technical replicates. D, E Cryo-EM maps of antibodies 8C7 (teal), 20G7 (orange), and 9A9 (pink) bound to either NiV-M1-F or HeV-α1.2-F, colored by domain. HNV Henipavirus, all other species abbreviations are described in Supplementary Note 1. Source data are provided as a Source Data file. Purification and antigenic characterization of henipavirus fusion proteins. A Representative SEC profiles of HNV-F Proteins. The oligomer-of-trimer peak (middle peak) and the main trimer peaks are collected separately. NiVop08 is a pre-fusion stabilized construct49 not part of the panel but purified for a immunization study and for comparison to wild-type proteins. An inset to the right shows mass photometry analysis of NiV-M1-F SEC components. Individual measurements are binned in 10 kDa bin sizes. The graph reports the count of events for each molecular weight bin with the main, middle, and mixed peak data overlayed. B Heatmap of antibody binding to HNV-F proteins, on a spectrum of blue for low binding to red for high binding. The left-side graph is based on ELISA data with immobilized F proteins, antibody analytes, and anti-Human Fab-conjugated HRP as a reporter. The values for each combination are based on the log of the area under the curve for the plotted absorbance over increasing concentrations. The right-side graph is based on BLI with immobilized F proteins and direct measurement of analyte to the BLI sensors, either mouse Fabs or mouse IgG (22F5 and 1C8). The values for each combination are based on the maximum wavelength shift measured during the association phase of the BLI experiment. HNV-F antigens colored as in Fig. 1, antibody analytes colored as follows: green, previously published anti-NiV-F binding; salmon, anti-LayV-F binding; gray, anti-influenza HA negative control antibody (ELISA only). C Heatmap of neutralizing potency in units of EC50 (µg/mL) of antibodies against a series of Nipah and Hendra pseudoviruses. The strain used for F and G proteins is listed with the corresponding protein from the panel below. In cases where the corresponding ectodomain was not purified in this study but is identical to one that was, the available protein is shown in brackets below. In cases where no identical ectodomain was purified, the name is marked with an asterisk. Colors for proteins and antibodies as in (B). EC50 values based on the average of two independent experiments (with the exception of one experiment for 22F5), that are in turn comprised of two technical replicates. D, E Cryo-EM maps of antibodies 8C7 (teal), 20G7 (orange), and 9A9 (pink) bound to either NiV-M1-F or HeV-α1.2-F, colored by domain. HNV Henipavirus, all other species abbreviations are described in Supplementary Note 1. Source data are provided as a Source Data file. Antigenicity of HNV-F ectodomains We next characterized the antigenicity of the F ectodomain panel. Since many HNV species have been recently identified and given the limited availability of antibodies that react with LayV and other members of our panel, we sought to identify antibodies specific to these species by vaccinating mice with HNV-F antigens (Supplementary Fig. 7 and Supplementary Table 1). One group was given four injections of the LayV-C1-F ectodomain, and for two others, the fourth injection of LayV-C1-F was replaced with NiV-F protein constructs, either the wild-type NiV-M1-F or NiVop08, a pre-fusion stabilized NiV-F construct49. By boosting with NiV-F constructs, we sought to elicit broadly reactive antibodies. Using hybridoma fusion technology50, we identified two murine IgG1 antibodies, named 1C8 and 22F5, both from group 2, with reactivity to LayV-C1-F. An additional immunization series using the same antigens was performed with a modified timeline, based on observations from the first series that antibody titers stabilized after the second immunization. This series led to the isolation of an additional four murine antibodies, all from group 2, named 9A9, 8C7, 20G7, and 8D4 (Supplementary Fig. 7 and Supplementary Tables 1 and 2). The binding of antibodies elicited in our immunization experiments, as well as previously identified anti-NiV or anti-LayV antibodies, to our panel of HNV-F ectodomains was assessed by enzyme-linked immunosorbent assay (ELISA), surface plasmon resonance (SPR), or Bio-layer interferometry (BLI)29,51 (Fig. 2B and Supplementary Figs. 8–10). 1H1 and 4B8, previously identified anti-NiV-F antibodies, displayed broad reactivity to all NiV and HeV strains, but minimal reactivity to all other species. A panel of previously characterized mouse anti-NiV-F Fabs broadly reacted to nearly all NiV and HeV antigens, as expected. 4G5, previously identified as an anti-LayV-F and MojV-F antibody that binds both the pre- and post-fusion conformations at an epitope involving the HRA and HRB regions29, strongly bound not only to LayV-F and MojV-F, but also to SHNV5-1-F and SHNV5-2-F, thus indicating that antibodies targeting the 4G5 epitope have greater breadth than was previously known (Fig. 2B and Supplementary Figs. 9 and 10C). Of the antibodies isolated from the immunization studies, 8C7 and 20G7 bound only to NiV-F ectodomains, while 9A9 bound to both NiV-F and HeV-F (Supplementary Fig. 10A–C). 1C8, 22F5, and 8D4 demonstrated strong reactivity not only to LayV-F proteins, but also to MojV-F and both SHNV5 strains (Supplementary Fig. 10C). Moreover, 22F5, which generated stronger binding signals than 1C8, bound to SHNV2-F as well (Fig. 2B). 8D4 demonstrated the broadest binding profile, with binding detected to F proteins from each of GakV-1, GakV-2, NinV, SHNV1, SHNV2, and SHNV3 (Supplementary Fig. 10C). Considering the phylogenetic relationships of these strains (Fig. 1), it appears that 8D4’s reactivity covers nearly the entirety of the known Parahenipaviruses. However, 8D4 demonstrated sharply reduced binding to pre-fusion stabilized F constructs, indicating that it specifically bound to the post-fusion F conformation (Supplementary Fig. 10C). To assess the potential to target the HNV-F protein glycans, we tested the binding of 2G12, a glycan-reactive antibody previously noted for its broad specificity to diverse viral fusion proteins and ability to neutralize influenza and HIV-152–54. We did not detect any binding between 2G12 and the HNV-F panel (Supplementary Fig. 10C). We also tested the reactivity of recombinant Griffithsin, a lectin discovered in Griffithsia red algae that has been investigated for antiviral properties55,56 and has been shown to be protective against NiV challenge in hamsters57. Griffithsin bound to most of the NiV/HeV strains in the panel, while also binding to two pre-fusion stabilized Parahenipavirus constructs, of SHNV5-1 and GakV-1, specifically (Supplementary Fig. 10C), underscoring the potential for a glycan-directed approach for broad spectrum targeting of HNVs. We further assessed the efficacy of these antibodies in the context of virus neutralizing activity, using either a fluorescence-based pseudovirus entry assay or an authentic virus plaque reduction assay, both utilizing various Nipah or Hendra virus strains. In the pseudovirus neutralization assays, antibodies 8C7, 20G7, and 9A9 neutralized several NiV strains with equivalent or even better potency than previously known neutralizing antibodies such as 4B851, m102.458, 5B327, and MBP-1F524 that were also assessed within the same assay (Fig. 2C and Supplementary Fig. 11A). Of these, 9A9 also neutralized two Hendra virus strains, a 2008 Redlands strain (Genbank accession number HM044317.1) and a 2015 Gympie strain (Genbank accession number MZ318101.1). As expected, given the lack of binding to Nipah and Hendra virus F proteins demonstrated by 8D4, 22F5, and 1C8, these Parahenipavirus-specific antibodies did not neutralize any Nipah or Hendra virus strain. Consistent with the results of the pseudovirus neutralization assay, 20G7 and 8C7 showed the highest levels of neutralization of the NiV isolate tested in the authentic virus assay, with 9A9 showing lower levels of neutralization activity (Supplementary Fig. 11B). All three antibodies, 20G7, 8C7, and 9A9, neutralized the authentic Hendra virus tested, albeit at different levels. Using single particle cryo-EM, we determined the epitopes of the NiV/HeV neutralizing antibodies 8C7 and 20G7 in complex with NiV-M1-F and 9A9 in complex with HeV-α1.2-F (Fig. 2D, E, Supplementary Figs. 12 and 13, and Supplementary Table 3). In all cases, the antibody bound with a 3:1 stoichiometry to the F protein, with epitopes positioned solely in the DIII region, a common site for NiV-F neutralizing epitopes24,26,51,59. The epitopes for 8C7 and 20G7 are very similar, both close to the F protein apex, whereas the 9A9 epitope is distinct and shifted toward the side face of the F protein. Taken together, our findings align with the current understanding that Henipaviruses and Parahenipaviruses have distinct antigenic profiles. Parahenipavirus F proteins are not bound by known anti-NiV/HeV antibodies. However, the breadth of strains bound by our LayV-F-reactive antibodies indicates that cross-neutralization of a sizeable portion of known Parahenipaviruses with a small number of antibodies may be an achievable goal. While the neutralizing potency of the Parahenipavirus-specific antibodies remains unknown, our immunization and antibody isolation efforts have generated several Henipavirus neutralizing antibodies. Structural characterization of antibody 22F5 Since there are very few well-characterized antibodies that have been described for Parahenipaviruses, we selected 22F5 for detailed characterization (Supplementary Figs. 14–16 and Supplementary Table 2), given its combination of a broad cross-reactivity profile and ability to bind pre-fusion F proteins. Using SPR, we tested the binding of 22F5 to LayV-F ectodomains in various conformational states, using two separate disulfide stabilized pre-fusion LayV-F constructs described previously, that we refer to here as DS47 and DSv229 (Supplementary Fig. 14A). Additionally, we tested binding with and without competition with 4G5 (Supplementary Fig. 14B). 22F5 bound to LayV-F in both pre- and post-fusion conformations, indicating conformational bispecificity, and its binding was mostly unaffected by the presence of 4G5, indicating a distinct epitope. Using single particle cryo-EM, we determined a structure of the 22F5 Fab bound to the post-fusion form of the LayV-C1-F protein, at a global resolution of 3.31 Å (Fig. 3A, Supplementary Fig. 15, and Supplementary Table 3), with local resolution of ~3 Å at the interface of 22F5 with the F protein. 22F5 bound to the DII region of LayV-F, with a few DI contacts as well (Regions of HNV-F proteins described previously22). While there have been several previously observed antibodies targeting DII epitopes in NiV-F and HeV-F proteins51, the only other available structure for an anti-LayV-F antibody, 4G5, details an epitope that includes contributions from HRA and HRB29. Additionally, we determined the structure of the 22F5 Fab bound to the LayV-F-DS ectodomain at ~4 Å local resolution at the interface (Fig. 3B, Supplementary Fig. 16, and Supplementary Table 3), confirming that the 22F5 epitope is accessible in both the pre- and post-fusion conformation of LayV-F. For both LayV-C1-F and LayV-F-DS, 22F5 was seen bound at all three DII sites.Fig. 3Structure and binding kinetics of 22F5 Fab with LayV-F ectodomain.A, B Cryo-EM maps of 22F5 Fab bound to post-fusion LayV-C1-F (A) or pre-fusion LayV-F-DS (B) with domains colored and labeled. C Views of the models (Left: Post-Fusion, Right: Pre-Fusion) built from the cryo-EM maps in cartoon representation and colored by domain or CDR. D Hydrophobic and polar contacts between 22F5 Fab and the post-fusion LayV-C1-F. Interacting residues labeled and shown in stick representation, with the color scheme continued from (C). E Measurement of affinity and kinetics of binding of 22F5 Fab to LayV-F ectodomains. The black lines in the graphs are the blank subtracted sensorgrams and the red lines are the fit of the data to a 1:1 Langmuir binding model. Kinetics are written into each graph. For LayV-C1-F WT: RMax (RU): 4.30E02, Chi2 (RU2): 18.3, SE (ka): 5.20E04, SE (kd): 1.60E-05, LayV-C1-F Heat Treated: RMax (RU): 433.4, Chi2 (RU2): 12.5, SE (ka): 6.70E04, SE (kd): 1.50E-05, LayV-F-DS: RMax (RU): 172.7, Chi2 (RU2): 18.3, SE (ka): 2.10E05, SE (kd): 5.60E-04. HRA Heptad repeat A, HRB Heptad repeat B, DI,II,III Domain I,II,III, LayV Langya virus. Source data are provided as a Source Data file. Structure and binding kinetics of 22F5 Fab with LayV-F ectodomain. A, B Cryo-EM maps of 22F5 Fab bound to post-fusion LayV-C1-F (A) or pre-fusion LayV-F-DS (B) with domains colored and labeled. C Views of the models (Left: Post-Fusion, Right: Pre-Fusion) built from the cryo-EM maps in cartoon representation and colored by domain or CDR. D Hydrophobic and polar contacts between 22F5 Fab and the post-fusion LayV-C1-F. Interacting residues labeled and shown in stick representation, with the color scheme continued from (C). E Measurement of affinity and kinetics of binding of 22F5 Fab to LayV-F ectodomains. The black lines in the graphs are the blank subtracted sensorgrams and the red lines are the fit of the data to a 1:1 Langmuir binding model. Kinetics are written into each graph. For LayV-C1-F WT: RMax (RU): 4.30E02, Chi2 (RU2): 18.3, SE (ka): 5.20E04, SE (kd): 1.60E-05, LayV-C1-F Heat Treated: RMax (RU): 433.4, Chi2 (RU2): 12.5, SE (ka): 6.70E04, SE (kd): 1.50E-05, LayV-F-DS: RMax (RU): 172.7, Chi2 (RU2): 18.3, SE (ka): 2.10E05, SE (kd): 5.60E-04. HRA Heptad repeat A, HRB Heptad repeat B, DI,II,III Domain I,II,III, LayV Langya virus. Source data are provided as a Source Data file. The 22F5 Fab interacts with both the pre- and post-fusion LayV-F ectodomains utilizing both heavy and light chains. The L1, L3, and H2 complementarity-determining regions (CDRs) are particularly involved (Fig. 3C and Supplementary Fig. 17). In the pre-fusion conformation of LayV-F, a stretch of residues within DIII known as the fusion peptide is inserted into an interprotomer pocket formed by a neighboring protomer’s DII28. During pre- to post-fusion conformational conversion, the fusion peptide leaves this pocket. The pre-fusion 22F5 epitope includes surface-accessible portions of the fusion peptide, but in the post-fusion conformation, these residues no longer contribute to the 22F5 epitope (Fig. 3C, Bottom). 22F5 binding appears to rely heavily on both polar and non-polar interactions (Fig. 3D). Aromatic residues from both chains, specifically Y33, W50, and W100 from the 22F5 heavy chain and W96 from the light chain, form hydrophobic interactions with LayV-F DII residues I386 and P395. Light chain residues N27D (Kabat numbering), R30, Y32, and Y92 contact LayV-F DII residues R388, M390, and D393. In the heavy chain, N54, R57, and N58 interact with DII residues N384, Q398, and DI residue D299. Additionally, when bound to the post-fusion conformation, the 22F5 light chain N-terminal residue D1 interacts with N285, a residue that belongs to domain DI of the neighboring protomer. When bound to the pre-fusion conformation, 22F5 contacts the neighboring protomer DI via its CDRH2 and heavy chain framework region residues. Due to the shifting of protomers relative to each other during pre- to post-fusion conformational conversion, 22F5 shifts its interaction with the neighboring protomer DI region, losing its CDRH2 and heavy chain framework region contacts and gaining the light chain D1 contact with the post-fusion F protein. (Fig. 3C, D). Using SPR, we determined the affinity of 22F5 for both ectodomain constructs, as well as a heat-treated sample of the wild-type LayV-C1-F that ensured uniform post-fusion conformation, with binding observed at low nM affinities in all cases (Fig. 3E). Interestingly, despite the loss of contacts with the fusion peptide residues when bound to the post-fusion conformation of LayV-F, 22F5 bound tighter to the post-fusion conformation than the pre-fusion conformation. Given that the fusion peptide is a flexible region, it is possible that contact with it is entropically unfavorable, thereby explaining the improved affinity with the post-fusion LayV-F, where this contact was absent. In summary, 22F5 binding and structural data describe an epitope that is accessible in both pre- and post-fusion conformations and reveal a region on the Parahenipavirus F protein that can be targeted for greater breadth than previously recognized. 22F5 is a valuable addition to the antibody toolbox for the Parahenipavirus genus for which relatively few antibodies are known. Differential scanning fluorimetry of HNV-F ectodomains We assessed the stability of the purified F ectodomains using differential scanning fluorimetry (DSF), a label-free method that measures changes in intrinsic fluorescence as the protein unfolds. The DSF profiles of the two wild-type proteins, NiV-M1-F and LayV-C1-F, and one pre-fusion stabilized NiV-F construct, NiVop0849, were consistent with our previous reports28. With each peak in a DSF profile indicating a conformational transition, for the unstabilized F ectodomains we observed one transition between 45 and 65 °C and one or more transitions at higher temperatures, up to over 90 °C (Fig. 4A). LayV-C1-F showed two transitions with a negative peak at 50 °C and a positive peak at 86 °C. NiV-M1-F showed three transitions with a negative peak at 56 °C, and positive peaks at 76 °C and 92 °C. NiVop08 showed two transitions, a negative peak at 66 °C and a positive peak at 76 °C. We next measured the DSF profiles of the same set of proteins after pre-incubation at 60 °C for 90 min, based on previous observations that heat treatment could force pre- to post-fusion conformational conversion in Paramyxoviridae F proteins60. Overlaying all DSF profiles revealed that the lower-temperature negative peak present in all three proteins prior to heat treatment was eliminated in only the two wild-type proteins after heat incubation (Fig. 4A). NSEM analysis confirmed that the wild-type proteins have been converted to the postfusion conformation, as revealed by the characteristic elongated six-helix bundle present in the 2D class averages (Fig. 4B). The NiVop08 pre-fusion stabilized construct retained the more compact shape characteristic of the prefusion state (Fig. 4B). By exposing NiV-M1-F and LayV-C1-F to a series of times and temperatures, we found that 50 °C, 15 min was the minimum amount of heat treatment sufficient to eliminate the pre- to post-fusion conversion DSF peak, indicating that the F protein population was predominantly in the post-fusion state (Supplementary Fig. 18). Together, these results indicate that the negative DSF peak present around 45–60 °C is indicative of a pre- to post-fusion conformational change taking place, and this DSF signature can be utilized to follow population level conformational transition from the pre- to post-fusion forms.Fig. 4Thermostability of henipavirus F proteins.A DSF profiles of a set of HNV proteins showing the first derivative of 350/330 nm emission ratio as a function of temperature. Dotted lines of the same color represent the DSF profile of the same proteins after being incubated at 60 °C for 90 min. B Two-dimensional NSEM class averages of a portion of the heat-treated samples from (A). C DSF profiles of diverse HNV-F species. Sections of the profile are noted, indicating the conformational steps occurring at those temperatures, with highlighting of peaks of interest. D A subset of DSF profiles from (C), containing only those species that do not demonstrate any noticeable conformational conversion peak. Below are representative two-dimensional NSEM class averages for select members of this subset. E A subset of DSF profiles from (C), containing only those species that demonstrate some presence of the standard conformational conversion signal. Below are representative two-dimensional NSEM class averages for select members of this subset. F A subset of DSF profiles from (C), containing only those species that demonstrate a profile notably different from the typical HNV-F species. Below are representative two-dimensional NSEM class averages for the two members of this subset. HNV Henipavirus, all other species abbreviations are described in Supplementary Note 1. Source data are provided as a Source Data file. Thermostability of henipavirus F proteins. A DSF profiles of a set of HNV proteins showing the first derivative of 350/330 nm emission ratio as a function of temperature. Dotted lines of the same color represent the DSF profile of the same proteins after being incubated at 60 °C for 90 min. B Two-dimensional NSEM class averages of a portion of the heat-treated samples from (A). C DSF profiles of diverse HNV-F species. Sections of the profile are noted, indicating the conformational steps occurring at those temperatures, with highlighting of peaks of interest. D A subset of DSF profiles from (C), containing only those species that do not demonstrate any noticeable conformational conversion peak. Below are representative two-dimensional NSEM class averages for select members of this subset. E A subset of DSF profiles from (C), containing only those species that demonstrate some presence of the standard conformational conversion signal. Below are representative two-dimensional NSEM class averages for select members of this subset. F A subset of DSF profiles from (C), containing only those species that demonstrate a profile notably different from the typical HNV-F species. Below are representative two-dimensional NSEM class averages for the two members of this subset. HNV Henipavirus, all other species abbreviations are described in Supplementary Note 1. Source data are provided as a Source Data file. Next, measuring DSF profiles for all purified F proteins for which there was sufficient material purified, typically at least 5 µg, we observed a wide range of conversion peak intensities and temperatures (Fig. 4C–F). While all proteins generated a large positive peak at high temperature, indicative of a denaturation event, not all proteins produced a detectable conversion peak. Fourteen proteins had essentially no detectable signal prior to the high temperature peak (Fig. 4D). For these F ectodomain samples, two-dimensional class averages from NSEM analysis revealed the predominance of post-fusion conformation for these proteins with no detectable pre-fusion F species observed in the datasets, supporting the DSF results. Though not exclusively, most of the proteins with no conversion peak are Parahenipaviruses, and inversely, all of the Parahenipaviruses have no conversion peak, except for LayV strains. Many of the Parahenipavirus F proteins could not be assessed due to insufficient purification yield, and these results may suggest it could be due to inherent instability. Among the F proteins that generated a conversion peak, there were shifts in temperature and intensity (Fig. 4E). These changes did not appear related to species differences, with NiV-M1-F and NiV-M3-F, separated only by one amino acid mutation, being right- and left-shifted compared to the average, respectively. Both LayV strains were the most left-shifted of the set, indicative of lower pre-fusion F stability. NSEM analysis showed a mix of pre- and post-fusion conformations in this set, therefore the presence of this negative DSF peak indicates some amount of pre-fusion conformation present, rather than the presence of a uniform set of pre-fusion proteins. Between the conversion peak and the final high temperature peak is typically a broad, positive peak, present with differing intensities and shapes. Further investigation would be required to determine if this peak is either part of a natural, yet undefined, conformational event or part of the denaturation process. Two proteins of the panel, AngV-F and the sequence-corrected version of GhV-F, GhV(+A)-F (Supplementary Note 1), display DSF profiles that are outliers to the rest of the panel (Fig. 4F). Interestingly, both proteins displayed a profile that seemed closer to NiVop08, with their high temperature positive peak significantly left-shifted. Two-dimensional NSEM classes for these two proteins showed the presence of largely, but not exclusively, pre-fusion conformations. Despite their apparent similarities to a pre-fusion stabilized F construct, these are wild-type proteins that are expected to undergo normal conformational conversion, barring unexpected and dramatic mechanistic differences. The ability of the AngV-F and GhV(+A)-F proteins to undergo conformational conversion from the pre- to post-fusion states is supported by presence of post-fusion F forms in the NSEM 2D classes (Supplementary Fig. 6). We incubated these outlier F proteins under both the 60 °C, 90 min and 50 °C, 15 min conditions. After incubation at 50 °C for 15 min, a condition under which NiV-F and LayV-F proteins show effectively complete transition to the post-fusion conformation, we found that AngV-F and GhV(+A)-F proteins retain much of their native DSF profile, suggesting greater pre-fusion stability. Incubation at 60 °C for 90 min, however, was sufficient to induce a dramatic change in their DSF profiles, suggesting more complete conformational conversion. Since DSF is reliant on primarily tryptophan residues to generate signal, it is possible that differences in DSF profile could arise from differing number and position of tryptophans; however, the F ectodomain has only one tryptophan that is almost universally conserved, with only CedV-1-F mutated at this spot (Supplementary Fig. 19). Furthermore, AngV-F contains one additional tryptophan that is exclusive in our panel to AngV-F. All F ectodomain constructs additionally contain one tryptophan in the foldon tag used to assist with trimerization and two in the Twin-Strep tag used for purification. In general, sequence variability is likely not a significant factor in the differences observed in the DSF profiles of HNV-F proteins. Our results support the utilization of DSF as a rapid analytical technique requiring a small amount of protein for assessing population-level pre- to post-fusion transition of HNV-F proteins. In this study, the DSF analysis was central to our identification of two species that exhibit F pre-fusion conformations that appear more stabilized than in other HNV species. Cryo-EM structures of AngV-F protein Based upon both amino acid sequence and pre-fusion stability as revealed by DSF analysis, AngV-F has divergent properties from the currently characterized F proteins. Therefore, we determined the structure of the AngV-F protein by cryo-EM single particle analysis (Fig. 5, Supplementary Figs. 20–22, and Supplementary Table 3). The cryo-EM dataset revealed three particle populations (Fig. 5A). The most populated class was of the isolated F ectodomain trimer, followed by a dimer-of-trimers population, and finally a population where the F proteins formed a hexameric array. The AngV-F trimer was refined to a global resolution of 4.0 Å and used for detailed examination of the F ectodomain structure and for comparing to the previously known structures of other F proteins. In cryo-EM structures of the HNV-F proteins, the stalk region is typically poorly defined24,28,48,51,61. Our AngV-F cryo-EM structure revealed well-defined density for the stalk, allowing atomic-level modeling.Fig. 5Cryo-EM structure of the AngV-F ectodomain.A Three different particle populations for AngV-F and their representative 2D classes. Cryo-EM reconstruction of top and side view shown for AngV-F trimer (left), dimer-of-trimers (middle), and hexameric lattice (right). B Comparison of the AngV-F, LayV-F, and NiV-F proteins. Top: A slice through the F protein visualizing the central cavity and the placement of the stalk region within the F protein. Bottom: A top-down view showing measurements. C Zoom-in view of the stalk region featuring amino acid present at the start of helix bundle (left) and the electrostatics of the region (right), with sequence alignment at the local region between each of the three species below. The HRB start residue is marked red in the sequence alignment. D AngV-F cryo-EM structure colored by protomer with the glycans shown as spheres and colored by element. Glycans at the different regions (apex, non-canonical NNV, stalk) are labeled. AngV Angavokely virus, NiV Nipah virus, LayV Langya virus, HRB Heptad repeat B. Cryo-EM structure of the AngV-F ectodomain. A Three different particle populations for AngV-F and their representative 2D classes. Cryo-EM reconstruction of top and side view shown for AngV-F trimer (left), dimer-of-trimers (middle), and hexameric lattice (right). B Comparison of the AngV-F, LayV-F, and NiV-F proteins. Top: A slice through the F protein visualizing the central cavity and the placement of the stalk region within the F protein. Bottom: A top-down view showing measurements. C Zoom-in view of the stalk region featuring amino acid present at the start of helix bundle (left) and the electrostatics of the region (right), with sequence alignment at the local region between each of the three species below. The HRB start residue is marked red in the sequence alignment. D AngV-F cryo-EM structure colored by protomer with the glycans shown as spheres and colored by element. Glycans at the different regions (apex, non-canonical NNV, stalk) are labeled. AngV Angavokely virus, NiV Nipah virus, LayV Langya virus, HRB Heptad repeat B. The AngV-F ectodomain exhibited a larger central cavity compared to NiV-F and LayV-F proteins. This expanded cavity is the result of both the stalk region not inserting as far into the cavity as in other HNV-F proteins and the domains surrounding the cavity being spaced farther apart (Fig. 5B). In previously determined HNV-F structures, the helical stalk region begins with a Lysine at its N-terminal end, thus positioning a trio of Lysines, one from each of the protomers, at the bottom of the internal cavity. In AngV-F, we instead observed a proline at the beginning of the stalk (HRB) helix (Fig. 5C). The surface electrostatics of this region are, therefore, strikingly different at this position for AngV-F compared to NiV-F and LayV-F structures. The proximity of the Lysines in LayV-F and NiV-F and their electrostatic repulsion could be a source of metastability in the pre-fusion F protein, as charge repulsion may be a mechanism to force apart the HRB helix during conformational conversion. The absence of this Lysine trio and replacement with a proline cap at the N-terminus of the HRB helix may lend greater stability to the pre-fusion form of the AngV-F protein. Notably, all species in our panel have either a lysine or arginine at this site, with only AngV-F substituting with a proline (Supplementary Data 2). Several glycans could be resolved (N61, N93, N353, N434, and N458), including ones analogous to those previously defined in HNV-F structures48,61. The glycans shield known sites of vulnerability in Paramyxoviridae F proteins, including at the apex (N61), near the fusion peptide (N93 in Domain III), and on the stalk domain (N458). Interestingly, the glycan proximal to the fusion peptide utilizes a non-canonical NNV sequence (Fig. 5D). The MojV-F and SHNV2-F proteins also have NNV sequences at this site that may be similarly glycosylated. Glycans in the apex region near N61 and the HRB/stalk region near N458 are observed in many HNV-F protein structures (Supplementary Fig. 23). However, there is no AngV-F analog to the N414 glycan of HeV-F and NiV-F. Instead, AngV-F contains two glycans in DI (N434) and DII (N353). The non-canonical NNV glycan at N93 is also found at the same site in HeV-F and NiV-F using canonical sequons, but interestingly, not in LayV-F. The differential glycosylation patterns observed across diverse HNV-F species may play a role in determining the potential for cross-reactivity, but critically, the presence of the NNV glycan in AngV-F indicates that prediction of glycosylation through sequence alone may be difficult. Interactions between the AngV-F trimers were mediated through intermolecular contacts along the DI and DIII domains. Similar oligomeric states between pre-fusion F protein molecules (trimer, dimer-of-trimers, trimer-of-trimers) were also found in the cryo-EM dataset of a Hendra virus F protein (HeV-α1.2-F) (Supplementary Fig. 22). AngV-F dimers resemble mirror-like protein dimers, with each exterior interaction surface of a protomer available for dimerization. These dimers, in turn, interact with protomers from three other trimers, forming hexameric structures. Each trimer is consistently aligned within the plane of the hexameric assembly, leading to a horizontally level lattice, distinct from an NiV-F crystal structure in which the lattice was not within one plane48. Taken together, our cryo-EM structures revealed that, while maintaining the overall conserved HNV-F architecture as observed in previously characterized HNV-F proteins, key mutations imparted notably distinct pre-fusion characteristics on the AngV-F protein. They also revealed multimerized pre-fusion F proteins, adding to the growing body of literature that suggests a role for the multimeric assemblies in the context of the virion and suggests a distinct mode of F protein multimerization leading to the formation of hexameric lattices for the AngV-F protein. Structure-guided pre-fusion stabilization of diverse HNV-F proteins We hypothesized that the proline substitution observed at the start of the HRB helix, exclusive in our panel to AngV-F, was the cause of the increased pre-fusion stability. We expressed a series of HNV-F constructs from a variety of HNV species incorporating this proline substitution. Available HNV-F structures indicate that the site analogous to K453 in NiV-M1-F is consistently positioned at the N-terminal start of HRB, and even among other species for which structures are not available, the region surrounding this site is relatively well-conserved (Fig. 6A and Supplementary Data 2).Fig. 6Transferable pre-fusion stabilization of henipavirus F proteins.A Location of HRB proline mutation in select HNV-F proteins. Left: Site of NiV K453P mutation shown as red spheres in the HNV-F ectodomain structure, with protomers colored in shades of blue. Right: A sequence alignment of the residues surrounding the HRB proline mutation, colored by residue type: hydrophobic, orange, polar, pink, negative charge, red, aromatic, white, conformationally significant, green, HRB proline mutation, dark green. B–D Characterization of the HRB proline mutation in three HNV-F species, B NiV-M1-F, C SHNV5-1-F, D GakV-1-F. Top: SEC overlay of the WT (right axis) and mutant (left axis) with mAU values scaled to a 1 L transfection volume. Middle: DSF analysis overlay of the WT and mutant. The middle line represents the average with standard deviation represented with thin lines above and below the average line and shaded in-between. Bottom: Comparison of representative NSEM 2D classes from WT (top) and mutant (Bottom). HRB Heptad repeat B, NiV Nipah virus, SHNV5 Shrew Henipavirus 5, GakV Gamak virus. Source data are provided as a Source Data file. Transferable pre-fusion stabilization of henipavirus F proteins. A Location of HRB proline mutation in select HNV-F proteins. Left: Site of NiV K453P mutation shown as red spheres in the HNV-F ectodomain structure, with protomers colored in shades of blue. Right: A sequence alignment of the residues surrounding the HRB proline mutation, colored by residue type: hydrophobic, orange, polar, pink, negative charge, red, aromatic, white, conformationally significant, green, HRB proline mutation, dark green. B–D Characterization of the HRB proline mutation in three HNV-F species, B NiV-M1-F, C SHNV5-1-F, D GakV-1-F. Top: SEC overlay of the WT (right axis) and mutant (left axis) with mAU values scaled to a 1 L transfection volume. Middle: DSF analysis overlay of the WT and mutant. The middle line represents the average with standard deviation represented with thin lines above and below the average line and shaded in-between. Bottom: Comparison of representative NSEM 2D classes from WT (top) and mutant (Bottom). HRB Heptad repeat B, NiV Nipah virus, SHNV5 Shrew Henipavirus 5, GakV Gamak virus. Source data are provided as a Source Data file. All of the proline mutants were able to be purified in sufficient amounts to permit downstream quantitation, with yields mostly exceeding those of the wild-type construct (Fig. 6B–D and Supplementary Fig. 24A). In the case of the SHNV5-1-K455P mutant, the yield sharply decreased, but DSF analysis of all mutants, including SHNV5-1-K445P-F, revealed significant pre-fusion stabilization (Fig. 6B–D and Supplementary Fig. 24B). The trademark fusion conversion peak was present in all constructs, often with greater intensity than the wild-type, or demonstrating a right-shift of conversion temperature, possibly indicating higher stability of the pre-fusion conformation. In the case of the NiV, HeV, and LayV mutants, the 70–80 °C peak was also present with higher intensity than the wild-type. Whereas GakV-1-F had displayed exclusively post-fusion conformation through both DSF and NSEM analysis (Fig. 6D), the GakV-1-K452P-F DSF profile was dramatically altered, even appearing similar to the profile of the pre-fusion NiVop08 mutant or outliers like AngV-F and GhV(+A)-F. NSEM analysis confirmed the findings from DSF, with pre-fusion 2D classes observable with all mutants. The altered 2D classes of GakV-1-K452P-F, as with its DSF profile, were particularly striking. Whereas the wild-type classes contained solely post-fusion conformations, the mutant classes were almost universally pre-fusion. Despite the strong pre-fusion stabilization across all species, it was still common to observe post-fusion classes, reflecting that as with AngV-F, this mutation stabilizes pre-fusion conformation without preventing conversion (Fig. 6B–D and Supplementary Fig. 24C). We assessed the binding of the NiV and HeV-based mutants, NiV-M1-K458P-F and HeV- α1.2-K453P-F, to the pre-fusion specific 4B8 antibody. Strong binding was observed between 4B8 IgG and all F constructs tested: both wild-type and mutant NiV-M1-F and HeV-α1.2-F, and NiVop08 (Supplementary Fig. 24D). However, after incubation at 50 °C for 15 min, while binding to NiVop08 was maintained, binding to wild-type NiV/HeV was nearly eliminated. The HeV proline mutant’s binding to 4B8 was similarly eliminated, while the NiV mutant’s binding was only moderately reduced. Taken together, our results indicate that pre-fusion stabilization of NiV/HeV through the substitution for the HRB start proline lends greater pre-fusion state stability but still allows for conformational conversion and antigenicity consistent with wild-type constructs. We collected larger NSEM datasets and reconstructed 3D classes to quantify pre- and post-fusion ratios for these mutants and their wild-type counterparts. In all cases, the proportion of pre-fusion particles was increased in the mutant construct compared to wild-type (Supplementary Fig. 25A–F). Interestingly, even for GakV-1-F and SHNV5-1-F, species where no pre-fusion conversion peak was present in DSF, a small amount of pre-fusion classes were detected in the larger NSEM datasets (Supplementary Fig. 25A, B, F). Most striking was the conversion of GakV-1-F from ~10% pre-fusion classes, the least of any species, to over 75% pre-fusion with the proline mutant, the most of any species (Supplementary Fig. 25A, B). Although the use of NSEM results in low-resolution 3D volumes, it is still apparent that the typical pre- and post-fusion class shapes are observed in all species, indicating that the proline mutation has not significantly altered the pre-fusion conformation. In summary, the introduction of this proline presents an effective and translatable means of pre-fusion stabilization for HNV-F proteins, and one that is expected to have minimal impact on any of the antigenic sites on the surface of the protein. Purification and characterization of diverse HNV-G head domains To analyze our panel of HNV-G proteins, in this study, we focused on the isolated head domain, which is responsible for receptor binding and is an important antigenic site22. To allow for the previously described head domain sequences to be secreted in a mammalian expression system, we added an artificial secretion signal62, and appended a C-terminal 8× His tag. The G head domain constructs were transiently transfected and affinity purified with cobalt resin affinity chromatography, followed by SEC. The typical SEC profile for the head domain proteins reflects the simple nature of their single-domain fold. All proteins eluted with a consistent single main peak (Supplementary Fig. 26), typically with a small, higher molecular shoulder or peak likely consisting of aggregates. Yields for the head domains ranged from 2.5 to 35.3 mg/L (Supplementary Fig. 1), with the majority yielding above 10 mg/L. It has previously been noted that LayV and MojV attachment proteins do not bind to Ephrin B2 or B3 ligands (EFNB2/B3), the receptors used by Henipaviruses29,63. Through BLI and SPR experiments, we assayed the binding of all G head domains in our panel to recombinant Fc-tagged Ephrin-B2 and -B3. As expected, all Nipah and Hendra virus head domains bound to both B2 and B3, with stronger binding observed for B2 rather than B3 (Fig. 7A and Supplementary Fig. 27A, B). CedV-1-G and GhV-G also bound to B2, with very faint binding to B3. Consistent with the previous observations, LayV and MojV G head domains, as well as the rest of the panel, including all Parahenipaviruses, did not show any appreciable binding to either B2 or B3.Fig. 7Structure and characterization of henipavirus G proteins.A Heatmap of recombinant Fc-Ephrin-B2 and -B3 binding to G head domains, on a spectrum of blue for low binding to red for high binding. The graph is based on BLI data with immobilized Ephrin and G head domain analytes. The value for each combination is based on the maximum response value from the association step. G head domain antigens in the table colored by group as in Fig. 1. B Heatmap of antibody binding to G head domains, with color scheme identical to (A). The graph is based on BLI data as described in (A) using immobilized anti-G antibodies. C Sequence alignment of the C-terminal end of select HNV-G proteins. D Three views of the GakV-G head domain colored in rainbow from N- to C-terminal. E Superposition of GakV-G head domain on Ephrin-bound NiV-G head domain (PDB: 2VSM [10.2210/pdb2vsm/pdb]76). GakV-G is colored gray, NiV-G cyan, and Ephrin light blue. Ephrin is shown as a transparent surface. The GakV-G glycan 506 is shown as blue spheres. F Topology diagram of GakV-G protein with blades of the beta propellor colored in rainbow from blade 1 through 6. G Left: Two views of the GakV-G head domain colored in rainbow with the N-terminal Asp to His218 shown as a transparent surface. The additional minidomain defined in the GakV-2-G head structure is highlighted within a dashed oval. Right: G head domains of LayV and NiV. All species abbreviations are described in Supplementary Note 1. Source data are provided as a Source Data file. Structure and characterization of henipavirus G proteins. A Heatmap of recombinant Fc-Ephrin-B2 and -B3 binding to G head domains, on a spectrum of blue for low binding to red for high binding. The graph is based on BLI data with immobilized Ephrin and G head domain analytes. The value for each combination is based on the maximum response value from the association step. G head domain antigens in the table colored by group as in Fig. 1. B Heatmap of antibody binding to G head domains, with color scheme identical to (A). The graph is based on BLI data as described in (A) using immobilized anti-G antibodies. C Sequence alignment of the C-terminal end of select HNV-G proteins. D Three views of the GakV-G head domain colored in rainbow from N- to C-terminal. E Superposition of GakV-G head domain on Ephrin-bound NiV-G head domain (PDB: 2VSM [10.2210/pdb2vsm/pdb]76). GakV-G is colored gray, NiV-G cyan, and Ephrin light blue. Ephrin is shown as a transparent surface. The GakV-G glycan 506 is shown as blue spheres. F Topology diagram of GakV-G protein with blades of the beta propellor colored in rainbow from blade 1 through 6. G Left: Two views of the GakV-G head domain colored in rainbow with the N-terminal Asp to His218 shown as a transparent surface. The additional minidomain defined in the GakV-2-G head structure is highlighted within a dashed oval. Right: G head domains of LayV and NiV. All species abbreviations are described in Supplementary Note 1. Source data are provided as a Source Data file. We assessed the antigenicity of our G head domain panel with a series of previously characterized anti-NiV/HeV-G antibodies: HENV-26 and −3225, m102.3 and .458, nAH1.364, and 1E535. We also included the same Griffithsin construct used with the F panel. The known anti-NiV/HeV antibodies demonstrated broad, strong binding to all tested NiV and HeV strains (Fig. 7B and Supplementary Fig. 27C, D). There were lower levels of binding observed for some of these antibodies outside of NiV/HeV strains, particularly with Ghana virus. Griffithsin showed moderate binding to several strains, albeit inconsistent and not seemingly linked to sequence similarity. It bound most tightly to the SHNV3-G and SHNV11-G head domains. Although it did not show consistent binding to NiV/HeV proteins, as with HNV-F, nevertheless, the ability to bind to several Parahenipavirus G proteins is notable, given the lack of known binders for these strains. Slight variations in binding were observed through ELISA, mostly in the form of low to moderate levels of 1E5 binding to GhV-G, DarV-K-G, and MelV-G (Supplementary Fig. 27D). We used DSF to assess the thermostability of the G head domains. Unlike F proteins, the head domains are not known to undergo a dramatic conformational change when heated, and most proteins in the panel generated a profile with a single peak between 60 and 70 °C, with some small shifts in inflection temperature that were largely consistent within species groupings. (Supplementary Fig. 28A, B). Several species generated profiles distinct from the most common result, such as broad, negative, or double peaks (Supplementary Fig. 28C, D). Unlike the relatively high conservation of tryptophan positions in F proteins, G proteins display a much higher level of sequence diversity (Supplementary Fig. 19). Indeed, there are substantial differences in tryptophan positions in species that generated outlier profiles; CedV, GhV, AngV, and SHNV2 especially. The variability of tryptophans in the G protein across species groupings precludes genus-wide comparisons of DSF profiles, yet these results provide interesting comparative insights within species groupings that may have implications for differences in receptor binding and virus fitness between species. As observed in the case of the SARS-CoV-2 spike protein, where thermostability of the receptor binding domain is a property that is optimized during virus evolution65–67, stability of the G head domain may impact its ability to bind receptor and antibodies, thus impacting its biological function. Structure of Gamak virus attachment protein head domain HNV-G proteins are tetramers (dimer of dimers), with each monomer composed of a helical stalk domain, a neck domain, and a 6-bladed, β-propeller head domain where the receptor-binding site is located22. When comparing the sequence diversity of the HNV-G proteins, we noted that the C-terminal end of G head domains are a site of substantial variability (Fig. 7C and Supplementary Data 3). GakV-G has among the longest C-terminal end, over 30 residues longer than NiV-G. A crystal structure of the GakV-2-G protein head domain at 1.4 Å resolution (Fig. 7D, Supplementary Fig. 29A, and Supplementary Table 4) showed the expected 6-bladed β-propeller architecture that is well-conserved among HNV-G proteins. A well-resolved glycan was observed in the region that is analogous to the Ephrin receptor binding surface in the NiV-G head domain (Fig. 7E). The presence of this glycan further illustrates the changed receptor specificity of the Parahenipaviruses, as it would likely cause a steric hindrance to Ephrin binding if present in NiV or HeV. The elongated C-terminal end was resolved in our structure, revealing a domain spanning residues 606 to 628 that formed a three-stranded β-sheet stacked against blade 5 of the β-propeller (Fig. 7D, F and Supplementary Fig. 29B). This additional compact and ordered domain is distinct from the previously known structures of Henipavirus G protein head domains25,68–71. Furthermore, the N-terminus of the head domain appeared well ordered and packed against the 6-blade β-propeller core (Fig. 7G). Residues 194–199 threaded between blades 5 and 6, contributing a β-strand to blade 6 as it continued to lead into blade 1. This stretch of residues also contacted blades 2, 4, and 5, thus forming a connection between different secondary structures within the β-propeller. Taken together, the structure of the GakV-G head domain illustrates the structural diversity that can be accommodated within the conserved β-propeller fold to possibly impact receptor tropism and antigenicity. Discussion Over the past few years, there has been a rapid expansion of identified species and strains in the Henipavirus genus, ultimately leading to the separation of these species into the Henipavirus and Parahenipavirus genera19. Given the history of zoonotic transmission with these viruses, characterization of these species is an important step for pandemic readiness. Here, we characterize a diversified panel of HNV-F and -G proteins to construct a panel that reflects the global diversity of HNVs. Until now, many of the strains in our expanded panel of HNVs were known by their sequences alone, their surface glycoproteins totally uncharacterized, and their relationships to other HNVs unknown. This study elucidates their genetic diversity, antigenic relationships, and structural and mechanistic features. Despite low sequence identity, HNV-F proteins share many architectural and biophysical similarities, a trend we previously noted by comparing LayV to NiV28. One notable similarity was the oligomerization between F trimers apparent in the SEC profiles and observed in cryo-EM datasets, but not detected by mass photometry or NSEM, where sample concentrations are much lower. Presence of F multimers across diverse HNV-F ectodomains suggests that F oligomerization is likely a conserved trait among HNVs. Hexameric arrangements of AngV-F ectodomain trimers were observed in cryo-EM datasets. Inter-trimer interactions were previously reported for NiV47, yet distinct in the implicated residues, with the AngV-F interaction involving DI and DIII, while the NiV-F interaction is dominated by DII residues. A Hexamer-of-trimers arrangement had been previously described in crosslinked NiV-F ectodomains48. Previous studies have described various lattice arrangements of Paramyxoviridae and Pneumoviridae F proteins. Hexameric lattices were observed in a whole-virion tomography structure of parainfluenza virus 360. Alternate arrangements of fusion and attachment proteins have been observed in measles virus and respiratory syncytial virus46,72. Further studies in the context of pseudovirus or authentic virus would be needed to elucidate the physiological relevance of the inter-F interactions observed in this study, such as the hexameric lattice observed for AngV-F. Assessing HNV-F protein thermostability using a DSF assay provided a rapid and inexpensive tool with which to observe conformational transitions. We identified distinct signatures for pre- to post-fusion conversion and F protein denaturation, revealing how differences in sequence or structure could impact metastability. Despite limitations in biological relevance working with soluble F0 ectodomains, this assay has proven reliable in quickly detecting altered pre-fusion stability in HNV-F proteins. It allowed the identification of the naturally pre-fusion stabilized character of AngV-F, leading to our discovery of the HRB helix proline substitution as a means for pre-fusion stabilization across diverse HNVs. The region surrounding the start of the HRB helix as well as specifically the positively charged residue substituted for a proline are so well-conserved that successful transfer of the stabilizing mutant to other species was possible without further structure-guided decision making, unlike previous attempts where successful transfer of pre-fusion stabilization strategies for NiV-F to Parahenipaviruses have required structure-based evaluation of analogous sites to transfer disulfide, cavity-filling, or proline substitution strategies29,47. While strategies like the triple mutation found in NiVop08 appear to fully lock the F protein in a pre-fusion conformation, the HRB start proline moderately stabilizes the F protein without completely preventing conformational conversion. For constructs like SHNV5-1-F and GakV-1-F, the use of the HRB proline substitution has allowed production of pre-fusion F ectodomains where wild-type constructs do not produce any pre-fusion protein. DSF analysis of G head domains also reveals a diversity of thermostability patterns. Although G head domains are comprised of a single domain 6-stranded β-propeller fold and are not known to have any major internal conformational rearrangements as part of their receptor-binding mechanism, some studies have indicated that receptor binding by HNV-G proteins could effect subtle changes in protein flexibility that do not manifest as large-scale conformational events73. The differences in thermostability could potentially correlate with mechanistic differences between strains; however, more structural data detailing the structural changes that take place after receptor binding in G proteins will be needed to confirm this. Our crystal structure of the Gamak virus head domain suggests additional sources of mechanistic differences between species. HNVs such as NiV, HeV, and CedV target known Ephrin receptors41,74,75, while receptors for Parahenipaviruses remain unknown. The presence of the glycan in the traditional Ephrin-binding site suggests an altered mode of receptor engagement in GakV. While the glycan may not be solely responsible for the lack of Ephrin binding in GakV-G, it would likely hinder binding of a similar protein receptor in the EFNB2/3 site, suggesting that GakV either uses a smaller receptor or perhaps a different receptor binding site entirely. This would not be without precedent in Paramyxoviruses. For example, though EFNB2/3 engages the same face of the head domain in NiV and HeV that sialic acid does in other Paramyxoviruses76,77, the measles virus receptor, CD150, binds at the side surface of the head domain, making major contacts along blades 5 and 678. The C-terminal diversity in the head domains, and especially the additional exterior blade revealed in the GakV-2-G structure, may be a source of mechanistic differences as well. Together with the altered arrangement of the N- and C-terminal ends of the head domain, these features could suggest a different pattern of quaternary interactions in GakV-G ectodomain tetramers, different types of conformational changes within GakV-G after receptor binding, or may even serve as a possible binding site for the as of yet undiscovered receptor. Through immunization in mice, we have elicited Henipavirus and Parahenipavirus F protein targeting antibodies. We have confirmed neutralization activity of three antibodies against Nipah and Hendra pseudoviruses and authentic viruses and discovered additional antibodies that target previously unknown antigens. By assessing the binding of a panel of both previously identified anti-HNV antibodies and antibodies elicited through immunization in this study, we can now better describe the antigenic landscape of HNV proteins. As expected, we saw little ability of anti-NiV/HeV antibodies to bind to non-NiV/HeV proteins. Most interestingly, we found clear antigenic overlap between the F proteins of LayV, MojV, SHNV5, and to a lesser extent, GakV. Both a previously reported anti-MojV-F antibody, 4G5, and our vaccine-elicited antibodies, 22F5 especially, bound to these strains. When considering our phylogenetic analysis, both by full genome and amino acid sequence, it can be seen this cross-reactivity spans distantly related strains. SHNV5 appears to be more distantly related to LayV than HeV is to NiV. GakV is even more distant phylogenetically, appearing to be roughly as distant from LayV as CedV is from NiV. While the G and F proteins are not particularly heavily glycosylated relative to other viral glycoproteins, they ultimately do exhibit a varied glycosylation landscape, with glycans appearing in important antigenic sites in some species. While the lectin griffithsin only showed moderate levels of binding to most antigens tested here, it was able to bind to a very diverse set of HNV-G head domains. Most interestingly, its strongest binding was to the SHNV3 and SHNV11-G head domains, proteins for which there are currently no known antibodies or ligands that can bind them. As investigation of griffithsin for antiviral potency across several families continues, it may prove a useful tool for analysis of diverse HNV proteins. With a group of viruses exhibiting such great sequence diversity, one major concern in the event of a possible widespread outbreak of disease in humans is that emergent infective strains may escape immunity conferred from prior infection or immunization with other HNV strains. This phenomenon can be observed with yearly outbreaks of new influenza strains or with the progression of SARS-CoV-2 evolution, where new strains emerged and quickly dominated the pool of circulating virus as they evaded prior immunity79,80. A key focus of efforts to address this possibility has been the design of immunogens that can elicit antibodies with broad coverage of the currently existing and future diversity of strains. Being able to draw antigenic boundaries within the Henipavirus and Parahenipavirus genera is a foundation for similar efforts, and here we see that while the prospect of cross-reactivity over the entire genus remains elusive, there is already demonstrated cross-reactivity among a large portion of the Parahenipaviruses. While it will be important to assess this observed cross-reactivity in neutralization assays and animal challenge models, there is currently limited knowledge of both the host receptors for Parahenipaviruses and of the proper cell lines or animal models to study infection. While some progress has been made in identifying permissive cell lines for Parahenipaviruses63,68, the rate of fusion is typically much lower than seen with more well-studied Nipah/Hendra models, indicating that such viral entry assays could be in need of further optimization, therefore, at this time, it remains impractical to perform similar assays with Parahenipavirus species. In the absence of robust experimental systems to study the virological properties and the breadth of protection from potential HNV infections, the discoveries reported here build a critical foundation for future exploration of Parahenipavirus antigenicity. The broad characterization of HNV proteins in this study provides a set of essential biophysical and biochemical data for many species and strains that, to this point, had only been known as a sequence in a database. As these species are studied in more depth, our analysis can be used to provide context for the comparison of other key aspects of Henipavirus biochemistry, structure, and immunology. Methods Ethics All mice were cared for in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. The study protocol, A081-24-04, and all veterinarian procedures were approved by the Duke University Institutional Animal Care and Use Committee (IACUC). All rooms are on a 12/12 light cycle unless otherwise requested. Heat and humidity are maintained within the parameters outlined in The Guide for the Care and Use of Laboratory Animals and animals are fed a standard rodent diet. Breeding or specialty diets will be provided upon request. Rodents and Guinea pigs are housed in individually ventilated micro-isolator caging on corn cob bedding. DLAR will provide alternative bedding based on scientific need. Mice are changed every two weeks, rats are changed twice a week, and Guinea pigs are changed three times a week or more frequently as needed. Group housing of up to five mice per cage is strongly encouraged. Environmental enrichment for singly housed animals will be provided unless an exception is granted by the IACUC. Phylogenetic analysis Candidate genome sequences were identified by using the NCBI Protein BLAST tool30. Nipah-Malaysia sequence NC_002728 was used as an input sequence. Amino acid sequences were aligned using the MAFFT algorithm81. We removed any sequences for which there were gaps in the amino acid sequence, ensuring each entry contained a full ectodomain protein sequence. The final tree was built as a UPGMA based on the distance matrix using the Tree Viewer software (v2.2.0)82. Protein production For HNV proteins, human codon-optimized DNA sequences were synthesized by GeneImmune Biotechnology and cloned into a pαH plasmid vector83. Fusion protein ectodomain constructs encoded residues 1 to the equivalent of 488 in NiV, followed by a foldon trimerization domain84, HRV3C protease site, 8× His tag, and a C-terminal Twin-Strep tag (IBA Lifesciences) (Supplementary Data 1 and 2). For all constructs, the putative natural cleavage site that allows processing from the F0 precursor to the fusion-competent F1/F2 complex was retained. Attachment protein head domains (Supplementary Data 1 and 3) had an artificial secretion signal, secrecon62, added to the N-terminal end and an 8× His tag added to the C-terminal end. HNV protein plasmids were transiently transfected into HEK 293 cell lines, either Expi293 (Thermo Fisher A14527) or 293F (Thermo Fisher R79007) (Supplementary Data 1). The cells were allowed to incubate at 37 °C, 8% CO2 and the supernatant was harvested 5 days post-transfection. Fusion proteins were purified by first sequestering free biotin by applying BioLock blocking solution (IBA) and then using the Strep-Tactin affinity chromatography resin (IBA). Eluted protein was then further purified using a Superose 6 Increase 10/300 GL size-exclusion column (Cytiva) with PBS buffer at pH 8. The elution from SEC was spin concentrated with 100 kDa filters and flash frozen with liquid nitrogen. Attachment protein head domains were purified first by sequestering EDTA by adding 2 mg of Cobalt Chloride per 1 L of supernatant. TALON Cobalt Resin (Takara) was used for purification, with 20 mM HEPES and 150 mM NaCl at pH 8 as a buffer, adding 50 mM Imidazole for elution. Proteins were further purified with a Superdex 75 Increase 10/300 GL SEC column (Cytiva), again using the HEPES-NaCl buffer. The elution from SEC was spin concentrated with 10 kDa filters and flash frozen with liquid nitrogen. Antibody human codon-optimized DNA sequences were synthesized by GeneImmune Biotechnology and cloned into pVRC8400 plasmid vectors. Heavy chains were designed to begin with the beginning of the antibody sequence and continue through the human IgG1 Fc region. Light chains included both the VL and CL domains. Depending on the sequence source, some antibody sequences either utilized mouse or human VL and VH domains. Mouse anti-NiV-F Fabs were designed to end after the mouse CH1 domain and had an 8× His tag added to the C-terminal end, consistent with their production when originally described51. Plasmids encoding the heavy and light chains of the antibodies IgG 1H1, 4B8, 1E5, nAH1.3, 4G5, HENV-26, HENV-32, m102.3, and m102.4, or an Fc-tagged Griffithsin construct, were co-transfected into Expi293 cells at a 1:1 ratio, using the ExpiFectamine 293 Transfection Kit (Thermo Fisher A14525), according to the manufacturer’s instructions. Following a six-day incubation period post-transfection, the cells were harvested, and the culture supernatant was clarified by filtration through a 0.2 µm PES membrane filter to remove cellular debris. IgG purification was subsequently carried out using Protein A affinity chromatography. The filtered supernatant was passed through the Protein A affinity column (PierceTM Recombinant Protein A Agarose 20334), and the bound IgG was eluted using IgG elution buffer by adjusting pH via Tris Base. Eluted fractions were pooled and concentrated with Centricon-70 30 kDA filters (Millipore Sigma), followed by size-exclusion chromatography using a HiLoad 16/600 Superdex 200 pg column (Cytiva), pre-equilibrated with 1× phosphate-buffered saline (PBS, pH 7.4) containing 0.02% sodium azide to prevent microbial contamination. The purified IgG was collected, concentrated, and analyzed for purity and integrity by SDS-PAGE. Anti-NiV-F Fabs 1A9, 1F2, 1F3, 1H1, 1H8, 2D3, 4B8, 4F6, and 4H3 were purified using the same Cobalt resin-based procedure as described above for head domains. To generate Fab fragments for 22F5 mouse IgG and 1C8 mouse IgG, 2 mg of each IgG was passed through Zeba Spin Desalting Columns (ThermoFisher), then washed with digestion buffer containing 35 mg cysteine. IgGs were incubated with 0.25 mL of the 50% slurry of Immobilized Papain resin (ThermoFisher), followed by incubation with digestion buffer containing 35 mg cysteine at 37 °C overnight in microcentrifuge tubes. The digested Fabs were separated from the Immobilized Papain by centrifuging at 5000 × g for 1 min, followed by washing with digestion buffer. After digestion, the mixture was subjected to Protein A affinity chromatography (ThermoFisher), then monitored by SDS-PAGE to confirm Fab generation. To further purify the Fab fragments, SEC was used for purification of the Fab fragments using Superose 6 Increase 10/300 column GL (Cytiva). The sample was loaded onto the column at a volume of 0.5 mL and the elution was carried out with PBS at a flow rate of 0.5 mL/min. Fractions were collected, yielding a final concentration of 3.2 mg/mL and 4.2 mg/mL for 22F5 Fab and 1C8 Fab, respectively. The purified Fab was stored at −80 °C. The influenza antibody Ab82 (also referred to as CH65)85 was commercially obtained (Catalent Cat_Ab82_AAA). MBP-1F524 and VRC0186 were expressed by co-transfecting heavy and light chain plasmids into Expi293 (Thermo Fisher A14527) cells using the ExpiFectamine 293 Transfection Kit (Thermo Fisher A14525). For small-scale expression, 15 μg each of heavy and light chain plasmid DNA were combined in 1.5 mL of Opti-MEM medium and then mixed with transfection medium containing 80 μL of ExpiFectamine 293 transfection reagent in 1.5 mL of Opti-MEM medium. The complex of DNAs and ExpiFectamine 293 transfection reagent was incubated for 20–30 min at room temperature before being mixed with 30 mL of Expi293F cell culture (2.5 × 106 cells/mL) in a 50 mL bio-reaction tube (CELLTREAT Scientific). The transfected cell culture was returned to suspension incubation for 24 h at 37 °C, 8% CO2, and 250 rpm and then fed with 150 μL ExpiFectamine 293 Transfection Enhancer 1 and 1.5 mL of ExpiFectamine 293 Transfection Enhancer 2. Four to six days post-transfection, the supernatant was harvested by centrifugation and flowed over a column packed with 1 mL rProtein A Sepharose Fast Flow antibody purification resin (Sigma Aldrich). The column was washed twice with IgG Binding Buffer (Pierce), and the IgG was eluted with IgG Elution Buffer (Pierce). The elution fraction was then neutralized with Tris pH 9.0 to a final pH of 7.5. Purified antibodies were dialyzed against 1 × phosphate-buffered saline (PBS), diluted to a final concentration of 1 mg/mL using a 30 kDa molecular weight cutoff Amicon Ultra spin concentrator (Millipore), and aliquoted and stored at −80 °C. DSF Samples were analyzed using a Prometheus Panta (Nanotemper Technologies, v1.8). All samples were diluted to 0.125 mg/mL using PBS, pH 7.4. Roughly 10 µL was added to glass capillaries in triplicate for measurements. The instrument was programmed to measure 350 and 330 nm fluorescence from 35.0 to 95.0 °C using a 6.0 °C/min temperature gradient. All curves reported represent averages from the triplicate readings. BLI BLI was employed to assess the binding between HNV proteins and various antibodies. For fusion proteins (Fig. 2), the analysis was performed using an Octet RH16 (Octet RED384) system (Sartorius) at 25 °C with kinetics buffer (Sartorius) as the running buffer. HNV-F proteins with the Twin-Strep tag at 6 µg/mL were captured onto Streptavidin (SA) Biosensors (Sartorius) for 300 s, followed by a 60 s wash in kinetics buffer. The biosensors were then exposed to 20 µg/mL mAbs or 50 µg/mL Fabs for 300 s to allow for association with the fusion protein. Dissociation in kinetics buffer was monitored for an additional 300 s. Sensor data was analyzed using Octet Analysis Studio (Sartorius, v12.2.2.26), with reference biosensor and zero-analyte control sensorgrams subtracted from the raw data to obtain the curves corresponding to specific binding events. For additional fusion protein experiments (Supplementary Fig. 10C), the analysis was performed with the same equipment, tips, and buffer. HNV-F proteins were loaded at 6 µg/mL for 300 s and equilibrated in buffer for 60 s before a 120 s association step and 200 s dissociation step. All analytes were at 50 µg/mL. Regeneration with 10 mM Glycine, pH 1.7, was employed, using tips for 6 cycles and regenerating with 10 s pulses between regeneration solution and buffer. SARS-S2-HexaPro87 was loaded onto a tip as a reference sensor. Kinetics buffer was used as a reference sample. Double reference subtraction was performed, except for in the case of Griffithsin IgG, which bound to the SARS-S2-HexaPro, so only reference sample subtraction was used. For each cycle, NiVop08 was loaded and bound to 4B8 Fab, and the signal relative to the first use of the tip was used to adjust the response values of all association curves during that cycle. For attachment head domains with human Fc-containing Abs, the analysis was performed using an Octet RH16 (Octet RED384) system (Sartorius) at 30 °C with HBS-EP+ Buffer (Cytiva) as the running buffer. Antibodies at 2 µg/mL were loaded onto anti-human antibody capture (AHC) biosensors (Sartorius) for 300 s, followed by a 60 s buffer baseline. Association with head domains at 25 µg/mL lasted 120 s, followed by dissociation in buffer for 200 s. Between cycles, biosensors were regenerated by alternating 3× in 10 mM Glycine pH 1.0 buffer and running buffer for 5 s each. Sensors were used for no more than 10 cycles of regeneration. Double reference subtraction was performed using both a control-loaded sensor and a zero-analyte reference sample. SPR For 22F5-LayV-F interactions, binding and kinetic assays were performed on a T-200 Biacore system (Cytiva) at 25 °C. All the binding and kinetic experiments were carried out in the HBS-EP+ (10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, and 0.05% surfactant P-20) running buffer. For the binding assay, an anti-Fc mouse SPR chip was used to capture 200 nM of 22F5 IgG at 10 µL/min flow rate and 50 nM of LayV-F ectodomains as analyte were run at 30 µL/min flow rate and association and dissociation were monitored at 120 s. In the competitive assay, 200 nM of 4G5 IgG was captured on an anti-Fc human SPR chip at 10 µL/min and analyte as 22F5 Fab with LayV-F ectodomain complex in 5:1 ratio was flowed at 30 µL/min flow rate for 60 s. SPR SA chip was used to capture the 50 nM of the LayV-F ectodomains at 5 µL/min flow rate by biotin tag for kinetics analysis. 22F5 Fab 50 nM with twofold dilution (50, 25, 12.5, 6.25, and 3.12 nM) as analyte was flowed with 50 µL/min flow rate and association and dissociation times were analyzed at 60 and 240 s, respectively. A flow cell without antibody capture was used as reference. The data were processed with Biacore T200 Evaluation Software version 3.1.1 (Cytiva) and double-referenced by subtracting the blank surface and a blank injection (buffer with analyte). Binding curves were globally fit to a 1:1 Langmuir binding model. For Ephrin-HNV-G interactions, binding experiments were performed on a Biacore T-200 (Cytiva) with HBS buffer supplemented with 3 mM EDTA and 0.05% surfactant P-20 (HBS-EP+, Cytiva, MA). All binding assays were performed at 25 °C. Ephrin binding to head domains was assessed using a Series S CM5 chip (Cytiva, MA), which was labeled with anti-human IgG (Fc) antibody using a Human Antibody Capture Kit (Cytiva, MA). Ephrin was coated onto the chip at 12.5 nM (120 s at 5 μL/min). Head domains were injected at 100 nM over the ephrin using a single injection (60 s on time, 180 s off time at 50 µL/min). The surface was regenerated with 3 pulses of a 3 M MgCl2 solution for 10 s at 100 μL/min. A flow cell without Ephrin-Fc capture was used as reference. The data were processed with Biacore T200 Evaluation Software version 3.1.1 (Cytiva) and double-referenced by subtracting the blank surface and a blank injection (buffer with analyte). NSEM Frozen samples from −80 °C were thawed at room temperature for 5 min. Samples were then diluted to 40 µg/mL with 0.02 g/dL Ruthenium Red in HBS (20 mM HEPES, 150 mM NaCl, pH 7.4) buffer containing 8 mM glutaraldehyde. After 5 min incubation, glutaraldehyde was quenched by adding sufficient 1 M Tris stock, pH 7.4, to give 80 mM final Tris concentration and incubated for 5 min. Quenched sample was applied to a glow-discharged carbon-coated EM grid for 10–12 s, blotted, consecutively rinsed with 2 drops of 1/20× HBS, and stained with 2 g/dL uranyl formate for 1 min, blotted, and air-dried. Grids were examined on a Philips EM420 electron microscope operating at 120 kV and a nominal magnification of 49,000×, and images were collected on a 76 Mpix CCD camera at 2.4 Å/pixel. Images were analyzed by 2D class averages using standard protocols with Relion 3.088. Mass photometry A TwoMP (Refeyn) mass photometer was used for analysis. The AcquireMP (Refeyn, v2024 software was used for calibration and data acquisition. Calibration was performed with a three-point curve consisting of β-amylase, Apoferritin, and Thyroglobulin. Samples were diluted to 40 nM in PBS. To each well, 10 µL of PBS was added to focus the system before adding 10 µL of sample and mixing for a final concentration of 20 nM. Recorded data for 60 s, with the machine counting individual adsorption events. The DiscoverMP processing software (Refeyn, v2024 R2.1) was used to assign molecular weights to individual events. Mouse immunizations Immunization in VH1-2RJH2/Vκ1-33RCS∆/hTdT (SE13) mice89 were intramuscularly immunized with HNV-F protein (25 µg/animal) adjuvanted with GLA-SE (IDRI EM-082, 5 µg/animal) at weeks 0, 4, 15, and 21 (first study) or 0, 4, 8, and 12 (second study). Serum titers were monitored by ELISA as described below. Mice with high-binding antibody titers were selected for the subsequent spleen cell fusion experiments. All mice were based on the 129S6/SvEvTac (Taconic Biosciences INC, 24028) and C57BL/6 (Jackson Laboratory 000664) strains. For the first study, each group had one male and one female mouse, all aged 15 weeks at first immunization. In the second study, groups 1 and 2 had 3 male and 3 female mice and group 3 had 2 male and 4 female mice, all aged 10–12 weeks at first immunization. Sex was not considered for study design as the goal was to generate antibodies without regard for sex-specific immune response. Hybridoma cell line generation and monoclonal antibody production Mice were boosted with the indicated priming antigen 3 days prior to fusion. Spleen cells were harvested and fused with NS0 murine myeloma cells using PEG1500 to generate hybridomas50,90. After 2 weeks, supernatant of hybridoma clones were collected and screened by binding ELISA as described below. Hybridomas that secreted LayV-F reactive antibodies were cloned by limiting dilution until the phenotypes of all limiting dilution wells were identical. IgG mAbs were purified by protein G purification. Indirect binding ELISA method ELISA assays were utilized for three purposes: measuring mouse serum antibody titers, screening for antigen-specific hybridoma B cell clones, and characterizing the binding of monoclonal human IgG and mouse IgG. 384-well ELISA plates (Costar #3700) were coated with 2 µg/mL streptavidin (Thermo Fisher Scientific Inc. Cat. No. S-888) or protein antigen in 0.1 M sodium bicarbonate overnight at 4 °C. Plates were washed with PBS/0.1% Tween-20 and blocked for one hour with assay diluent (PBS containing 4% (w/v) whey protein/15% Normal Goat Serum/0.5% Tween-20/0.05% Sodium Azide). Streptavidin coated plates were washed and followed by 10 μL Twin-Strep-tag® (IBA Lifesciences GmbH) protein at 2 µg/mL in assay diluent for one hour. All plates were then washed and samples added in 10 μl volumes with dilution scheme and secondary specifics depending on sample type as follows. Mouse serum was titrated from 1 to 30 in threefold steps and incubated for 1 h. Plates were washed and followed with 10 µL goat anti-mouse IgG-HRP secondary antibody diluted 1:10,000 (Southern Biotech #1030-05). Hybridoma supernatants were added undiluted as a single well for 1 h, washed, and followed with 10 µL goat anti-mouse IgG-HRP. Monoclonal mouse IgG and Human IgG were diluted to 100 µg/mL, titrated in threefold steps and incubated for 1 h. Plates were washed and followed with 10 µL goat anti-Mouse IgG Fab-HRP diluted 1:10,000 (Southern Biotech 1015-05) for mouse IgGs or 10 µL of goat anti-human IgG Fab-HRP diluted 1:15,000 (Jackson Immunoresearch, 109-035-097) for human IgG monoclonal mAbs. After 1 h, plates were washed again and detected with 20 µL SureBlue Reserve (Seracare 5120-0081) for 15 min. Reactions were stopped with the addition of 20 µL HCl stop solution. Plates were read at 450 nm using a Spectramax 384 plus. Antibody gene cloning and sequencing Cell suspensions of the 22F5 hybridoma cell line were pelleted by centrifugation at 300 × g for 5 min. The cell pellets were washed twice with Phosphate-buffered saline (PBS, 1×) to remove residual culture medium. The 22F5 antibody genes were amplified from the hybridoma cells by RT-PCR. Briefly, RNA was isolated from the cell pellets by adding them to reverse transcription solution. Immunoglobulin genes were reverse-transcribed using Superscript III (Thermo Fisher Scientific) with random hexamer oligonucleotides (Gene Link, Hawthorne, NY) as primers. The cDNA served as a template for nested PCR amplification of the antibody heavy and light chain genes. The PCR used mouse antibody gene-specific primers91 and was carried out with AmpliTaq Gold 360 (Thermo Fisher Scientific). The PCR products were purified using the Biomek FX Laboratory Automation Workstation (Beckman Coulter, Brea, CA). Purified PCR products were then sequenced using Sanger sequencing to obtain the antibody gene sequences. V, D, J gene segment usage, somatic hypermutation frequency, and complementarity-determining region 3 length of the VH (heavy chain variable region) and VK/L (light chain variable region) genes were determined using Partis92 (https://github.com/psathyrella/partis) with a custom knock-in mouse reference that combined the knocked-in human VH1-2*02, JH2*01, and VK1-33*01 with mouse endogenous V, D, and J gene segments. Cryo-EM grid preparation and imaging To prepare grids, Quantifoil R1.2/1.3 (Cu, 300-mesh; Electron Microscopy Sciences, PA) grids were glow discharged for 15 s at 15 mA using a PELCO easiGlow Cleaning System (Ted Pella Inc.). The final concentration of purified HNV-F ectodomains was maintained as 1.5 mg/mL in PBS, with a 5× molar ratio of antibody to HNV-F ectodomain, excepting a 4.87× ratio for LayV-F-DS and 22F5. 0.005% (w/v) of n-dodecyl β-D-maltoside (DDM) was added to the AngV-F samples and 0.005% (w/v) DDM and 2% glycerol to all other samples before applying to the grid to prevent the protein from adhering to the air-water interface. 3.0 μL of the protein sample was applied to the grid and incubated for 30 s at >95% humidity. Excess protein was blotted away for 2.5 s before being plunge frozen into liquid ethane using Leica EM GP2 plunge freezer (Leica Microsystems). Cryo-grids were imaged using either an FEI Titan Krios (Thermo Scientific) or Tundra cryo TEM (Thermo Scientific). The Titan Krios TEM was used for the determination of the AngV-F trimer, dimer-of-trimer, and 22F5 Fab-LayV-C1-F complex structures. The Krios was equipped with a K3 camera (Gatan) at 81 k magnification, operated at 300 keV. For AngV-F, approximately 24,000 micrographs were collected at nominal defocused range between −2.8 and −1.7 μm with a dose range of 63–65 e/A2. For LayV-C1-F, 9715 micrographs were collected at a nominal defocused range between −2.8 and −1.2 µm with a dose range of 58.6–60.6 e/A2. The Tundra TEM was used for determination of the AngV-F hexameric lattice and 22F5, 20G7, 8C7, and 9A9-containing complex structures. For the AngV-F hexameric lattice, the Tundra was equipped with a CETA-F camera (Thermo Scientific) at 180 k magnification, operated at 100 keV. Approximately 4000 micrographs were collected at nominal defocused range between −3.0 and −1.5 μm with a dose range of 23–25 e/A2. For the antibody complexes, the Tundra was equipped with a Falcon 4i camera (Thermo Scientific) at 180 k magnification, operated at 100 keV. A range of micrographs, doses, and defocus ranges were used, as shown in Supplementary Table 3. Cryo-EM data processing Cryo-EM image quality was monitored on-the-fly during data collection using automated processing routines. Data processing was carried out using cryoSPARC (v4.6.2)93,94. Micrographs were curated through contrast transfer function, where greater than >30 Å were discarded. Automated blob picker was used to assign the particle position. Different box sizes are utilized to accommodate the varying dimensions of specific molecular assemblies. For instance, monomer particles are extracted using a box size of 320 pixels, while dimer particles require a larger box size of 512 pixels. For even larger molecular assemblies, such as hexamer lattices, a box size of 1024 pixels was employed. Following particle extraction, multiple rounds of 2D classification was performed to remove junk. ab-initio 3D reconstruction was used to create 3D reconstructions and poor-quality particles were discarded after thorough heterogeneous refinement. Final resulting volumes were subjected to non-uniform refinement to build a high-resolution 3D reconstruction. Phenix (v1.21-5286)95, Coot (v0.9.8.95)96, Pymol (v2.5.5), nextPYP97, and ChimeraX (v1.6.1)98 were used for model building and refinement. Models 8TVE29 and 8U1R47 were used as starting models for post-fusion and pre-fusion LayV-F model building, respectively. X-ray crystallography Crystals were grown via the vapor diffusion method at room temperature in a siting drop well format. GakV-2-G head domain protein crystals were formed by mixing 0.20 μL of protein at 15 mg/mL with 0.20 μL of well solution containing 0.2 M magnesium sulfate heptahydrate, 20% w/v PEG3350 at pH 7.1. Crystals appeared within one week. Cryoprotection of the crystals was done by mixing glycerol to a final concentration of 50% with the well solution. Crystals were moved to the drop containing the well solution and glycerol and soaked for one minute before flash-cooling in liquid nitrogen, with data were collected at the NSLS-II, Beamline 19-ID at 100 K and a wavelength of 1 Å. All data were indexed and integrated using iMosflm99 and scaled using AIMLESS. Phaser100 in the PHENIX (v1.21rc-5127-000) suite was used to perform molecular replacement using a model generated from the AlphaFold3 prediction server101 using the GakV-2-G head domain sequence. Phenix refine102 was used for data refinement, and manual refinement was done in Coot (v0.9.8.92)96. Pseudovirus neutralization Human embryonic kidney cells (293 T, ATCC, CRL-3216) were maintained in complete DMEM (cDMEM) (i.e., DMEM supplemented with 1% (v/v) penicillin-streptomycin, 10% (v/v) FBS) and 2 mM L-glutamine) at 37 °C, with 5% CO2. NiV F and G pseudo-typed viruses were prepared by co-transfecting exponentially dividing 293T (293T cells seeded in T225 flask 72 h prior to transfection) with a 6:6:1:1 ratio of a lentiviral packaging plasmid pCMVdR8.2 (VRC5602)103, transducing plasmid pHR’ CMV-Luc (VRC01)104, NiV F plasmid and NiV G with cytoplasmic tail truncation plasmid105 using Lipofectamine3000 reagent (ThermoFisher) as described by the manufacturer. The transfection complexes were allowed to form for 20 min at room temperature (18–25 °C) before adding to 293T cells, and incubated for 48 h with a change of medium after 24 h. The cell supernatant containing NiV F and G pseudo-typed virus particles was removed from the flasks, filtered through a 0.45 μm filter, and frozen in single-use aliquots at −80 °C. Neutralization was measured by reduction in Luc reporter gene expression following single round of virus infection in 293T cells. Briefly, NiV F and G pseudo-typed virus (titrated to achieve RLU of ~50,000) was incubated with serial dilutions of antibody (nine fivefold dilutions starting at 50 μg/mL) in duplicate for 30 min at 37 °C in a total volume of 20 μL cDMEM in 384-well flat-bottom black, optically clear culture plates (Revvity) with white BackSeal backing tape (Revvity). Freshly trypsinized cells (~ 2600 cells in 20 μL of cDMEM) were added to each well and luminescence measured with a SpectraMax Paradigm luminometer. The 50% inhibitory concentration (IC50) was defined as the antibody concentration that caused a 50% reduction in RLU compared to virus control wells after subtraction of background RLU. Percent neutralization and neutralization ID50s were calculated using GraphPad Prism 8.0.2. Antibodies 1H1 and 4B851, m102.458, 5B327, and MBP-1F524 were included as positive controls and VRC0186 as a negative control. Plaque reduction neutralization test Assays were based on an established protocol106. Antibodies were serially diluted in 1:2 folds for ten dilutions in Dulbecco’s Phosphorous buffered saline (Gibco, NY, USA). 1000 PFU/mL each of NiV and HeV were diluted in DMEM (Gibco) with 2% (FBS). Each dilution of antibodies was incubated at 37 °C and 5% CO2 for 1 h with equal volume of 1000 PFU/mL of prepared NiVB/HeV). 200 µL of each dilution of incubated antibodies with NiVB or HeV virus along with virus or PBS only controls were added to confluent monolayers of NR Vero E6 cells in a 6 well plate (8 × 105 cells/well) in triplicates and plates were incubated for 1 h at 37 °C and 5% CO2. Plates were gently rocked frequently to prevent drying of monolayers. An overlay of 2.5% Avicel® RC-591 microcrystalline cellulose and Carboxymethyl Cellulose Sodium (DuPont Nutrition & Biosciences) was mixed in 1:1 ratio with 2× Modified Eagle Medium (Temin’s modification, Gibco) and 10% FBS (Gibco). Plates were incubated for 4 days at 37 °C and 5% CO2. After incubation period, the monolayers were fixed with 10% neutral buffered Formalin for 6 h and then were stained with 0.2% aqueous Gentian Violet (RICCA Chemicals), followed by rinsing. Plates were dried and plaques were counted manually and plots were prepared using GraphPad Prism 10. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Source data Source Data Source Data