Prefusion-stabilized herpesvirus glycoprotein b trimers

Abstract

Herpesviridae glycoprotein B (gB) polypeptides comprising a modified ectodomain is stabilized in the prefusion state, enabling development of inhibitors and vaccines directed against these viral pathogens. Modifications to DI, DII, and DV subdomains are important to stabilization of gB in the prefusion state, and additional modifications result in a further improved stabilization of gB in the prefusion state. The gB can be derived from an Epstein Barr Virus (EBV), a Human Cytomegalovirus (CMV), a Human Herpesvirus 6 (HHV6), a Herpes Simplex Virus 1 (HSV1), or a Varicella Zoster Virus (VZV). Polypeptides, nucleic acid constructs, compositions, and methods of using same to elicit an immune response are described.

Description

PREFUSION-STABILIZED HERPESVIRUS GLYCOPROTEIN B TRIMERS

[0001] This application claims benefit of United States provisional patent application number 63 / 664,923, filed June 27, 2024, and United States provisional patent application number 63 / 693,647, filed September 11 , 2024, the entire contents of each of which are incorporated by reference into this application.REFERENCE TO A SEQUENCE LISTING

[0002] The content of the XML file of the sequence listing named “UW086_seq”, which is 79 kb in size, created on June 25, 2025, and electronically submitted herewith the application, is incorporated herein by reference in its entirety.ACKNOWLEDGEMENT OF GOVERNMENT SUPPORT

[0003] This invention was made with government support under 75N99224R00001 awarded by Advanced Research Projects Agency for Health APECx, and DP1AI158186, 75N93022C00036 awarded by National Institute of Allergy and Infectious Diseases. The government has certain rights in the invention.BACKGROUND

[0004] The ubiquitous human-infecting members of the Herpesvirus family are responsible for considerable morbidity and mortality globally. Epstein Barr Virus (EBV) - a gamma-Herpesvirus - causes 1 .3-1 .9 % of all human cancers, and has been associated with the development of multiple sclerosis. Infection of the fetus by Cytomegalovirus (CMV), a beta-Herpesvirus, is a leading cause of hearing loss and permanent disability. Infection of older adults with alphaherpesviruses, such as Herpes Simplex Viruses (HSV1 / 2) or Varicella Zoster Virus (VZV), is involved in the etiology of dementia. With respective seroprevalences of >90%, 45-100%, -50%, -10%, and 98% for EBV, CMV, HSV1 , HSV2, and VZV, the development of effective countermeasures are urgently needed.

[0005] Although vaccine trials for most Herpesviruses have not yet achieved their goals, there are effective vaccines against VZV disease. Live-attenuated vaccines against VZV reduce acute disease (Chickenpox) by 87% and reactivation (Shingles) by 51 %; consequently, the incidence of acute VZV symptoms have dramatically decreased following vaccination of children15. More recently, a subunit vaccine against VZV proved effective at reducing VZV reactivation by -90%. This subunit vaccine functions by eliciting neutralizing antibodies and cell-mediated immunityagainst glycoprotein E (gE) which is involved in viral entry. gE is not conserved or is absent in other Herpesviruses, so alternative entry glycoproteins have been explored as immunogens. Most notably, neutralizing antibodies have been identified that are protective in animal models against the conserved Herpesvirus glycoprotein H and L heterodimers (gH / gL) and glycoprotein B (gB) which play a central role in viral entry into cells. Herpesvirus entry is typically initiated by binding of gH / gL to the host receptor, either directly or along with other virus-specific glycoproteins. Upon receptor engagement, gH / gL triggers gB to change its conformation and promote fusion of the viral and host membranes. While several gH / gL-targeted neutralizing antibodies compete for binding with host receptors, the mechanisms of action of most gB- directed neutralizing antibodies has remained elusive. The gB ectodomain is composed of five subdomains, designated DI, Dll, Dill, DIV, and DV, followed by C-terminal transmembrane and cytoplasmic domains. CMV and HSV1 gB share four antigenic sites which are targeted by neutralizing antibodies, located in DI, Dll, DIV, and the flexible N-terminus. Recently, the DIV- targeted 93k VZV neutralizing antibody was proposed to be incompatible with a DIV-bound conformation of the gB N-terminus; though, the underlying molecular mechanism of neutralization remains unknown. Our limited understanding of gB-targeted neutralizing antibody mechanisms highlights gaps in our understanding of Herpesvirus fusion.

[0006] Stabilizing the prefusion conformation of gB can enhance our understanding of fusion and facilitate the development of countermeasures against this important family of pathogens. Although the conformation of viral fusion glycoproteins is finely tuned to promote entry with exquisite spatial and temporal control, irreversible refolding to the postfusion state often occurs upon purification. As a result, stabilizing the prefusion conformation is typically necessary to produce an immunogen eliciting potent immune responses. For example, the 2P and DS-CaV1 prefusion-stabilizing mutations enhance glycoprotein stability and enable elicitation of potent immune responses for approved vaccines against COVID-19 and RSV, respectively. Whereas SARS-CoV-2 and RSV use class I fusion proteins, Herpesvirus gB is a class III fusion protein, which is markedly distinct and much less well characterized with limited prefusion-stabilization success despite extensive prior attempts. The H516P HSV1 gB mutation was shown to increase the prefusion to postfusion ratio; however, the applicability of this strategy is limited to full-length membrane-embedded protein and fails to stabilize the prefusion recombinant ectodomain trimer adequately.

[0007] There remains a need for a stabilized prefusion Herpesvirus gB ectodomain that can be utilized as an immunogen.SUMMARY

[0008] The constructs, compositions, and methods described herein meet this need and more. Described herein are Herpesviridae glycoprotein B (gB) polypeptides comprising a modified ectodomain, and which is stabilized in the prefusion state. In some embodiments, the ectodomain comprises at least one of the following modifications. In some embodiments, the gB ectodomain comprises deletion of residues 402-452 or 394-452 of SEQ ID NO: 1. In some embodiments in residues 402-452 or 394-452 have been deleted, and a GSPPGSPP (SEQ ID NO: 51) linker is inserted in its place. In some embodiments, the gB ectodomain comprises a disulfide bond between Q527C-E634C. In some embodiments, the gB ectodomain comprises an A293P mutation, a oP3H trimer, and / or a I5350A modified to harbor an extendable and structured link to DI (l5350A2h or l5350A2hN). In some embodiments, the transmembrane and / or cytoplasmic domains have been deleted. In some embodiments, gB is derived from an Epstein Barr Virus (EBV), a Human Cytomegalovirus (CMV), a Human Herpesvirus 6 (HHV6), a Herpes Simplex Virus 1 (HSV1), or a Varicella Zoster Virus (VZV). In some embodiments, the gB polypeptide further comprises a mutation of the fusion loop. In some embodiments, the mutation of the fusion loop is one of the fusion loop mutations indicated in Table 1. In some embodiments, the gB polypeptide comprises gB-deliDoP as described in Table 1. In some embodiments, the gB polypeptide comprises gB-deliDoP-l5350A2h or gB-deliDoP-l5350A2hN as described in Table 1 or in FIG. 7.

[0009] SEQ ID NO: 1 provides the unmodified amino acid sequence of gB of EBV. Corresponding residues of CMV, HHV6B, HSV1 , and VZV can be identified in the alignment shown in FIG. 6.

[0010] In some embodiments, the gB polypeptide comprises an amino acid sequence having at least 80% sequence identity to an amino acid sequence of SEQ ID NO: 25-43, or of Table 1 (SEQ ID NO: 6-24). In some embodiments, the sequence identity is at least 85%. In some embodiments, the sequence identity is at least 90%. In some embodiments, the sequence identity is at least 95%. In some embodiments, the gB polypeptide comprises an amino acid sequence of SEQ ID NO: 25-43, or of Table 1 (SEQ ID NO: 6-24). In some embodiments, the gB polypeptide comprises EBV gB-deliDoP-l5350A2h (SEQ ID NO: 12) or EBV gB-deliDoP- l5350A2hN (SEQ ID NO: 15). In some embodiments, the gB polypeptide is in soluble form.

[0011] Also described herein is a nucleic acid molecule that encodes the gB polypeptide. In some embodiments, the nucleic acid molecule is an mRNA or DNA. In some embodiments, thenucleic acid molecule comprises a modified nucleotide. In some embodiments, the modified nucleotide comprises a methylpseudouridine.

[0012] Also described herein is a composition comprising the gB polypeptide or the nucleic acid molecule encoding the gB polypeptide. Such compositions optionally include a pharmaceutically acceptable carrier, excipient, and / or adjuvant.

[0013] Methods of making and using such polypeptides, nucleic acid molecules, and compositions are also described. In some embodiments, the method of use is for eliciting an immune response to a Herpesviridae family virus in a mammalian subject. The method comprises administering a composition as described herein to the subject. In some embodiments, the amount to be administered to the subject is a therapeutically effective amount.

[0014] In some embodiments, the composition comprises additional peptides or nucleic acids encoding additional peptides. In some embodiments, the additional peptides are additional immunogens of the Herpesviridae family. In some embodiments, the additional peptides are selected from peptides of EBV, CMV, HHV6B, HSV1 , and / or VZV. In some embodiments, the additional peptides are from other members of the Herpesviridae family. In some embodiments, the additional peptides are unrelated to the Herpesviridae family.

[0015] Also provided is a method of isolating gB-directed B-cells and / or antibodies from a blood sample obtained from a mammalian subject. In some embodiments, the method comprises contacting the blood sample with a gB polypeptide described herein and isolating gB-directed B- cells and / or antibodies that specifically recognize and bind the gB polypeptide. In some embodiments, the subject has been previously infected or immunized with a Herpesviridae gB polypeptide. In some embodiments, the isolating comprises ELISPOT assay, magnetic cell separation, affinity chromatography, fluorescence-activated cell sorting (FACS), biolayer interferometry (BLI), surface plasmon resonance (SPR), mass photometry (MP), or enzyme- linked immunosorbent assay (ELISA). In some embodiments, the gB-directed B-cells and / or antibodies are directed against a gB polypeptide derived from Epstein Barr Virus (EBV), a Human Cytomegalovirus (CMV), a Human Herpesvirus 6 (HHV6), a Herpes Simplex Virus 1 (HSV1), or a Varicella Zoster Virus (VZV).BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIGS.1A-1G show prefusion-stabilization of the EBV gB ectodomain trimer. (1A) Schematic of the domain organization of full-length Epstein Barr Virus (EBV) gB with thedesigned modifications to produce the gB-deliDoP-l5350A2hN ectodomain indicated. All constructs have deletion of the transmembrane and cytoplasmic domains, fusion loop mutations to enhance solubility, and a C-terminal polyhistidine tag (Table 1). de: deletion of residues 402- 452; A: deletion of residues 394-452; li: GSPPGSPP loop-linker (SEQ ID NO: 51), D: disulfide bond between residues Q527C and I634C (D); P: A293P mutation; o: oP3h trimer domain; l5350A2h(N): the redesigned I5350A chimerized with DI. (1 B) Left, ribbon representation of the AlphaFold2-predicted structure of full length prefusion EBV gB using prefusion CMV gB as a template (PDB 7kdp). Right, schematic of the 3D organization of prefusion gB domains in one protomer. (1C) EBV gB ectodomain design iterations as panels, including gB-deli, gB-A, gB- deD, gB-deliDoP-l5350A, gB-deliDoP-l5350A2h, and gB-deliDoP-l5350A2hN. The locally sharpened cryoEM map (left, surface), model built into the map (middle, ribbons), and schematic of one protomer (cartoon, right) is shown with mutations labeled as in panel (1 A). A composite map from locally refined and sharpened cryoEM reconstructions is shown for EBV gB-deliDoP-l5350A2h and EBV gB-deliDoP-l5350A2hN. Domain I (DI), domain II (Dll), domain III (Dill), domain IV (DIV), domain V (DV), glycans (dark gray), oP3h (black), and l5350A2h (gray). All constructs have the A684-857 mutation, fusion loop mutations, and a C-terminal HIS tag. (1 D) Comparison of the DV C-termini among EBV gB-deliDoP-l5350A2h, CMV gB (PDB 7kdp) and HSV1 gB (PDB 6z9m). DIV is omitted from the renderings for viewing DV. (1 E) Comparison of the DI-DII relative orientation and intervening hinge among EBV, CMV, and HSV1 gB structures. Open triangles show helically restructured hinge residues characteristic of the postfusion state; P: A293P mutation. (1 F) Comparison of the organization of the Dill central helices among EBV gB-deliDoP-l5350A2h, EBV gB-deliDoP-l5350A2hN, prefusion CMV gB (PDB 7kdp), and prefusion HSV1 gB (PDB 6z9m), rendered as ribbons (upper panel) or schematics (lower panel; SEQ ID NOs: 44-46) (1G) Comparison of the organization of the DI I- Dll I interface among EBV gB-deliDoP-l5350A2h, EBV gB-deliDoP-l5350A2hN, prefusion CMV gB (PDB 7kdp), and prefusion HSV1 gB (PDB 6z9m), rendered as ribbons (upper panel) or schematics (lower panel).

[0017] FIGS.2A-2H show design and validation of oP3h and l5350A2hN trimers for prefusion gB stabilization. (2A) Wheel diagram (top; SEQ ID NO: 47) and AlphaFold2-predicted structure of the oP3h trimer, shown in ribbon representation with residues labeled (bottom). (2B) Top, Coomassie stained SDS-PAGE of SARS-CoV-2 S ectodomain constructs with C-terminally chimerized foldon, one to six oP3h (RINEIER; SEQ ID NO: 48) repeats, or four oP3h (RINAIET ; SEQ ID NO: 49) repeats. Glycoproteins were expressed and purified in parallel and normalized to a standard volume for comparison of relative yields after affinity purification. Bottom, theproportion of S ectodomains (after affinity purification) that are clearly folded in the prefusion state, as determined by 2D classification of negatively stained samples. Exemplar prefusion 2D class averages are shown underneath. (2C) CryoEM maps of SARS-CoV-2 S with three RINEIER (top; SEQ ID NO: 48) or four RINAIET (bottom; SEQ ID NO: 49) C-terminal oP3h repeats with each protomer colored distinctly. Insets show local refinement maps (mesh) and refined models (ribbons) encompassing the oP3h fusion. (2D) Top, Coomassie stained nonreducing SDS-PAGE analysis of EBV gB-deliD with (+) and without (-) C-terminal oP3h. Bottom, cryoEM map (mesh) and model (ribbons) of the EBV gB-deliDoP-l5350A2h domain V (DV) and oP3h fusion. (2E) AlphaFold2-predicted models of the l5350A2h and l5350A2hN designs, showing the predicted relative orientation of DI to which they are chimerized with. (2F Left, Coomassie stained non-reducing SDS-PAGE and 2D EM class averages of negatively stained gB-deliD ectodomains with (+) and without (-) oP3h, A293P (P), or l5350A2h. (2G) Top, Schematic showing the addition of 1 (center) or 4 (right) heptad repeats to the coiled coils of l5350A2h. Bottom, EM reconstructions of negatively stained EBV gB-deliDoP-l5350A2h with the corresponding heptad repeat extensions. (2H) Local refinement cryoEM maps (transparent surface) and model (black ribbons) of the l5350A2h or l5350A2hN regions within EBV gB- deliDoP-l5350A2h and EBV gB-deliDoP-l5350A2hN, respectively. Open arrows: relative orientation of DI. The diameter at which DI is held by l5350A2h or l5350A2hN is labeled on the right.

[0018] FIGS.3A-3X show a broadly generalizable herpesvirus gB prefusion-stabilization strategy. Domain I (DI), Domain II (Dll), Domain III (Dill,), Domain IV (DIV), and Domain V (DV) are shown, along with the N-terminal residues (black). Structural analysis is presented for prefusion-stabilized constructs of HSV1 gB-deliD (3A-3I), VZV L687K gB-deliD (3J-3N), HHV6B gB-deliDoP-l5350A2h (3O-3S), and CMV gB-deliDoP-l5350A2hN (3T-3Y). The panels on the left show sharpened cryoEM maps (3A,3I), composite cryoEM maps of locally refined and sharpened maps (30, 3T), and the corresponding ectodomain ribbon diagrams (3B,3F,3K,3P,3U). In the middle panels (3C,3G,3L,3Q,3V), DI, Dll, and N-terminal residue are depicted as ribbons with the magnitude of DI rotation between the prefusion and postfusion states indicated. Insets show zoomed-in views of N-terminal residue interactions with the Dl- hinge (black box) or of the postfusion hinge (grey dashed box, PDB 3nwf for HSV-1 gB and 6vlk for VZV). Restructuring of the Dl-hinge is indicated (triangle). In the right panels, Dll and Dill are shown for the trimer (3D,3H,3M,3R,3W; SEQ ID NOs: 45, 45, 45, 50, 46, respectively) or for the protomer (3E,3I,3N,3S,3X), represented as ribbon diagrams (upper panel) andcorresponding schematics (lower panel). (3Y) Inset shows the binding of CMV gB N-terminal residues to DIV, including the sharpened cryoEM density map (mesh).

[0019] FIGS.4A-4L show antigenicity of prefusion-stabilized gB. (4A) Binding of the 3A3 and 3A5 antibodies to immobilized gB ectodomain constructs analyzed by ELISA. Bars represent geometric mean (± SD) EC50s obtained from four technical replicates. (4B-4C) CryoEM structure of EBV gB-deliDoP-l5350A2hN in complex with the 3A3 and 3A5 neutralizing antibodies. (4B) A global view of gB bound to 3A3 and 3A5 Fabs, with Domain I (DI), Domain II (Dll), Domain III (Dill), Domain IV (DIV), Domain V (DV), and glycans (dark) rendered as surfaces. The 3A3 and 3A5 Fab fragments are shown as light and dark gray ribbons, respectively. (4C) Local refinement-derived models of the antibody-bound regions, showing 3A3 engagement of Dll (upper panel) and 3A5 binding to DIV (lower panel), with protein chains depicted as ribbons. (4D-4I) Structural alignment of Dll-directed antibody complex structures, including 3A3 (from 4B), D48 (PDB 8kfa), and SM5-1 (PDB 4ot1), with the prefusion-1 and prefusion-2 conformations of the DII-DIII interface (4D-4F) or between protomers (4G-4I).Antibody Fab variable domains are shown as transparent surfaces outlined with black dashed lines (4D-4F) or as black ribbons (4G-4I). Dll and Dill are rendered as ribbons, with insets highlighting steric clashes in the prefusion-1 conformation (indicated by a star). (4J-4L) Mapping of the footprints of DIV-directed antibodies (dashed outlines), including 3A5 (from 4C), 93k (PDB 6vn1), and 3-25 (PDB 6uoe), with the ectodomains (rendered as surfaces) of EBV gB-deliDoP- l5350A2hN, VZV gB-deliD + L687K, and CMV gB-deliDoP-l5350A2hN, respectively. The predicted N-terminal residues that interact with DIV are shown as light ribbons, taken from the corresponding region in postfusion HSV1 gB (PDB 3nwf).

[0020] FIG. 5 shows anticipated and deduced molecular mechanisms of gB fusion. Domain I (DI), Domain II (Dll), Domain III (Dill), Domain IV (DIV), and Domain V (DV) are shown, along with the DII-DIII loop (black line), the cytoplasmic and transmembrane regions of gB (gray), the viral envelope (light gray), and the host membrane (dark gray). While gB is a trimer, a single protomer is shown here for clarity, (a-c) Schematic representation of the anticipated mechanism of gB-mediated membrane fusion at the outset of this study: beginning with prefusion gB (a), Dll rotates around Dill, repositioning DI for host membrane binding (b). Subsequently, DV undergoes a conformational flip around Dill, transitioning gB into the postfusion state and merging the viral and host membranes (c). (d-h) Mechanism of gB-mediated fusion inferred from structural analyses in this study (upper) with corresponding full-length (*) or ectodomain structures listed below, including PDB accession codes. Prefusion gB Dill can adopt twoconformations: prefusion-1 (d) and prefusion-2 (e). Conformational rearrangement of the DI-DII hinge leads to DI rotation, disengaging the fusion loops from the viral envelope and releasing DV, yielding intermediate-1 (f). DV then refolds into its postfusion conformation, displacing Dll on Dill and releasing Dll. As DV is anchored to the viral envelope, this rearrangement reorients the entire ectodomain, positioning the DI fusion loops for host membrane engagement, yielding intermediate-2 (g). Finally, Dll binds its postfusion interface on Dill, stabilizing the postfusion gB conformation (h).

[0021] FIG. 6 shows alignment of EBV gB (SEQ ID NO: 1) with corresponding ectodomain sequences for CMV (SEQ ID NO: 2), HHV6B (SEQ ID NO: 3), HSV1 (SEQ ID NO: 4), and VZV (SEQ ID NO: 5). Note that alternating sheets of this figure contain first and second halves of each row, as indicated by numerical markers. See FIG. 7 and Table 1 for exemplary modifications to these ectodomains.

[0022] FIG. 7 provides exemplary modifications to the gB ectodomains, namely EBV_gB_deliDoP_l5350A2h (SEQ ID NO: 12) and EBV_gB_deliDoP_l5350A2hN (SEQ ID NO: 15). Note that alternating sheets of this figure contain first and second halves of each row, as indicated by numerical markers.DETAILED DESCRIPTION

[0023] The invention disclosed herein is based on the discovery of Herpesviridae glycoprotein B (gB) polypeptides comprising a modified ectodomain, and which is stabilized in the prefusion state. This provides a solution to the problem of inability to develop inhibitors and vaccines directed against these viral pathogens.

[0024] The data presented herein demonstrate that modifications to DI, Dll, and DV subdomains are important to stabilization of gB in the prefusion state. This discovery is contrary to the expectation one skilled in the art might have held that stabilizing Dll only, or the link between Dll and Dill would be the best target for development of an immunogenic target. Surprisingly, the additional modifications of gB-deliDoP-l5350A2hN relative to gB-deliDoP-l5350A2h were found to result in a further improved stabilization of gB in the prefusion state. As demonstrated in Example 1 below, EBV gB-deliDoP-l5350A2hN folds as a prefusion gB trimer.

[0025] Definitions

[0026] All scientific and technical terms used in this application have meanings commonly used in the art unless otherwise specified. As used in this application, the following words or phrases have the meanings specified.

[0027] As used herein, a “therapeutically effective amount” is an amount sufficient to effect an alleviation of symptoms or to hinder or delay the onset of symptoms. In some embodiments, a therapeutically effective amount is an amount sufficient to elicit an immune response that, for example, reduces viral infection or limits the bodily damage resulting from viral infection.

[0028] As used herein, a “significant difference” means a difference that can be detected in a manner that is considered reliable by one skilled in the art, such as a statistically significant difference, or a difference that is of sufficient magnitude that, under the circumstances, can be detected with a reasonable level of reliability. In one example, an increase or decrease of 10% relative to a reference sample is a significant difference. In other examples, an increase or decrease of 20%, 30%, 40%, or 50% relative to the reference sample is considered a significant difference. In yet another example, an increase of two-fold relative to a reference sample is considered significant.

[0029] “Nucleotide sequence” or “nucleic acid molecule” refers to a heteropolymer of deoxyribonucleotides, ribonucleotides, or peptide-nucleic acid sequences that may be assembled from smaller fragments, isolated from larger fragments, or chemically synthesized de novo or partially synthesized by combining shorter oligonucleotide linkers, or from a series of oligonucleotides, to provide a sequence which is capable of expressing the encoded protein.

[0030] As used herein, the term “active fragment” refers to a substantial portion of an oligonucleotide that is capable of performing the same function of specifically hybridizing to a target polynucleotide.

[0031] As used herein, "hybridizes," "hybridizing," and "hybridization" means that the oligonucleotide forms a noncovalent interaction with the target DNA molecule under standard conditions. Standard hybridizing conditions are those conditions that allow an oligonucleotide probe or primer to hybridize to a target DNA molecule. Such conditions are readily determined for an oligonucleotide probe or primer and the target DNA molecule using techniques well known to those skilled in the art. The nucleotide sequence of a target polynucleotide is generally a sequence complementary to the oligonucleotide primer or probe. The hybridizing oligonucleotide may contain nonhybridizing nucleotides that do not interfere with forming the noncovalent interaction. The nonhybridizing nucleotides of an oligonucleotide primer or probemay be located at ari end of the hybridizing oligonucleotide or within the hybridizing oligonucleotide. Thus, an oligonucleotide probe or primer does not have to be complementary to all the nucleotides of the target sequence as long as there is hybridization under standard hybridization conditions.

[0032] The term "complement" and "complementary" as used herein, refers to the ability of two nucleic acid molecules to base pair with each other. For example, in DNA, adenine (A) is complementary to thymine (T). In RNA, adenine (A) is complementary to uracil (U). In some embodiments, complementarity refers to an antisense compound that is capable of base pairing with its target nucleic acid. For example, if a nucleobase at a certain position of an antisense compound is capable of hydrogen bonding with a nucleobase at a certain position of a target nucleic acid, then the position of hydrogen bonding between the oligonucleotide and the target nucleic acid is considered to be complementary at that nucleobase pair. Nucleobases comprising certain modifications may maintain the ability to pair with a counterpart nucleobase and thus, are still capable of nucleobase complementarity. Typically, two DNA molecules are complementary if they hybridize under the standard conditions referred to above. Typically, two DNA molecules are complementary if they have at least about 80% sequence identity, preferably at least about 90% sequence identity.

[0033] As used herein, "pharmaceutically acceptable carrier" or “excipient” includes any material which, when combined with an active ingredient, allows the ingredient to retain biological activity and is non-reactive with the subject's immune system. Examples include, but are not limited to, any of the standard pharmaceutical carriers such as a phosphate buffered saline solution, water, emulsions such as oil / water emulsion, and various types of wetting agents. Preferred diluents for aerosol or parenteral administration are phosphate buffered saline or normal (0.9%) saline.

[0034] Compositions comprising such carriers are formulated by well-known conventional methods (see, for example, Remington's Pharmaceutical Sciences, 18th edition, A. Gennaro, ed., Mack Publishing Co., Easton, PA, 1990).

[0035] As used herein, the term "subject" includes any human or non-human animal. The term "non-human animal" includes all vertebrates, e.g., mammals and non-mammals, such as non- human primates, horses, sheep, dogs, cows, pigs, chickens, and other veterinary subjects. In a typical embodiment, the subject is a human.

[0036] As used herein, “a” or “an” means at least one, unless clearly indicated otherwise.

[0037] Polypeptides and Nucleic Acid Molecules

[0038] Described herein are Herpesviridae glycoprotein B (gB) polypeptides comprising a modified ectodomain, and which is stabilized in the prefusion state. In some embodiments, the ectodomain comprises at least one of the following modifications. In some embodiments, the gB ectodomain comprises deletion of residues 402-452 or 394-452 of SEQ ID NO: 1. In some embodiments in residues 402-452 or 394-452 have been deleted, and a GSPPGSPP (SEQ ID NO: 51) linker is inserted in its place. In some embodiments, the gB ectodomain comprises a disulfide bond between Q527C-E634C. In some embodiments, the gB ectodomain comprises an A293P mutation, a oP3H trimer, and / or a I5350A modified to harbor an extendable and structured link to DI (l5350A2h or l5350A2hN). In some embodiments, the transmembrane and / or cytoplasmic domains have been deleted. In some embodiments, gB is derived from an Epstein Barr Virus (EBV), a Human Cytomegalovirus (CMV), a Human Herpesvirus 6 (HHV6), a Herpes Simplex Virus 1 (HSV1), or a Varicella Zoster Virus (VZV). In some embodiments, the gB polypeptide further comprises a mutation of the fusion loop. One example of a mutation of the fusion loop is provided in Table 1. In some embodiments, the gB polypeptide comprises gB- deliDoP as described in Table 1. In some embodiments, the gB polypeptide comprises gB- deliDoP-l5350A2h or gB-deliDoP-l5350A2hN as described in Table 1 or in FIG. 7.

[0039] SEQ ID NO: 1 provides the unmodified amino acid sequence of gB of EBV (NCBI Reference Sequence: YP_401713.1):

[0040] MTRRRVLSVWLLAALACRLGAQTPEQPAPPATTVQPTATRQQTSFPFRVCELSSHGDLFRF SSDIQCPSFGTRENHTEGLLMVFKDNIIPYSFKVRSYTKIVTNILIYNGWYADSVTNRHEEKFSVDSYE TDQMDTIYQCYNAVKMTKDGLTRVYVDRDGVNITVNLKPTGGLANGVRRYASQTELYDAPGWLIWT YRTRTTVNCLITDMMAKSNSPFDFFVTTTGQTVEMSPFYDGKNKETFHERADSFHVRTNYKIVDYD NRGTNPQGERRAFLDKGTYTLSWKLENRTAYCPLQHWQTFDSTIATETGKSIHFVTDEGTSSFVTN TTVGIELPDAFKCIEEQVNKTMHEKYEAVQDRYTKGQEAITYFITSGGLLLAWLPLTPRSLATVKNLT ELTTPTSSPPSSPSPPAPSAARGSTPAAVLRRRRRDAGNATTPVPPTAPGKSLGTLNNPATVQIQFA YDSLRRQINRMLGDLARAWCLEQKRQNMVLRELTKINPTTVMSSIYGKAVAAKRLGDVISVSQCVPV NQATVTLRKSMRVPGSETMCYSRPLVSFSFINDTKTYEGQLGTDNEIFLTKKMTEVCQATSQYYFQ SGNEIHVYNDYHHFKTIELDGIATLQTFISLNTSLIENIDFASLELYSRDEQRASNVFDLEGIFREYNFQ AQNIAGLRKDLDNAVSNGRNQFVDGLGELMDSLGSVGQSITNLVSTVGGLFSSLVSGFISFFKNPFG GMLILVLVAGWILVISLTRRTRQMSQQPVQMLYPGIDELAQQHASGEGPGINPISKTELQAIMLALHE QNQEQKRAAQRAAGPSVASRALQAARDRFPGLRRRRYHDPETAAALLGEAETEF

[0041] See FIG. 6 for alignment of SEQ ID NO: 1 with corresponding ectodomain sequences for CMV (SEQ ID NO: 2), HHV6B (SEQ ID NO: 3), HSV1 (SEQ ID NO: 4), and VZV (SEQ ID NO: 5). See FIG. 7 and Table 1 for exemplary modifications to these ectodomains.

[0042] In some embodiments, the gB polypeptide comprises an amino acid sequence having at least 80% sequence identity to an amino acid sequence of SEQ ID NO: 25-43, or of Table 1 (SEQ ID NO: 6-24). In some embodiments, the sequence identity is at least 85%. In some embodiments, the sequence identity is at least 90%. In some embodiments, the sequence identity is at least 95%. In some embodiments, the gB polypeptide comprises an amino acid sequence of SEQ ID NO: 25-43, or of Table 1 (SEQ ID NO: 6-24). In some embodiments, the gB polypeptide comprises EBV gB-deliDoP-l5350A2h (SEQ ID NO: 12) or EBV gB-deliDoP- l5350A2hN (SEQ ID NO: 15). In some embodiments, the gB polypeptide is in soluble form.

[0043] Also described herein is a nucleic acid molecule that encodes the gB polypeptide. In some embodiments, the nucleic acid molecule is RNA, including mRNA, or DNA. In some embodiments, the nucleic acid molecule comprises a modified nucleotide. In some embodiments, the modified nucleotide comprises a methylpseudouridine.

[0044] In some embodiments, the nucleic acid molecule comprises an open reading frame capable of expressing the gB polypeptide or an immunogenic fragment thereof. The nucleic acid molecule can optionally encode a second immunogen, or multiple immunogenic fragments of gB or another glycoprotein.

[0045] Compositions

[0046] Described herein is a composition comprising the gB polypeptide or the nucleic acid molecule encoding the gB polypeptide. Such compositions optionally include a pharmaceutically acceptable carrier, excipient, and / or adjuvant. In some embodiments, the composition is free of adjuvant. In some embodiments, the composition is formulated for use as a vaccine. In some embodiments, the composition is formulated for delivery of multiple immunogens.

[0047] In some embodiments, the composition comprises additional peptides or nucleic acids encoding additional peptides. In some embodiments, the additional peptides are additional immunogens of the Herpesviridae family. In some embodiments, the additional peptides are selected from peptides of EBV, CMV, HHV6B, HSV1 , and / or VZV. In some embodiments, the additional peptides are from other members of the Herpesviridae family. In some embodiments, the additional peptides are unrelated to the Herpesviridae family.

[0048] Methods

[0049] Methods of making and using the polypeptides, nucleic acid molecules, and compositions described herein are known in the art, and include representative methods of synthesizing and cloning described in Example 1 below. In some embodiments, the method of use is for eliciting an immune response to a Herpesviridae family virus in a mammalian subject. The method comprises administering a composition as described herein to the subject. In some embodiments, the amount to be administered to the subject is a therapeutically effective amount. The method comprises administering a composition as described herein to the subject. In some embodiments, the amount to be administered to the subject is an amount sufficient to elicit an immune response.

[0050] In some embodiments, the composition is delivered in conjunction with an adjuvant or an additional therapeutic or immunogenic agent. In some embodiments, the composition is administered in the absence of adjuvant or additional therapeutic or immunogenic agent.

[0051] In some embodiments, the composition is delivered to the subject by intradermal or intramuscular injection.

[0052] In some embodiments, the method of eliciting an immune response involves a single administration of the composition. In some embodiments, the method further includes administering to the subject a booster dose of the composition. A booster dose can comprise any of the gB molecules disclosed herein. In some embodiments, multiple booster doses are administered to the subject.

[0053] Also provided is a method of isolating gB-directed B-cells and / or antibodies from a blood sample obtained from a mammalian subject. In some embodiments, the method comprises contacting the blood sample with a gB polypeptide described herein and isolating gB-directed B- cells and / or antibodies that specifically recognize and bind the gB polypeptide. In some embodiments, the subject has been previously infected or immunized with a Herpesviridae gB polypeptide. In some embodiments, the isolating comprises ELISPOT assay, magnetic cell separation, affinity chromatography, fluorescence-activated cell sorting (FACS), biolayer interferometry (BLI), surface plasmon resonance (SPR), mass photometry (MP), or enzyme- linked immunosorbent assay (ELISA). In some embodiments, the gB-directed B-cells and / or antibodies are directed against a gB polypeptide derived from Epstein Barr Virus (EBV), a Human Cytomegalovirus (CMV), a Human Herpesvirus 6 (HHV6), a Herpes Simplex Virus 1 (HSV1), or a Varicella Zoster Virus (VZV).EXAMPLES

[0054] The following examples are presented to illustrate the present invention and to assist one of ordinary skill in making and using the same. The examples are not intended in any way to otherwise limit the scope of the invention.

[0055] Example 1 : Computational design of prefusion-stabilized Herpesvirus qB trimers

[0056] In the absence of effective vaccines, most human-infecting members of the Herpesvirus family cause considerable morbidity and mortality. Herpesviruses infect most of the global population, establishing lifelong infections linked to cancer, congenital disorders, and neurodegeneration. The conserved gB fusion protein is essential for viral entry and a prime target for immune intervention. However, gB readily undergoes an irreversible transition from a prefusion to a postfusion conformation, altering its antigenic landscape, and limiting mechanistic studies of viral entry into cells. In this Example, we stabilize the prefusion state of Epstein-Barr virus (EBV) gB using structure-guided and computational protein design, and determine its structure using cryoEM. We show that this prefusion-stabilization strategy is generalizable to Human Cytomegalovirus, Human Herpesvirus 6, Herpes Simplex Virus 1 , and Varicella zoster virus. The conformation of the gB N-terminus is revealed adding valuable mechanistic insight into the antibodies that neutralize gB. These structures and the identified conformational intermediates support a molecular mechanism for gB distinct from previous assumptions. These results strengthen our understanding of class III fusion proteins, leaving us better prepared to stabilize vaccine antigens for several viral families.

[0057] A major barrier to understanding Herpesvirus entry has been the inability to stabilize the prefusion conformation of gB. This metastable state collapses to the postfusion conformation during purification. Prefusion-stabilization has been critical for designing effective vaccines against other enveloped viruses, including SARS-CoV-2 and RSV (34-36). While SARS-CoV-2 and RSV rely on class I fusion proteins, Herpesvirus gB is a class III fusion protein, which is structurally distinct and has proven difficult to stabilize in the prefusion state despite extensive efforts (37, 38). The observation that full length gB frequently adopts the postfusion conformation even on intact virions — possibly as a result of sample preparation — emphasizes the extraordinary gB metastability and the need for prefusion-stabilizing mutations (39, 40).

[0058] The H516P mutation increases the full-length HSV1 gB prefusion-to-postfusion ratio, as does its counterpart in VZV gB (40, 41). However, this approach is only effective for full-length, membrane-embedded alpha-Herpesvirus gB but does not sufficiently stabilize the purifiedrecombinant HSV1 gB ectodomain, nor other Herpesvirus gBs (40, 42). Full length prefusion HSV1 H516P gB and the corresponding VZV H527P gB mutant were visualized at ~1 nm resolution using cryo-electron tomography (cryoET), as was full-length CMV gB at ~2-3 nm resolution, revealing their prefusion domain organization (32, 40, 41). Using a cryo-electron microscopy (cryoEM) structure of full-length CMV gB stabilized with a CMV-specific inhibitor and chemical crosslinking (43) a recent study identified mutations stabilizing the CMV gB ectodomain in a prefusion-like conformation (44). This structure shares many features with the prefusion state, though differences in domain I (DI) orientation relative to the prefusion state were noted (44). As a result the immunogenicity of a fully prefusion-stabilized Herpesvirus gB remains to be explored. No structural information exists for prefusion EBV gB or any other gamma-Herpesviruses, hindering our understanding of cell entry, our ability to isolate prefusionspecific antibodies, and our capacity to design countermeasures against this important class of pathogens.

[0059] This Example demonstrates how to stabilize the prefusion conformation of the EBV gB ectodomain trimer. This was achieved by leveraging structure-based and machine-learning guided protein design, and identifying conformational intermediates that illuminate the molecular mechanism of gB-mediated fusion. This Example shows that this stabilization strategy is broadly applicable to alpha-, beta-, and gamma-Herpesviruses, pointing to a shared fusion mechanism and paving the way to study Herpesvirus entry into cells, isolate prefusion gB-specific neutralizing antibodies, and develop inhibitors and vaccines against these pathogens.

[0060] METHODS

[0061] Computational stabilization of gB and design of oP3h and !5350A2h

[0062] Alphafold2 (v2.2) (48) via ColabFold (v1 .3) (88) was used to predict the trimeric prefusion conformation structure of EBV gB using the prefusion conformation structure of CMV gB (43) as a template. Using this structure, ProteinMPNN(47) and Disulfide by Design 2.0 (89) were used in parallel to identify stabilizing mutations, including the A293P and Q527C-E634C intermolecular disulfide bond mutations in EBV gB. The relatively soluble HSV1 fusion loop residues were initially used in EBV gB ectodomains as previously described (90), but were replaced with ProteinMPNN(47) optimized fusion loop mutations subsequent to the addition of l5350A2h.

[0063] To design an ultrastable trimer that was narrow and short enough to fit partly in the interior of gB and not clash with additional stabilizing trimers, we iteratively cycled thecoordinates of the GCN4 trimer (PDB 4dme) between ProteinMPNN and Alphafold2(47, 48) to optimize the design until convergence on a three-helix bundle (designated optimized parallel 3- helix bundle, herein oP3h) harboring the RINEIER heptad repeat motif. These repeat sequences were predicted to tightly pack isoleucine side chains in the coiled-coil core with interdigitating arginine, glutamate, and asparagine side chains forming a network of salt bridges and hydrogen bonds at the periphery. During optimization, the heptad motif was adjusted to RINAIET to increase the surface presentation of hydrophobic moieties, anticipating hydrophobic interactions with the air-water interface may favor cryoEM particle views perpendicular to the C3 symmetry axis. To add oP3h to the coiled coils of SARS-CoV-2 S and EBV gB, the fusion protein coiled coil abcdefg repeats (where a and d are hydrophobic residues) were first matched to oP3h. While not necessary for all applications of oP3h, a smoother transition between the coiled coil of the fusion protein and oP3h was accomplished using mutations identified in ProteinMPNN(47) and screened in Alphafold2(48) before adding to ectodomain constructs.

[0064] To redesign the N- and C-termini of I5350A to create l5350A2h, RFDiffusion(91) was first used to place de novo backbone residues with a regular secondary structure; cycling between ProteinMPNN and Alphafold2(47, 48) was then used to identify residues that would promote formation of this structure. This enabled creating a helix-turn-helix, antiparallel coiled coil, and p- strands that match strands of gB DI to I5350A. Similar to the repeats of oP3h, the use of a coiled coil in this design enabled extension of the structured link between gB and l5350A2h simply by adding heptad repeats: RIQELER repeats to the N-terminus of l5350A2h plus RLREEIN repeats to the C-terminus of l5350A2h. To further redesign l5350A2h to create l5350A2hN, the coordinates of EBV gB-deliDoP-l5350A2h were used with ProteinMPNN to optimize the coiled coil residues. Next, RFDiffusion(91) was used to place the p-strands that match strands of gB DI in a direction ~30° tilted from the corresponding residues in l5350A2h. In the context of trimeric l5350A2h these residues point away from the C3 symmetry axis, while they point ~30° more parallel to this axis in l5350A2hN; consequently l5350A2hN was anticipated to position DI with a narrower diameter than l5350A2h. Cycling between ProteinMPNN and Alphafold2(47, 48) was then used to identify residues that would promote formation of this structure.

[0065] DNA constructs

[0066] Sequences of gB incorporating the gB-deli modifications (Table 1) were synthesized and cloned into a pCMV vector by Genscript. Mutations were introduced using In-Fusion cloning (Takara, 638949). Antibody variable heavy and light chain sequences for monoclonal antibodies3A3 and 3A5 (PDB ID 7FBI) were cloned into matched human IgG 1 heavy and kappa light chain constant regions (PDB ID 8FXB_H and Uniprot P01834, respectively). Each chain was preceded by an N-terminal azurocidin signal peptide; the heavy chain also contained a C- terminal AVI-HIS affinity tag. Constructs were synthesized and cloned into pCMV by Genscript.

[0067] Recombinant ectodomain and monoclonal antibody production

[0068] Recombinant gB ectodomains and monoclonal antibodies were expressed in 25 mL cultures of Expi293F cells (ThermoFisher Scientific), a 293-derived suspension-adapted cell line, grown in Expi293 Expression Medium at 37 °C in a humidified incubator with 8% CO2, shaking at 130 rpm. Cultures were transfected at a density of 3 x 106 cells / mL using the ExpiFectamine 293 Transfection Kit (ThermoFisher Scientific), and supernatants were harvested four days post-transfection. His-tagged ectodomains and antibodies were purified from clarified supernatants using 2 mL cobalt-affinity resin (TALON, Takara Bio), supplemented with 18 pL of 1 M CoCI2to offset cobalt depletion. Columns were washed with 200 column volumes of buffer (50 mM Tris-HCI pH 8.0, 150 mM NaCI, 5 mM imidazole), and proteins were eluted with 600 mM imidazole in the same buffer. Ectodomains and antibodies were concentrated and buffer-exchanged into Tris-HCI pH 8.0, 150 mM NaCI using 100 kDa centrifugal filters (Amicon Ultra 0.5 mL, MilliporeSigma). At this stage, ectodomains were screened by negative-stain electron microscopy (see below). Samples selected for cryoEM analysis, as well as monoclonal antibodies, were further purified by size-exclusion chromatography (SEC) using a Superose 6 Increase 10 / 300 GL column (Cytiva) equilibrated in Tris-HCI pH 8.0, 150 mM NaCI, followed by final concentration (100 kDa centrifugal filters, Amicon Ultra) and flash freezing in liquid nitrogen. Fab fragments were generated by digesting IgG with LysC (NEB, P8109S) at 1 :3000 w / w enzyme-to-substrate ratio at 37 °C for 16 h. Undigested IgGs and Fc fragments were removed using a Protein A affinity resin (GenScript, L00210), and the flow-through containing Fabs was passed through a 100 kDa centrifugal filter to remove any additional undigested IgG, followed by final concentration over a 30 kDa filter (Amicon Ultra 0.5 mL, MilliporeSigma).

[0069] Negative-stain EM sample preparation

[0070] All constructs in this study were negatively stained at a final concentration of 0.01 mg ml-1 using Gilder Grids overlaid with a thin layer of carbon and 2% uranyl formate. Data were acquired using the Leginon software(92) to control a Tecnai T12 transmission electron microscope operated at 120 kV and equipped with a Gatan 4K Ultrascan CCD detector. The dose rate was adjusted to 50 electrons per A2 and each micrograph was acquired in 1 s.For each ectodomain, ~100 micrographs were collected with a defocus range between -1.0 and -2.5 pm. Data were subsequently processed using CryoSPARC(93).[00711 ELISA

[0072] 384-well Maxisorp plates (ThermoFisher Scientific, 464718) were coated with 30 pL per well of 4 pg / mL recombinant EBV gB ectodomain constructs in TBS and incubated for 2 hours at 37 °C. Plates were washed once with TBST and blocked with Blocker Casein in TBS (ThermoFisher Scientific, 37532) for 1 hour at 37 °C. Following blocking, plates were washed four times with TBST using a 405 TS Microplate Washer (BioTek). 3A3 or 3A5 antibodies were applied in a 1 :5 serial dilution series starting at 0.01 mg / mL in TBST and incubated for 1 hour at 37 °C. Plates were then washed four times with TBST, followed by the addition of goat antihuman HRP (ThermoFisher Scientific, A18817) diluted 1 :5000 in TBST and incubated for 1 hour at 37 °C. After four additional washes, 30 pL of TMB Microwell Peroxidase Substrate (Seracare, 5120-0083) was added per well. After 4 minutes, the reaction was quenched with 30 pL of 1 N HCI, and absorbance at 450 nm was measured using a BioTek plate reader. Data were normalized and fit using non-linear regression (inhibitor vs. normalized response, variable slope) in GraphPad Prism version 10.2.3. All conditions were performed in four technical replicates.

[0073] CryoEM sample preparation and data collection

[0074] To prepare the CryoEM grids for all ectodomains, the following concentrations of proteins with (or without as indicated) detergent were prepared: 5 mg / mL EBV gB-deli with 0.02 % (w / v) fluorinated octyl-maltoside (FOM, Anatrace) detergent, 2 mg / mL EBV gB-A, 2 mg / mL EBV gB- deD, 0.1 mg / mL EBV gB-deliDoP-l5350A, 3 mg / mL EBV gB-deliDoP-l5350A2h with 0.01 % (w / v) FOM, 5 mg / mL EBV gB-deliDoP-l5350A2hN with 0.02 % (w / v) FOM, 5 mg / mL CMV gB- deliDoP-l5350A2hN with 0.02 % (w / v) FOM, 3 mg / mL HHV6B gB-deliDoP-l5350A2h with 0.01 % (w / v) FOM, 0.5 mg / mL HSV1 gB-deliD, 1.25 mg / mL VZV gB-deliD + L687K, 1 mg / mL SARS-CoV-2 S-2P + 3X-RINEIER, and 1 mg / mL SARS-CoV-2 S-2P + 4X-RINAIET. FOM was added as indicated to reduce issues of preferred orientation or unfolding at the air water interface. These solutions (3 pL) were added onto a freshly glow discharged 2.0 / 2.0 UltraFoil (94) grid (200 mesh), and plunge frozen using a vitrobot MarkIV (ThermoFisher Scientific) using a blot force of -1 and 6 second blot time at 100% humidity and 23°C. For EBV gB-deliDoP- I5350A, prior to glow discharging the 2.0 / 2.0 UltraFoil (94) grid (200 mesh) was manually overlaid with a thin layer of carbon. Data were acquired using the Leginon software (92) to control an FEI Titan Krios transmission electron microscope operated at 300 kV equipped with a Gatan K3 Summit direct detector and Gatan Quantum GIF energy filter, operated in zero-lossmode with a slit width of 20 eV. The dose rate was adjusted to 15 counts / pixel / s, and each movie was acquired in 75 frames of 40 ms with pixel size 0.83 A. Data were collected with a defocus range between -0.5 and -2.8 pm. For EBV gB-deliDoP-l5350A, HSV1 gB-deliD, SARS-CoV-2 S-2P + 4X-RINEIER, and SARS-CoV-2 S-2P + 4X-RINAIET data were acquired using the Leginon software (92) to control a Glacios transmission electron microscope equipped with a Gatan K3 Summit direct detector and operated at 200 kV. The dose rate was adjusted to 7.5 counts / pixel / s, and each movie was acquired in 100 frames of 50 ms with pixel size 0.89 A. For EBV gB-deD, 3141 movies were collected at 0° tilt, 3712 movies were collected at 15° tilt, 3783 movies were collected at 30° tilt, and 519 movies collected at 45° tilt. For HSV1 gB-deliD and VZV gB-deliD + L687K, data was collected at 30° tilt. All other datasets were collected as a single session without tilt.

[0075] Overall CryoEM data processing

[0076] For all datasets, movie frame alignment and pixel binning by 2-fold was initially carried out using Warp(95), except for EBV gB-deD and EBV gB-deliDoP-l5350A, for which movie frame alignment was performed in CryoSPARC(93) and 2-fold binning was performed at the particle extraction stage (for these datasets the maps were resolved better this way). For all datasets, estimation of the microscope contrast-transfer function parameters using Patch CTF, particle picking using Topaz(96), and particle extraction was carried out in CryoSPARC(93). Reference-free 2D classification was performed using CryoSPARC to select well-defined particle images. Ab initio structure reconstruction and non-uniform refinement in CryoSPARC were then performed, followed by 3D classification in Relion(97) for EBV g B-deli, SARS-CoV-2 S-2P + 4X-RINEIER, SARS-CoV-2 S-2P + 4X-RINAIET, and VZV gB-deliD + L687K datasets. Alternatively, for EBV gB-deliD, EBV gB-deliDoP-l5350A2h, EBV gB-deliDoP-l5350A2hN, EBV gB-deliDoP-l5350A2hN + 3A3 / 3A5 Fabs, CMV gB-deliDoP-l5350A2hN, HHV6B gB-deliDoP- l5350A2h, and HSV1-deliD datasets, ab initio structure reconstruction with multiple classes followed by heterogeneous refinement was performed in CryoSPARC, as this yielded better resolved maps for these datasets. At this stage, particles were subjected to reference-based motion correction in CryoSPARC(93) (for datasets with movie frames aligned in CryoSPARC(93)) or Bayesian polishing using Relion(98) (for datasets with movie frames aligned in WARP) during which the pixels were unbinned. Reference-based motion correction / Polishing was not performed on the EBV gB-A and EBV gB-deliDoP-l5350A datasets, as this did not yield better resolved maps. At this stage, for the HSV1 -deli D dataset ab initio was performed followed by heterogeneous refinement to separate three distinct maps andcorresponding particles. For each dataset, another round of non-uniform refinement in CryoSPARC was performed with global and per-particle defocus refinements as well as beam tilt refinement, parameters which were optimized for each dataset(97, 99). HSV1 -deliD (prefusion-1 / prefusion-1), HSV1-deliD (prefusion-2 / prefusion-1), HSV1-deliD (prefusion- 2 / prefusion-2), VZV gB-deliD + L687K, HHV6B gB-deliDoP-l5350A2h, CMV gB-deliDoP- l5350A2hN, EBV gB-deli, EBV gB-A, EBV gB-deD, EBV gB-deliDoP-l5350A, EBV gB-deliDoP- l5350A2h, EBV gB-deliDoP-l5350A2hN, EBV gB-deliDoP-l5350A2hN + 3A3 / 3A5 Fabs, SARS- CoV-2 S-2P + 4X-RINEIER, and SARS-CoV-2 S-2P + 4X-RINAIET datasets were refined with D3, C3, D3, D3, C3, C3, C3, C3, C1 , C1 , C1 , C1 , C3, and C3 symmetries, respectively. Reported resolutions are based on the gold-standard Fourier shell correlation (FSC) of 0.143 criterion and Fourier shell correlation curves were corrected for the effects of soft masking by high-resolution noise substitution(100, 101).

[0077] CryoEM data processing for local features

[0078] For SARS-CoV-2 S-2P + 3X-RINEIER and SARS-CoV-2 S-2P + 4X-RINAIET datasets, a mask surrounding oP3h was generated in UCSF Chimera(102), and used for local refinement of this region applying C3 symmetry in CryoSPARC(93). UCSF Chimera(102) was used to create a mask surrounding the region corresponding to an individual trimer, which was used for local refinement of this region applying C3 symmetry in CryoSPARC(93).

[0079] For refining the map of l5350A2h from EBV gB-deliDoP-l5350A2h, masks were created in UCSF Chimera(102) surrounding the gB ectodomain and l5350A2h, these masks were used for multibody refinement followed by particle signal subtraction in Relion (103, 104) to remove signal from the gB ectodomain, and then the mask surrounding l5350A2h was used for local refinement of this region applying C3 symmetry in CryoSPARC(93). For refining the map of DI I- DV from EBV gB-deliDoP-l5350A2h, masks were created in UCSF Chimera(102) surrounding the DII-DV from the gB ectodomain and DI with l5350A2h, these masks were used for multibody refinement followed by particle signal subtraction in Relion(103) to remove signal from DI and l5350A2h, and then the mask surrounding DII-DV was used for local refinement of this region in CryoSPARC(93). For refining the map of DIII-DV from EBV gB-deliDoP-l5350A2h, a mask of this region was created in UCSF Chimera(102) and used to locally refine the particles from the local refinement of DII-DV above, applying C3 symmetry in CryoSPARC(93). DI is framed by two semi-flexible pivot points and remains poorly resolved in EBV gB-deliDoP-l5350A2h maps, and it is too small for local refinement strategies on its own.

[0080] To improve the local resolution of the EBV gB-deliDoP-l5350A2hN + 3A3 / 3A5 Fab map, we performed localized refinement with region-specific masking. A mask encompassing domains DII-DV and the Fab variable regions was generated in UCSF Chimera(102) and used for local refinement with C3 symmetry, followed by symmetry expansion in CryoSPARC(93). To further resolve individual Fab interactions, additional masks were generated: one targeting Dll and the 3A3 Fab variable region, and another encompassing DIV and the 3A5 Fab variable region. These masks were used for localized 3D classification of the symmetry-expanded particles, followed by local refinement in CryoSPARC(93).

[0081] For locally refining the map of EBV gB-deliDoP-l5350A2hN, masks were created in UCSF Chimera(102) surrounding the gB ectodomain and l5350A2hN, these masks were used for multibody refinement followed by particle signal subtraction in Relion(103) to remove signal from the gB ectodomain or l5350A2hN. Then a mask surrounding l5350A2hN or gB, was used for local refinement of this region applying C3 or C1 symmetry, respectively, in CryoSPARC(93), yielding maps for l5350A2hN and the gB ectodomain. Finally, to improve the local resolution of DII-DV of EBV gB-deliDoP-l5350A2hN, a mask of this region was created in UCSF Chimera(102) and used to locally refine the particles from the local refinement of DI-DV above, applying C3 symmetry in CryoSPARC(93).

[0082] For locally refining the map of EBV gB-deliDoP-l5350A2hN, masks were created in UCSF Chimera(102) surrounding the gB ectodomain and l5350A2hN, these masks were used for multibody refinement followed by particle signal subtraction in Relion(103) to remove signal from the gB ectodomain or l5350A2hN. Then a mask surrounding l5350A2hN or gB, was used for local refinement of this region applying C3 or C1 symmetry, respectively, in CryoSPARC(93), yielding maps for l5350A2hN and the gB ectodomain. Finally, to improve the local resolution of DII-DV of EBV gB-deliDoP-l5350A2hN, a mask of this region was created in UCSF Chimera(102) and used to locally refine the particles from the local refinement of DI-DV above, applying C3 symmetry in CryoSPARC(93).

[0083] Cryo-EM model building and analysis

[0084] For VZV gB-deliD, EBV gB-deli, EBV gB-deD, EBV gB-deliDoP-l5350A2h, and EBV gB- deliDoP-l5350A2hN maps, EMready(105) was used to improve map quality for model building; otherwise the unsharpened maps and locally sharpened maps were used for model refinement and deposition. Initial models of ectodomains were predicted using Alphafold2(48) and rigidly docked into corresponding maps in UCSF Chimera(102). Models were then refined and rebuilt by iteratively cycling between Coot, ISOLDE, and Rosetta: Briefly, self-restraints between atomsin these models were created to maintain structured features prior to flexibly fitting the model coordinates to the map in Coot(106, 107). At this stage, ISOLDE(108) (implemented in ChimeraX(109)) was used to relax the model into the map with an AMBER forcefield and remodel as necessary. Rosetta(1 10, 1 1 1 ) was then used to further optimize fit and geometry. Models were validated using MolProbity(1 12), EMringer(1 13), and Phenix(1 14). Figures were generated using UCSF ChimeraX(109) and UCSF Chimera(102).Table 1 : Amino acid sequences of designed ectodomain constructs.Note: fusion loop mutations are framed by square brackets; deletion mutations are demarcated with a slash; designed trimers and mutations designed to transition to these trimers areunderlined; point mutations are underlined and bold; the HIS-tag is in italics. These same sequences, with His-tag omitted, are presented in SEQ ID NO: 25-43.

[0085] RESULTS

[0086] The results described below are available in greater detail at McCallum M, Veesler D. Computational design of prefusion-stabilized Herpesvirus gB trimers. bioRxiv [Preprint], 2024 Oct 24:2024.10.23.619923. doi: 10.1101 / 2024.10.23.619923. Supplemental Figures S1-S10 and Supplemental Tables S2-S4 referenced below can be found in McCallum & Veesler 2024. Detailed description of the content of Supplemental Figures S1 -S10 can be found at the end of the Results section. For all Figures referenced herein, including Figs. 1-7 and Supplemental Figures S1-S10, colors referenced in the descriptions can be found in the online version of these figures.

[0087] Iterative computational and structure-guided design

[0088] The large-scale structural rearrangements that gB undergoes during viral entry, from the metastable prefusion state to the postfusion conformation, define the challenge of prefusion stabilization. The gB ectodomain consists of five subdomains, designated DI, Dll, Dill, DIV, and DV, followed by C-terminal transmembrane and cytoplasmic domains £45) (Fig. 1 a-b). Prior work on CMV gB and HSV1 gB revealed that DI, Dll, and DV undergo a ~180° reorientation relative to Dill and DIV during membrane fusion [32, 46). To stabilize the prefusion EBV gB ectodomain and prevent these conformational changes, we used ProteinMPNN £47) along with the AlphaFold2-predicted [48) structure of prefusion gB to identify over 200 stabilizing mutations (table 1). We experimentally evaluated these mutations via negative stain EM or cryoEM, and the resulting structural insights were used to iteratively guide subsequent design rounds (Fig. 1c).

[0089] Preventing gB domain II inversion

[0090] We hypothesized that deleting the loop between Dll and Dill would restrict Dll — and consequently gB as a whole — to the prefusion conformation, as the loop spans a distance of 20 A in the predicted prefusion conformation and 40 A in the postfusion conformation (Fig. 1a-b and fig. S1 a-b). This loop is not anticipated to play a significant role in fusion, aside from the presence of a polybasic protease cleavage site which enhances (but is not necessary for) gB- mediated fusion [49). Therefore, we deleted this 51 -residue loop between Dll and Dill (residues 402-452) and replaced it with an eight residue linker (herein gB-deli). CryoEM structure determination unexpectedly revealed that gB-deli adopted the postfusion conformation (Fig. 1 c,fig. S2a, and table S2), underscoring the need for further stabilization. For subsequent screening, we used 2D classification of negatively stained particles, as the postfusion conformation and deviations from it were readily distinguishable by this method.

[0091] Anticipating that a larger deletion might be required to restrict Dll movement, we subsequently designed gB-A, with a deletion of the DII-DIII loop along with that of the adjacent residues 394-401 (predicted to form an a-helix bridging Dll to the loop, herein the Dll C-terminal helix). This larger deletion indeed prevented gB refolding to the postfusion conformation (fig. S2a-c). However, DI and Dll did not adopt a defined orientation relative to the rest of the ectodomain, inconsistent with the prefusion conformation (fig. S3c), and were not resolved upon cryoEM analysis (Fig. 1c, fig. S2b, and table S2). Furthermore, in the gB-A structure, DV interacts with Dill in a postfusion-like manner, thereby occluding the site on Dill that would be occupied by Dll in the prefusion state (fig. S1 c). This observation explains the conformational heterogeneity of DI and Dll, as Dll cannot bind to its prefusion binding site on Dill whereas the 394-401 loop deletion prevents Dll from moving to its postfusion binding site on Dill. This finding is reminiscent of the structure of pseudorabies (PRV) gB with DI and Dll deleted, in which DV also binds Dill in a postfusion-like configuration £50) (fig. S1d). These results show that preventing Dll inversion alone is insufficient to prevent DV rearrangement.

[0092] Preventing gB domain V inversion

[0093] Since DV moves to the postfusion conformation in gB-A, we targeted DV to introduce additional prefusion-stabilizing mutations. To facilitate this, we used a gB ectodomain with a 402-452 deletion (gB-de), which still adopts a postfusion conformation similar to gB-deli but lacks the inserted linker. This was expected to impose greater conformational constraints on Dll than gB-deli, thereby lowering the threshold for detecting structural changes induced by stabilizing mutations, while still enabling discernment of the postfusion conformation by 2D classification of negatively stained particles. We identified an intermolecular disulfide bond between Dill and DV (Q527C-E634C) in the gB-de ectodomain that prevents the transition to the postfusion conformation as the two cysteine residues participating in the disulfide are ~5C)A apart in postfusion gB (Fig. 1 c, fig. S2c, fig. S3d). Non-reducing SDS-PAGE analysis of this ectodomain construct (gB-deD herein) was consistent with formation of the designed intermolecular disulfide bond (fig. S3d).

[0094] To probe the role of the loop deletion in stabilization, we reintroduced an eight residue loop-linker (herein gB-deliD) which resulted in dampened intermolecular disulfide bond formation and refolding to the postfusion state (fig. S3e). This indicates that the conformationalrestriction of Dll imposed by the loop deletion contributed to enhancing the stability of prefusion gB-deD, and that release of this restriction by reintroducing the loop-linker in gB-deliD favors the postfusion state. A cryoEM structure of gB-deD reveals prefusion-like interfaces between Dll and Dill as well as between DIV and DV, whereas the predicted DV C-terminal 3-helix coiled- coil was not resolved in the map (Fig. 1 c, fig. S2c, and table S2). In summary, we identified the Q527C-E634C intermolecular disulfide bond which participates in stabilizing the prefusion gB conformation.

[0095] A designed helical trimer stabilizing qB domain V

[0096] To further enhance the stability of EBV gB DV, we computationally designed a sevenresidue trimerization motif, designated oP3h, to promote formation of a C-terminal trimeric helical bundle present in the AF2-predicted EBV gB prefusion model and the full-length CMV gB prefusion structure (Fig. 2a). To validate the structural behavior of oP3h, including its repeat length-dependent properties, we initially tested it in the well-characterized SARS-CoV-2 S-2P ectodomain £57) by replacing the native foldon with oP3h variants (Fig. 2b). Recombinant production yields were inversely correlated with the number of oP3h repeats with three repeats leading to optimal balancing of production yields and proper folding to the prefusion state (Fig. 2b and fig. S4a-h). CryoEM structure determination of the S-2P ectodomain trimer with two variations of oP3h confirmed the predicted structure of oP3h and the accuracy of our designs (Fig. 2c, fig. S4i-j, and table S3), supporting its use as a generalizable trimerization domain (Fig. 2c).

[0097] We subsequently added oP3h in-frame with the EBV gB DV C-terminus using the gB- deliD ectodomain, including residue substitutions that gradually transition DV to oP3h (designated herein gB-deliDo) (Fig. 2d). Addition of the oP3h domain effectively promoted formation of the Q527C-E634C intermolecular disulfide bond, confirmed by non-reducing SDS- PAGE, consistent with DV adopting the prefusion conformation (Fig. 2d). Given that gB-deliDo remains locked in a prefusion-like conformation (fig. S3f), oP3h-mediated increased DV stabilization compensates for the enhanced conformational freedom of Dll resulting from the presence of the flexible loop-linker region.

[0098] Domain I qB hinge stabilization

[0099] Although gB-deD and gB-deliDo resist Dll and DIV refolding to the postfusion state, DI points laterally away from the rest of gB in the EM reconstructions of these two constructs, unlike the Dl-mediated clamping of DV observed for full-length prefusion HSV1 gB or CMV gB(Fig. 1 c, fig. S3g, and fig. S5). Comparing the relative orientation of DI and Dll in gB-deD with that of prefusion and postfusion CMV gB shows that DI has a postfusion-like orientation with hinge residues 290-295 adopting a helical conformation (fig. S5f-i). Analogous domain rotations between prefusion and postfusion states, including helical hinge formation, are observed in distantly related Rhabdovirus class III fusion proteins (52) (fig. S5j-k). The gB-deD DI orientation and Dl-hinge structure are similar to that observed in a recently described disulfide-stapled CMB gB structure (44), which also lacks a resolved DV C-terminal bundle, suggesting a similar degree of stabilization. We hypothesized that the postfusion Dl-hinge is rigidly locked, whereas the extended conformation observed for the prefusion Dl-hinge enables rotation — consistent with a metastable prefusion conformation and highly stable postfusion conformation. We therefore designed the A293P residue mutation to disfavor helix formation and stabilize the prefusion-like Dl-hinge conformation. Negative stain 2D classification and 3D reconstructions of gB-deliD or gB-deliDo harboring A293P (gB-deliDP and gB-deliDoP, respectively) show that DI is dynamic relative to the rest of the ectodomain in both prefusion and postfusion conformations (Fig. 2f lane 2-3 and fig. S3h-i). Similar to Dll and DV stabilization not preventing DI reorientation in gB-deliDo, the A293P substitution does not suppress large scale Dll postfusionlike conformational changes (observed via negative staining EM), nor does it greatly increase prefusion DV stability (observed via non-reducing SDS-PAGE analysis of Q527C-E634C bond formation) (fig. S3e-h). These data suggest that Dl-hinge conformational changes occur largely independently of Dll and DV rearrangement in the gB ectodomain, defining DI rotation as a third fusogenic conformational step.

[0100] gB stabilization via chimerization to a designed trimer

[0101] In the context of a membrane-anchored prefusion gB, the fusion loops are anticipated to be embedded in the viral envelope, thereby restricting DI motion. To mimic the envelope-bound orientation of DI, we genetically inserted the trimeric I5350A domain (53) in the DI fusion loops using flexible linkers. CryoEM analysis of the chimeric gB-deliDoP-l5350A revealed that the flexible linkers allowed I5350A to rotate markedly relative to gB, antithetical to the goal of using I5350A to guide the orientation of DI (Fig. 1 c, fig. S2d, fig. S3j, and table S2). To restrict this conformational freedom, I5350A was redesigned to harbor a structured linker connecting to DI (designated l5350A2h) — compatible with various linker lengths — and we evaluated this design when added to gB-deliDo and gB-deliDoP constructs (Fig. 2e-h, and fig. S3k-m). In the absence of the A293P mutation, we did not detect production of the gB-deliDo-l5350A2h ectodomain chimera (Fig. 2f), consistent with a postfusion-like Dl-hinge positioning DI in a mannerincompatible with the presence of the l5350A2h trimer. Conversely, in the presence of the A293P mutation, the EBV gB trimer folded as designed, enabling visualization of l5350A2h and of oP3h in the cryoEM structure of EBV gB-deliDoP-l5350A2h (Fig. 1 c, Fig. 2d,f,h, fig. S3k, fig. S6a and table S2). However, the l5350A2h chimera places DI too far apart to allow contact with DIV and DV (fig. S5l-o). As these interactions are observed in membrane-anchored prefusion HSV1 gB and CMV gB, we redesigned the l5350A2h helices connecting to DI with four fewer residues, thereby rotating and reorienting DI to enable DV clamping. This trimeric domain was designated l5350A2hN and enabled visualization of DI contacting DV and DIV in the cryoEM structure of EBV gB-deliDoP-l5350A2hN (Fig. 1 c, Fig. 2e-h, fig. S3n, fig. S5I-O, fig. S6b, and table S2).

[0102] The prefusion-stabilized EBV gB trimer

[0103] All key structural features of EBV gB-deliDoP-l5350A2hN align with the anticipated prefusion EBV gB architecture (Fig. 1 c-e, fig. S1 , fig. S5). Comparisons with full-length prefusion CMV gB and HSV1 gB reveal many similarities, including the orientation of Dll and the conformation of DV (Fig. 1 d,e). Additionally, a 45° rotation of DI relative to Dll distinguishes gB- deliDoP-l5350A2hN from postfusion gB, mirroring the structural shifts observed between prefusion and postfusion CMV and HSV1 gBs (Fig. 1e). This rotation enables DI to clamp DV, another defining feature of the prefusion state (Fig. 1 d,e). These findings validate the stabilization strategy, confirming that it preserves key structural features of prefusion gB.

[0104] EBV gB adopts two distinct prefusion conformations

[0105] Structural analysis revealed that EBV gB exists in at least two distinct prefusion conformations. In the gB-deliDoP-l5350A2hN construct, Dill central helices form a tightly packed, symmetrical coiled-coil structure, whereas in the gB-deliDoP-l5350A2h and gB-deD constructs these helices are partially splayed and arranged asymmetrically (Fig. 1f). Comparison of EBV gB-deliDoP-l5350A2h with EBV gB-deliDoP-l5350A2hN reveals these differences derive from the DII-DIII interaction interface, defining two distinct prefusion conformations: prefusion-1 (as seen in gB-deliDoP-l5350A2h) and prefusion-2 (as seen in gB- deliDoP-l5350A2hN) (Fig. 1 g).

[0106] In prefusion-2, Dll binds Dill with minimal contact to the W483 sidechain, allowing W483 to pack internally within the Dill coiled-coil bundle — resulting in a compact helical arrangement that closely resembles the Dill conformation in postfusion gB (Fig. 1 f,g). In contrast, in prefusion-1 , the Dill helices rotate and adopt a splayed arrangement, orienting W483 towardsthe interface with Dll (Fig. 1 f,g). Furthermore, the Dll C-terminal helix is perpendicular to the Dill helix in prefusion-2 but runs antiparallel to the Dill helix in prefusion-1 (Fig. 1 g).

[0107] The relative positioning of Dll and Dill in membrane-anchored CMV gB (43) resembles the prefusion-2 conformation, consistent with the bundled, symmetric Dill arrangement observed in EBV gB-deliDoP-l5350A2hN (Fig. 1 f,g). Conversely, full-length HSV1 gB-H516P (40) is consistent with the prefusion-1 conformation, showing a similar DII-DIII interface to EBV gB-deliDoP-l5350A2h. However, unlike the asymmetric splaying in EBV gB-deliDoP-l5350A2h, HSV1 gB-H516P exhibits symmetric splaying of the Dill helices (Fig. 1f,g). These structural similarities suggest that these prefusion-1 and -2 states represent conserved functional intermediates across Herpesvirus gBs.

[0108] A broadly generalizable qB stabilization strategy

[0109] The successful prefusion stabilization of the EBV gB ectodomain provides a framework for extending this approach to other Herpesvirus gBs. Given the structural conservation of the mutated regions in EBV gB with CMV and HSV1 gB, we hypothesized that a similar stabilization strategy could be applied to additional human-infecting Herpesviruses. To evaluate this possibility, we engineered CMV, HSV1 , HHV6B, and VZV gB-deliDoP-l5350A2h(N) ectodomains and evaluated their structure using electron microscopy analysis of negatively stained samples. Two-dimensional classification confirmed that CMV and HHV6B gB-deliDoP- l5350A2h(N) trimers adopt conformations resembling prefusion EBV gB, indicating successful stabilization (fig. S7a,c). Although production yields were low for VZV and HSV1 gB harboring these modifications, a subset of mutations — specifically gB-deliD — successfully enabled prefusion stabilization of the HSV1 gB and VZV gB ectodomains, which formed dimers of prefusion trimers (fig. S8a,c and Fig. 3a-n). Dimerization of trimers is primarily mediated by the fusion loops for HSV1 gB and by interactions between the fusion loops and DIV for VZV gB, thereby guiding the orientation of DI in the absence of l5350A2hN (Fig. 3a, j). VZV gB-deliD was a mixture of prefusion and postfusion conformations (fig. S8c), so to promote prefusion VZV gB stability, we introduced the L687K mutation into a gB-deliD background, as this residue is selectively buried in a hydrophobic pocket in the postfusion state, which would be unfavorable (fig. S1 b and fig. S8c). CryoEM reconstructions confirmed that HSV1 gB-deliD, VZV L687K gB- deliD, CMV gB-deliDoP-l5350A2hN, and HHV6B gB-deliDoP-l5350A2h adopt the prefusion conformation, as evidenced by the orientations of DI, Dll, and DV (Fig. 3). These findings support the broad applicability of these prefusion-stabilizing mutations across Herpesvirussubfamilies, providing a generalizable approach for stabilizing gB in the prefusion state (Fig. 3, fig. S7, fig. S8, and table S4).

[0110] Two distinct prefusion qB conformations across Herpesviruses

[0111] Similar to EBV gB, the prefusion stabilized gB trimers of HSV1 , VZV, CMV, and HHV6B can be categorized as prefusion-1 and prefusion-2 conformations based on examination of the interface between Dll and Dill (Fig. 3). Notably, cryoEM analysis revealed the presence of both prefusion ectodomain trimer conformations in the HSV1 gB-deliD dataset (Fig. 3a-i). One conformation closely resembles that of VZV L687K gB-deliD: the conserved tryptophan (HSV1 gB W528 or VZV gB W539) is centrally packed within the Dill coiled-coil bundle (Fig. 3h,m), and the Dll C-terminal helix is oriented roughly perpendicular to the Dill helices (Fig. 3i,n). These features are characteristic of the prefusion-2 conformation. The second HSV1 gB-deliD conformation resembles that of CMV gB-deliDoP-l5350A2hN and HHV6B gB-deliDoP-l5350A2h (Fig. 3b-e, p-s, u-x): the Dill helices are rotated such that the conserved tryptophan residue (HSV1 gB W528; CMV gB W505; or HHV6B gB W437) interacts extensively with Dll, instead of being buried in the core of the Dill helical bundle. This interaction forces an alternative packing arrangement of Dill along with an antiparallel configuration of the Dll C-terminal helix relative to Dill (Fig. 3d,e,r,s,w,x) — hallmarks of the prefusion-1 conformation.

[0112] Whereas prefusion-1 EBV gB and HSV1 gB exhibit splayed or partially splayed Dill helices, prefusion-1 CMV gB and HHV6B gB harbor bundled Dill helices (Fig. 3r,w). This superficial similarity to prefusion-2 highlights the need for DII-DIII interface analysis to distinguish the two states. The distinct DII-DIII interactions and the antiparallel orientation of the Dll C-terminal helix relative to Dill confirm that these are bona fide prefusion-1 states. Moreover, the bundled Dill helices in CMV and HHV6B gB prefusion-1 differ in the internal residue arrangement from those in prefusion-2, as the conserved tryptophan is buried within the Dill bundle of prefusion-2 structures but not for CMV and HHV6B gB prefusion-1 structures. These observations suggest that the key distinction between prefusion-1 and prefusion-2 is not the presence of Dill bundling per se, but the configuration of the DII-DIII interface. In prefusion- 2 — but not in prefusion-1 — Dll binds Dill in a way that permits the distinct prefusion-2-like Dill bundling.

[0113] HSV-1 gB-deliD folds as membrane-anchored HSV-1 qB

[0114] The prefusion-1 conformation of HSV1 gB-deliD resembles the membrane-anchored HSV-1 gB H516P model and resolves additional structural features that improve the fit to thelow resolution cryoET map (40). For instance, the Dll C-terminal helix is fully resolved, running antiparallel to Dill, a defining feature of the prefusion-1 conformation. Unique to HSV-1 gB among prefusion-1 ectodomains, Dill is remodeled and contacts the N141 glycan (Fig. 3e and fig. S9), which is also visible in the full-length gB-H516P cryoET map although it was not previously modeled (fig. S9). The HSV1 gB-deliD structure suggests that the H516P mutation likely reinforces an o-helix kink in Dill, stabilizing Dill remodeling in prefusion-1 (Fig. 3e, fig. S9d,e). Dill helix remodeling is unique to prefusion-1 of alpha-Herpesviruses as mutations corresponding to H516P improve VZV gB stability but fail to stabilize KSHV gB (41 , 42), and the Dill helices remain intact in prefusion-1 gB from CMV, HHV6B, and EBV (Fig. 1 g and Fig. 3e,i,n,s,x). Beyond Herpesviruses, remodeling of the Dill helix parallels the changes observed in prefusion RABV G, with the HSV-1 gB H516P mutation and the RABV G H270P mutation producing comparable effects (fig. S9f). However, the broader conformational distinctions that define the prefusion-1 and prefusion-2 states in Herpesvirus gB are not observed in Rhabdovirus G. In contrast, this region of Dill is helical in HSV1 gB prefusion-2 and postfusion states, and would be incompatible with a helix-destabilizing proline mutation (Fig. 3i, fig. S9d,e). This constraint explains why only the prefusion-1 conformation was observed for HSV1 gB- H516P. Therefore, our prefusion-1 HSV1 gB-deliD structure not only recapitulates key features of the native architecture of membrane-embedded gB but also explains previously ambiguous regions of density in the cryoET map, demonstrating that the applied prefusion stabilization strategy preserves the native gB conformation (fig. S9). Collectively, these findings demonstrate that purified gB ectodomains from alpha-, beta-, and gamma-Herpesviruses adopt two structurally distinct prefusion states, underscoring a flexible prefusion architecture conserved across the Herpesvirus family.

[0115] Resolving the N-terminal antigenic site

[0116] Our cryo-EM maps reveal the previously unresolved N-terminal region of HSV1 , VZV, and CMV gBs (Fig. 3c,g,l,y). Some VZV, CMV, and HSV1 neutralizing antibodies target these residues which have been proposed to participate in VZV gB-mediated membrane fusion (54- 57). In both HSV1 gB-deliD and VZV L687K gB-deliD structures, the N-terminal residues adopt a helical conformation and interact with the Dl-hinge through polar interactions involving a conserved arginine — HSV1 gB R98 and VZV gB R104 (Fig. 3c, g, I). In postfusion gB, a conserved leucine residue in Dll — HSV1 gB L409 and VZV gB L414 — interacts with the Dl- hinge instead of R98 or R104 (Fig. 3c, g, I). In the HSV1 gB postfusion structure and closely related alphaherpesvirus PRV, the N-terminal residues adopt a distinct conformation bound toDIV (58, 59). Conversely, in the prefusion CMV gB-deliDoP-l5350A2hN structure, the N-terminal residues adopt a conformation bound to DIV (Fig. 3y and fig. S7), consistent with unmodeled density observed in the full-length CMV gB map (fig. S7). These findings indicate distinct prefusion N-terminal residue conformations for alpha- and beta-Herpesviruses.

[0117] Convergent Mechanisms of qB-Directed Neutralizing Antibodies

[0118] To evaluate the antigenicity of prefusion-stabilized EBV gB, we assessed binding of the 3A3 and 3A5 neutralizing antibodies to various gB ectodomain constructs by ELISA (Fig. 4a and fig. S3o). Both antibodies bind comparably to prefusion and postfusion states, indicating that their epitope remains accessible irrespective of the gB conformation (Fig. 4a). We next determined a cryoEM structure of the prefusion-2 EBV gB-deliDoP-l5350A2hN construct in complex with the 3A3 and 3A5 neutralizing antibody Fab fragments revealing three 3A3 Fabs and three 3A5 Fabs bound to the gB trimer (Fig. 4b-c). Local refinement of the Fab-bound gB domains revealed previously unresolved features, including a fucosylated N-glycan at position N348 contributing ~20% of the gB surface buried at the interface with 3A3 (Fig. 4c), and docking of the 3A5 heavy chain CDR3 (including residues Y99 and 1101) to DIV (Fig. 4c). To contextualize these interactions, we aligned our prefusion stabilized gB ectodomain structures with previously described antibody-bound gB complexes (Fig. 4d— i).

[0119] The 3A5 epitope overlaps with that of the VZV-neutralizing antibody 93k on DIV (Fig. 4h). The 93k antibody has been proposed to block the gB N-terminal residues from engaging DIV (60, 61). Supporting this model, mutations in the N-terminal region significantly impair VZV replication (56). Structural alignment with the prefusion VZV gB-deliD structure shows that 93k would block binding of the N-terminal region to DIV but not to the Dl-hinge (Fig. 4h), thereby inhibiting membrane fusion (Fig. 4h). A related mechanism is indicated by the prefusion CMV gB-deliDoP-l5350A2hN structure, in which the N-terminus engages DIV (Fig. 3y, Fig. 4i). While the molecular basis for why the N-terminus binds DIV in prefusion CMV gB but engages the Dl- hinge in prefusion VZV and HSV-1 gB remains unclear, CMV-neutralizing antibodies such as 3- 25 and related TRL345-like antibodies are expected to block the N-terminus-DIV interaction by binding to the N-terminus (the AD-2 epitope in CMV gB), thereby mirroring the effects of the 3A5 and 93k mAbs through engagement of a distinct antigenic site. These findings support a convergent neutralization mechanism involving blockade of the N-terminus-DIV interaction, underscoring the importance of this interaction in regulating membrane fusion across Herpesviruses.

[0120] The 3A3 antibody, along with Dll-targeted antibodies against other Herpesviruses like D48 (HSV1) and SM5-1 (CMV), neutralize viral entry through mechanisms that remain only partially defined (Fig. 4d-f). While 3A3 reduces virus-cell binding (61), interference with gB conformational changes is also plausible given that SM5-1 dysregulates the fusogenic function of gB following attachment (62). While not studied for 3A3, Fab fragments of SM5-1 and D48 remain inhibitory, indicating that crosslinking neighboring domains is not required for neutralization (63, 64). Furthermore, co-expression of D48 with gB does not stabilize the prefusion conformation (64). When modeled on prefusion-1 gB structures, each antibody exhibits predicted sterically clashes — 3A3 with the N395 glycan of the protomer its recognizes and a neighboring gB protomer, D48 with the restructured Dill and Dll C-terminal helix positioning, and SM5-1 with the repositioned Dll and N409 glycan from the neighboring protomer — suggesting reduced accessibility to this conformation (Fig. 4d-i). In contrast, all three antibodies are structurally compatible with prefusion-2 gB — consistent with structural data showing 3A3 bound to prefusion-2 EBV gB (Fig. 4b, c) and SM5-1 co-purified with full-length prefusion-2 CMV gB (43). Notably, 3A3 binds both EBV gB-deliDoP-l5350A2h and EBV gB- deliDoP-l5350A2hN (Fig. 4a), which adopt prefusion-1 and prefusion-2 conformations, respectively, indicating that 3A3 can access its epitope across multiple conformational states. However, the modeled clashes in prefusion-1 indicate that such binding likely destabilizes this state. Dll-directed antibodies such as 3A3, D48, and SM5-1 may bind transiently to prefusion-1 , thereby biasing the conformational equilibrium toward prefusion-2, possibly triggering premature triggering or interfering with the refolding trajectory of gB and disrupting membrane fusion. The convergence of neutralizing antibodies on the N-terminus-DIV interface and the prefusion-1 conformation underscores the functional importance and vulnerability of these epitopes.

[0121] qB-mediated Membrane Fusion

[0122] The iterative stabilization of prefusion gB offers critical insight into the mechanism of gB- mediated membrane fusion (Fig. 5). For class III fusion proteins like gB, it has long been proposed that the fusion loops insert into the host membrane first, followed by conformational rearrangements of the domains anchored to the viral membrane (32, 42, 65-67). This model predicts that Dll inverts before DV (Fig. 5a-c). Indeed, the postfusion DV interface on Dill overlaps with the prefusion Dll-Dll I interface (fig. S1 a-b), implying that Dll must disengage from Dill prior to DV adopting its postfusion conformation. However, our data show that DV transitions to a postfusion conformation even though Dll has not yet adopted its final postfusion configuration in gB-A (Fig. 1 c). Furthermore, when DV is stabilized — via interprotomer disulfidebonding and genetic chimerization with oP3h — Dll remains locked in the prefusion state, despite minimal restraints (gB-deliDo; fig. S3f-g). These results are consistent with Dll rearrangement to its final postfusion conformation depending on prior rearrangement of DV, which is itself dependent on Dll release from Dill. Interactions between Dll and DV observed in prefusion structures (Fig. 1 g and Fig. 3e,i,n,s,x) provide a structural basis for coordinated release and rearrangement of these domains. In other words, while DV rearrangement requires Dll release, our data indicate that DV transitions to its postfusion conformation before Dll does, reversing the order proposed in prior models. This is consistent with the domain layering seen in all postfusion gB structures and distantly related class III fusion proteins, where DI and Dll wrap around DV (Fig. 5h and fig. S10). In this case, EBV gB-A may represents a fusion intermediate in which Dll has disengaged from Dill permitting DV to bind Dill in its postfusion conformation (Fig. 5g).

[0123] Given that DV is surrounded by DI in both prefusion and postfusion states (fig. S5), the conformational changes leading to membrane fusion likely also involve the release of prefusion DV by DI. In gB-deD and gB-deliDo ectodomains, DI adopts a conformation that releases its clamp on DV, with DI pointing laterally away from the rest of the gB trimer (Fig. 1 c, fig. S3g, and fig. S6). This DI reorientation would pull the fusion loops out of the viral envelope (Fig. 5e), a critical early step in membrane fusion (32). Therefore, gB-deliDo may represent another conformational intermediate of the gB-mediated fusion reaction (Fig. 1 e and Fig. 5e).

[0124] We observed two distinct conformations of prefusion-stabilized gB ectodomains, designated prefusion-1 and prefusion-2 (Fig. 1 g and Fig. 3e,i,n,s,x). Because HSV-1 gB-deliD adopts both states, this transition may be reversible (Fig. 3b, f and Fig. 5d). Structural comparisons suggest that gB progresses from prefusion-1 to prefusion-2 prior to adopting the postfusion conformation as the Dill central helices of prefusion-2 — but not prefusion-1 — closely resemble those in the postfusion structure. In the postfusion state, DV binds along the periphery and above the bundled Dill helices (fig. S1 b), indicating that this bundling occurs before DV refolds. In prefusion-1 , this postfusion-like Dill helical bundle is absent, obstructing DV refolding (Fig. 1 g and Fig. 3e,s,x). The shifting of Dll upon Dill to its prefusion-2 conformation would enable Dill helix bundling, thereby priming DV for refolding (Fig. 1 g and Fig. 3i,n). Consistent with this model, the HSV-1 gB H516P mutation — which disrupts the Dill o-helix and consequently Dill helix bundling characteristic of the prefusion-2 and postfusion state — impairs the transition to postfusion gB (40). These analyses support a sequence in which prefusion-1 precedes prefusion-2.

[0125] These findings support a revised model of Herpesvirus membrane fusion in which gB initially adopts the prefusion-1 conformation (Fig. 5d,e). Transition to the prefusion-2 conformation involves Dill adopting a bundled conformation of its central helices (Fig. 5e). Conformational remodeling of the Dl-hinge reorients DI, dislodging the fusion loops from the viral envelope and releasing DV (Fig. 5f). Dll release from Dill then enables DV to refold to its postfusion conformation. As DI is linked to Dll, this displacement repositions DI such that its fusion loops can insert in the host cell membrane (Fig. 5g). Finally, reassociation of DI / DII with the rest of the gB ectodomain in the postfusion state brings the viral and host membranes together, promoting membrane fusion (Fig. 5h). Reassociation of DI and Dll with the gB core leads to burial of approximately 12,000 A2 of surface area per trimer, similar to the 11 ,000 A2 surface area buried by HR1-HR2 interactions in coronavirus spikes (68, 69), suggesting that this conformational change can release sufficient free energy to promote membrane fusion. Collectively, these insights define a sequential refolding pathway for gB, culminating in membrane fusion.

[0126] Description of Supplemental Figures S1-S10

[0127] FIGS. S1a-d show DI, Dll, and DV conformational changes between the Alphafold2 predicted prefusion EBV gB model (a), postfusion EBV gB structure (b), EBV gB-A structure (c), and PRV gB with DI and Dll deletion structure (d), aligned relative to the central helices of Dill. Panels a-b show a gB trimer (left) or a single protomer (right). Domain I (DI, blue), domain II (Dll, green), domain III (Dill, yellow), domain IV (DIV, orange), domain V (DV, red). In the prefusion structure prediction, the loop between Dll and Dill as well as the N-terminus are rendered black whereas regions not included in the predicted structure are shown as dotted black lines. The distance spanned by the DII-DIII loop between rigidly structured elements is ~20 A in the prefusion conformation and ~40 A in the postfusion conformation. Insets show a cross-section through Dill, highlighting the relative positions of Dll and DV binding to Dill in the prefusion and postfusion conformations. In the postfusion conformation, DV binds the bundled or closed central helices of Dill and occupies the Dll prefusion binding site. Residue T630 in EBV gB (L687 in VZV gB) is labelled.

[0128] FIGS. S2a-d show cryoEM data processing flowcharts of EBV gB-deli (a), gB-A (b), gB- deliD (c), and gB-deliDoP-l5350A (d) datasets. Representative electron micrograph and 2D class averages are shown at the start of the flowchart. At the end of the flowchart is the gold- standard Fourier shell correlation curve (0.143 cutoff: horizontal dashed line), angulardistribution of particles (heat map), and local resolution estimation plotted on the maps (calculated in CryoSPARC). CTF: contrast transfer function; NUR: non-uniform refinement.

[0129] FIGS. S3a-o show non-reducing SDS-PAGE, negative-stain EM, and ELISA binding data for EBV gB ectodomain constructs, (a-f, h-n) A summary of the construct is listed (left), as well as non-reducing SDS-PAGE (middle), and negative stain 2D classes (right). Included are variants of the EBV gB-deliDoP-l5350A2h construct (k) with one (I) or four (m) additional heptad repeats inserted between the l5350A2h trimer and domain I (DI) of gB. (g) Left, EBV gB-deliDo negative stain map reconstruction with the cryoEM structure of EBV gB-deD (black cartoon) rigidly fit into this map. Center, EBV gB-deliDP negative stain map reconstruction. Right, EBV gB-deli negative stain map reconstruction. Dashed lines highlight the DI orientation in gB-deD, gB-deliDo, and gB-deli, and the apparent movement of DI in EBV gB-deliDP. (o) ELISA doseresponse curves for binding of the 3A3 and 3A5 antibodies to the indicated immobilized EBV gB ectodomain constructs.

[0130] FIGS. S4a-j show a summary of negative stain EM and cryoEM data for SARS-CoV-2 S- 2P - oP3h. (a-h) negative stain micrographs (top) and 2D classifications (bottom) of SARS- CoV-2 S-2P with the indicated C-terminal trimerization domains, (i-j) CryoEM data processing flowcharts of SARS-CoV-2 S-2P - oP3h (3X RINEIER) (i) and SARS-CoV-2 S-2P - oP3h (4X RINAIET) (j) datasets. At the start of the flowchart, representative electron micrograph and 2D class averages of SARS-CoV-2 S ectodomains embedded in vitreous ice are shown. At the end of the flowchart is the gold-standard Fourier shell correlation curve (0.143 cutoff: horizontal dashed line), angular distribution of particles (heat map), and local resolution estimation plotted on the maps (calculated in CryoSPARC). At the bottom, the local refinement strategies employed for improving map resolution are also shown. CTF: contrast transfer function; NUR: non-uniform refinement.

[0131] FIGS. S5a-o show a comparison of gB prefusion and postfusion hinge conformations with gB-deD. Domain I (DI, blue), domain II (Dll, green), domain III (Dill, yellow), domain IV (DIV, orange), and domain V (DV, red) are shown as ribbons, (a-e) Side-view comparisons of prefusion CMV gB (a), prefusion HSV1 gB (b), the EBV gB-deD structure determined here (c), prefusion VSV G (d), and prefusion RABV G (e) highlighting how prefusion structures clamp or bind DV (Cter in VSV and RABV G). (f-k) Comparisons of Dl-hinge structure showing DI and Dll (or FD and PHD for VSV) from prefusion CMV gB (f), prefusion HSV1 gB (g), the gB-deD structure determined here (h), postfusion EBV gB (i), prefusion VSV G (j), and postfusion VSV G (k). Insets show zoomed-in view of backbone atoms (side chains are omitted for clarity) withdisulfide bonds and prolines shown to highlight the restructuring of prefusion and postfusion Dl- hinge residues, (l-o) Comparisons of gB cross-sections showing DI in relation to DIV and DV, including EBV gB-deliDoP-l5350A2hN (I), EBV gB-deliDoP-l5350A2h (m), prefusion CMV gB (n), and prefusion HSV1 gB (o).

[0132] FIGS. S6a-c show cryoEM data processing flowcharts of EBV gB-deliDoP-l5350A2h (a) EBV gB-deliDoP-l5350A2hN (b), and EBV gB-deliDoP-l5350A2hN bound to 3A3 and 3A5 (c) datasets. Representative electron micrograph and 2D class averages are shown at the start of the flowchart. At the end of the flowchart is the gold-standard Fourier shell correlation curve (0.143 cutoff: horizontal dashed line), angular distribution of particles (heat map), and local resolution estimation plotted on the maps (calculated in CryoSPARC). This flowchart additionally has a summary of local refinement strategies employed for improving local map resolution. CTF: contrast transfer function; NUR: non-uniform refinement.

[0133] FIGS.S7a-e show a summary of negative stain EM and cryoEM data for CMV gB- deliDoP-l5350A2hN (a-b) and HHV6B gB-deliDoP-l5350A2h (c-d). (a,c) non-reducing SDS- PAGE (left) and 2D classification (right) of negatively stained ectodomains. (b,d) CryoEM data processing flowcharts of CMV gB-deliDoP-l5350A2hN (b), and HHV6B gB-deliDoP-l5350A2h(d) datasets. At the start of the flowchart, representative electron micrograph and 2D class averages of gB ectodomain particles embedded in vitreous ice are shown. At the end of the flowchart is the gold-standard Fourier shell correlation curve (0.143 cutoff: horizontal dashed line), angular distribution of particles (heat map), and local resolution estimation plotted on the maps (calculated in CryoSPARC). At the bottom, the local refinement strategies employed for improving map resolution are also shown. The Alphafold3 predicted structure of the N-terminus of CMV is shown beside the refined model for comparison^ 15). The local map of full length prefusion CMV gB (shown as black mesh) with its model (PDB 7kdp) are shown with the N- terminal residues L76-N85 of CMV gB-deliDoP-l5350A2hN for comparison (map was blurred by applying a B-factor of +70 A2). CTF: contrast transfer function; NUR: non-uniform refinement.(e) Schematic of mutations used for CMV gB-deliDoP-l5350A2hN and HHV6B gB-deliDoP- l5350A2h.

[0134] FIGS.S8a-e show a summary of negative stain EM and cryoEM data for HSV1 and VZV gB ectodomains. (a,c) non-reducing SDS-PAGE (left) and 2D classification of negatively stained (right) HSV1 and VZV gB ectodomains. 2D classes resembling postfusion conformations are marked with blue asterisks; those resembling prefusion conformations are indicated with yellow daggers. (b,d) CryoEM data processing flowcharts of HSV1 gB-deliD (b), and VZV-deliD +L687K (d) datasets. At the start of the flowchart, representative electron micrograph and 2D class averages of gB ectodomain particles embedded in vitreous ice are shown. At the end of the flowchart is the gold-standard Fourier shell correlation curve (0.143 cutoff: horizontal dashed line), angular distribution of particles (heat map), and local resolution estimation plotted on the maps (calculated in CryoSPARC). At the bottom, the local refinement strategies employed for improving map resolution are also shown for HSV-1 gB-deliD. CTF: contrast transfer function; NUR: non-uniform refinement, (e) Schematic of mutations used for VZV gB-deliD + L687K and HSV1 gB-deliD.

[0135] FIGS.S9a-f show a structural comparison of HSV1 gB structures with cryo-ET data, (a) Ribbon representations of full-length, membrane-embedded HSV1 gB-H516P(40) (orange) and HSV1 gB-deliD prefusion-1 (blue) aligned to highlight structural similarities and differences, (b-c) Cryo-ET subtomogram averaged map of HSV1 gB-H516P(40) low-pass filtered at 9 A resolution with docked HSV1 gB-H516P pseudo-atomic model (b) and HSV1 gB-deliD prefusion-1 structure (c). Only part of domains II (Dll, green) and III (Dill, yellow) are shown for clarity. The HSV1 gB-H516P model fits well in the sharpened cryo-ET map although unmodeled densities (*) suggest additional structural elements. These densities are better accounted for by the HSV1 gB-deliD prefusion-1 structure, (d-e) Close-up comparison of Dill in HSV1 gB-deliD prefusion-1 and -2 conformations, shown as cartoons (left inset) or sticks with the sharpened map (right). In the prefusion-2 conformation, the H516 backbone amide forms a hydrogen bond with the H512 carbonyl oxygen (dashed line). In contrast, in the prefusion-1 conformation, the H516 backbone amide (hydrogen bond donor) remains unsatisfied, as there is a kink in the Dill helix (**).Because the side chain of proline residues form a covalent interaction with their main-chain nitrogen, the H516P mutation prevents hydrogen bonding at this site, which would destabilize the prefusion-2 conformation while stabilizing prefusion-1 , (f) Comparison of HSV-1 Dll and Dill with the equivalent domains in RABV G, noting the position of H270P (RABV G) and H516 (HSV1).

[0136] FIGS. S10a-h show a comparison of class III fusion protein postfusion structures. Postfusion structures shown as ribbons for EBV gB (a), CMV gB (b), VZV gB (c), HSV1 gB (d), PRV gB (e), Thogotovirus Gp (f), Baculovirus GP64 (g), and VSV G (h). Domain I (DI, blue), domain II (Dll, green), domain III (Dill, yellow), domain IV (DIV, orange), and domain V (DV, red) are shown as ribbons. Insets show a zoomed-in view highlighting DI and DV (FD and Cter for VSV).

[0137] DISCUSSION

[0138] We leveraged machine-learning- and structure-guided approaches to stabilize the prefusion conformation of the EBV gB ectodomain, paving the way for its evaluation as a vaccine candidate. During this process, we stabilized conformational intermediates of EBV gB, unveiling a mechanism of membrane fusion that may extend to all class III viral fusion proteins. In parallel, we designed trimeric proteins to stabilize the metastable prefusion gB, demonstrating applicability to other fusion glycoproteins, as shown with SARS-CoV-2 S. Finally, we ported the designed prefusion-stabilizing mutations to CMV, HHV6B, HSV1 , and VZV gBs, showcasing the generalizability of our approach and revealing conserved architectural principles among these viruses. The tools developed here provide a molecular blueprint for designing next generation Herpesvirus vaccines and antivirals to address major unresolved public health challenges.

[0139] The conformational snapshots obtained define an alternative model of gB-mediated fusion involving movement between two prefusion conformations, turning of DI, followed by DV reorientation, and subsequent rotation of Dll. Multiple checkpoints are a common theme across viral entry proteins, as they must exhibit exquisite control of the timing and specificity of the irreversible fusogenic conformational changes they undergo (70). The splayed prefusion-1 and bundled prefusion-2 gB conformations described here are reminiscent of those for the HIV-1 Env class I fusion protein, which exhibits similar structural rearrangements in entry intermediates induced by host-receptor binding (71-74). Reminiscent of DI rotation in gB, domain rotations exposing the fusion loops triggered by low pH were observed for the class II Flavivirus fusion proteins, and recently for the class III Baculovirus fusion protein (75-82). By analogy, these gB conformational changes may correspond to critical fusogenic triggers.

[0140] Our structural analysis indicates that Dl-hinge restructuring could be a key player in fusogenic triggering. Whereas the N-terminus of VZV and HSV1 gB binds to the prefusion Dl- hinge, an alternate DIV-bound conformation was captured for HSV1 postfusion gB at low pH, a known fusion trigger for this virus (59, 83). Therefore, we hypothesize that the alphaHerpesvirus gB N-terminal residues maintain the Dl-hinge in its prefusion state until fusogenic triggering. Although the N-terminal residues are unresolved in prefusion EBV gB, the DIV-bound conformation in CMV gB reveals that N-terminal positioning and function likely varies across viruses or conditions, with distinct functional roles yet to be defined.

[0141] An additional possible role for the N-terminal residues may be to mask epitopes targeted by neutralizing antibodies. For instance, when the N-terminus engages DIV, it would occlude epitopes recognized by 3A5 and 93k antibodies. Epitope masking may also be one selectivepressure shaping the existence of two prefusion conformations, as antibodies such as 3A3, D48, and SM5-1 selectively clash with prefusion-1. Careful regulation of transitions between the prefusion-1 , prefusion-2, and N-terminal conformations may enable immune evasion analogous to the conformational masking and glycan shielding strategies used by coronaviruses to promote immune evasion (84-86). Inversely, engineering increased access to these epitopes may enhance the potency of vaccine-elicited antibody responses. For example, removal of the HSV1 gB Dll N141 glycan with the N141 Q mutation increases viral neutralization with seropositive human sera indicating that its removal reveals a key neutralizing epitope (87). As the N141 glycan forms a core contact in the prefusion-1 conformation, this phenotype likely reflects destabilization of prefusion-1 and a shift toward prefusion-2, further implicating prefusion-1 in immune evasion.

[0142] The generalizability of the prefusion-stabilizing mutations identified here to alpha-, beta-, and gamma-Herpesvirus gB trimers indicates that the underlying molecular mechanism of gB- mediated fusion is conserved across Herpesviruses. The observed conservation of structural features among other class III fusion proteins is consistent with a shared fusion mechanism, supporting the potential application of this strategy to more distantly related viral families, analogous to the prefusion-stabilizing proline substitutions in the central helix of class I fusion proteins (34, 87, 88). Finally, these data provide a blueprint to design inhibitors targeting gB and isolate prefusion-specific monoclonal antibodies, offering a path toward next-generation therapeutics against the most pervasive chronically infecting human pathogens.

[0143] References

[0144] 1. K. Bjornevik, et al. Science 375. 2022. 296-301.

[0145] 2. Y. Wong, et al. J. Cancer Res. Clin. Oncol. 2022. 148, 31-46.

[0146] 3. B. Damania, et al. Cell 185. 2022. 3652-3670.

[0147] 4. M. J. Cannon. J. Clin. Virol. 2009. 46 Suppl 4, S6-10.

[0148] 5. R. F. Itzhaki. Vaccines (Basel). 2021. 9.

[0149] 6. S. Shah, et al. Brain Behav. 2024. 14, e3415.

[0150] 7. M. Eyting, et al. Nature. 2025. doi: 10.1038 / s41586-025-08800-x.

[0151] 8. M. K. Smatti, et al. Front. Oncol. 2018. 8, 211.

[0152] 9. M. J. Cannon, et al. Rev. Med. Virol. 2010. 20, 202-213.

[0153] 10. H. Chemaitelly, et al. PLoS One. 2019. 14 e0214151.

[0154] 11. M. A. Reynolds, et al. Public Health Rep. 2010. 125, 860-869.

[0155] 12. M. Vazquez, et al. JAMA 291 . 2004. 851-855.

[0156] 13. M. N. Oxman, et al. N. Engl. J. Med. 2005. 352, 2271-2284.

[0157] 14. S. Lopez, et al. MMWR Morb. Mortal. Wkly. Rep. 2016. 65, 902-905.

[0158] 15. N. Lecrenier, et al. Expert Rev. Vaccines. 2018. 17, 619-634.

[0159] 16. C. Tricco, et al. BMJ. 2018. 363, k4029.

[0160] 17. L. Cunningham, et al. N. Engl. J. Med. 2016. 375, 1019-1032.

[0161] 18. W. D. Harshbarger, et al. Vaccines (Basel). 2024. 12, 1111.

[0162] 19. F. C. Bender, et al. J. Virol. 2007. 81 , 3827-3841 .

[0163] 20. A. Bootz, et al. PLoS Pathog. 2017. 13, e1006601.

[0164] 21. T. M. Cairns, et al. B. J. Virol. 2014. 88, 2677-2689.

[0165] 22. J. Hong, et al. Front. Immunol. 2022. 13, 920467.

[0166] 23. J. Snijder, et al. Immunity. 2018. 48, 799-811.e9.

[0167] 24. C. Sun, et al. Cell Host Microbe. 2023. 31 , 1882-1897.e10.

[0168] 25. W. Bu, et al. Immunity. 2019. 50, 1305-1316.e6.

[0169] 26. J. Parsons, et al. Commun Biol. 2022. 5, 387.

[0170] 27. Q.-Y. Zhu, et al. Nat. Commun. 2021. 12, 6624.

[0171] 28. E. Fouts, et al. Proc. Natl. Acad. Sci. U. S. A. 2014. 111 , 8209-8214.

[0172] 29. D. I. Bernstein, et al. Vaccine. 2016. 34, 313-319.

[0173] 30. R. F. Pass, et al. N. Engl. J. Med. 2009. 360, 1191-1199.

[0174] 31. S. A. Connolly, et al. Nat. Rev. Microbiol. 2021. 19, 110-121.

[0175] 32. Z. Si, et al. PLoS Pathog. 2018. 14, e1007452.

[0176] 33. W.-H. Chen, et al. Immunity. 2022. 55, 2135-2148.e6.

[0177] 34. R. W. Sanders, et al. Cell Host Microbe. 2021. 29, 327-333.

[0178] 35. J. E. Bowen, et al. Sci Immunol. 2022. 7, eadf1421.

[0179] 36. J. S. McLellan, et al. Science. 2013. 342, 592-598.

[0180] 37. E. Vitu, et al. J. Mol. Biol. 2013. 425, 2056-2071.

[0181] 38. J. L. Silverman, et al. J. Virol. 2010. 84, 2001-2012.

[0182] 39. J. Zhen, et al. J. Virol. 2024. 98, eO119424.

[0183] 40. B. Vollmer, et al. Sci Adv 2020. 6.

[0184] 41. M. Zhou, et al. Nat. Commun. 2023. 14, 7958.

[0185] 42. F. Ito, et al. J. Virol. 2025. 99, eO153324.

[0186] 43. Y. Liu, et al. Sci Adv. 2021. 7.

[0187] 44. M. R. Sponholtz, et al. PNAS. 2024. p. 2024.02.10.579772.

[0188] 45. R. S. Cooper, et al. Viruses. 2015. 7, 6552-6569.

[0189] 46. S. C. Harrison. Nat. Struct. Mol. Biol. 2008. 15, 690-698.

[0190] 47. J. Dauparas, et al. Science. 2022. 378, 49-56.

[0191] 48. J. Jumper, et al. Nature. 2021. 596, 583-589.

[0192] 49. J. Sorem, et al. J. Gen. Virol. 2009. 90, 591-595.

[0193] 50. X. Li, et al. PLoS Pathog. 2017. 13, e1006777.

[0194] 51. C. Walls, et al. Cell. 2020. 181 , 281-292.e6.

[0195] 52. F. Yang, et al. Cell Host Microbe. 2020. 27, 441-453.e7.

[0196] 53. J. B. Bale, et al. Science. 2016. 353, 389-394.

[0197] 54. K. Manley, et al. Cell Host Microbe. 2011. 10, 197-209.

[0198] 55. M. Ohlin, et al. J. Virol. 1993. 67, 703-710.

[0199] 56. S. L. Oliver, et al. PLoS Pathog. 2021. 17, e1008961.

[0200] 57. S. Chandramouli, et al. Nat. Commun. 2015. 6, 8176.

[0201] 58. M. Vallbracht, et al. J. Virol. 2018. 92.

[0202] 59. D. Stampfer, et al. J. Virol. 2010. 84, 12924-12933.

[0203] 60. S. L. Oliver, et al. Nat. Commun. 2020. 11 , 4141.

[0204] 61. X. Zhang, et al. Proc. Natl. Acad. Sci. U. S. A. 2022. 119, e2202371119.

[0205] 62. S. Pbtzsch, et al. PLoS Pathog. 2011. 7, e1002172.

[0206] 63. N. Spindler, et al. PLoS Pathog. 2014. 10, e1004377.

[0207] 64. C. Sun, et al. HLife. 2024. 2, 141-146.

[0208] 65. S. Roche, et al. Science. 2007. 315, 843-848.

[0209] 66. J. M. White, et al. Biochem. Mol. Biol. 2008. 43, 189-219.

[0210] 67. S. Roche, et al. Science. 2006. 313, 187-191.

[0211] 68. V. M. Supekar, et al. Proc. Natl. Acad. Sci. U. S. A. 2004. 101 , 17958-17963.

[0212] 69. C. Walls, et al. Proc. Natl. Acad. Sci. U. S. A. 2017. 114, 11157-11162.

[0213] 70. M. S. Maginnis, et al. J. Mol. Biol. 2018. 430, 2590-2611 .

[0214] 71 . X. Ma, et al. Elite. 2018. 7.

[0215] 72. G. Ozorowski, et al. Nature. 2017. 547, 360-363.

[0216] 73. D. Lyumkis, et al. Science. 2013. 342, 1484-1490.

[0217] 74. J. P. Julien, et al. Science. 2013. 342, 1477-1483.

[0218] 75. Y. Modis, et al. Nature. 2004. 427, 313-319.

[0219] 76. F. A. Rey, et al. Nature. 1995. 375, 291-298.

[0220] 77. I. Medits, et al. EMBO Rep. 2020. 21 , e50069.

[0221] 78. K. Stiasny, et al. PLoS Pathog. 2007. 3, e20.

[0222] 79. S. Bressanelli, et al. EMBO J. 2004. 23, 728-738.

[0223] 80. V. A. Kostyuchenko, et al. Nature. 2016. 533, 425-428.

[0224] 81. D. Sirohi, et al. Science. 2016. 352, 467-470.

[0225] 82. J. Guo, et al. Nat. Commun. 2024. 15, 7668.

[0226] 83. V. Nicola, et al. Traffic. 2016. 17, 965-975.

[0227] 84. C. Walls, et al. Nat. Struct. Mol. Biol. 2016. 23, 899-905.

[0228] 85. Y.-J. Park, et al. Cell. 2025. 188, 1711-1728.e21 .

[0229] 86. C. Walls, et al. Cell. 2019. 176, 1026-1039.e15.

[0230] 87. A. Fukui, et al. MBio. 2023. 14, e0099223.

[0231] 88. M. Mirdita, et al. Nat. Methods. 2022. 19, 679-682.

[0232] 89. D. B. Craig, et al. BMC Bioinformatics. 2013. 14, 346.

[0233] 90. M. Backovic, et al. Proc. Natl. Acad. Sci. U. S. A. 2009. 106, 2880-2885.

[0234] 91. L. Watson, et al. Nature. 2023. 620, 1089-1100.

[0235] 92. C. Suloway, et al. J. Struct. Biol. 2005. 151 , 41-60.

[0236] 93. A. Punjani, et al. Nat. Methods. 2017. 14, 290-296.

[0237] 94. C. J. Russo, et al. Science. 2014. 346, 1377-1380.

[0238] 95. D. Tegunov, et al. Nat. Methods. 2019. 16, 1146-1152.

[0239] 96. T. Bepler, et al. Nat. Methods. 2019. 16, 1153-1160.

[0240] 97. J. Zivanov, et al. Elife. 2018. 7.

[0241] 98. J. Zivanov, et al. lUCrJ. 2019. 6, 5-17.

[0242] 99. A. Punjani, et al. Nat. Methods. 2020. 17, 1214-1221.

[0243] 100. P. B. Rosenthal, et al. J. Mol. Biol. 2003. 333, 721-745.

[0244] 101. S. Chen, et al. Ultramicroscopy. 2013. 135, 24-35.

[0245] 102. E. F. Pettersen, et al. J. Comput. Chem. 2004. 25, 1605-1612.

[0246] 103. T. Nakane, et al. Methods Mol. Biol. 2021. 2215, 145-160.

[0247] 104. D. Kimanius, et al. Nat. Methods. 2024. 21 , 1216-1221.

[0248] 105. J. He, et al. Nat. Commun. 2023. 14, 3217.

[0249] 106. P. Emsley, et al. Acta Crystallogr D Struct Biol. 2010. 66, 486-501.

[0250] 107. A. Casahal, et al. Protein Sci. 2020. 29, 1069-1078.

[0251] 108. T. I. Croll. Acta Crystallogr D Struct Biol. 2018. 74, 519-530.

[0252] 109. T. D. Goddard, et al. Protein Sci. 2018. 27, 14-25.

[0253] 110. R. Y. Wang, et al. Elife. 2016. 5.

[0254] 111. B. Frenz, et al. Structure. 2019. 27, 134-139.e3.

[0255] 112. V. B. Chen, et al. Acta Crystallogr D Struct Biol. 2010. 66, 12-21.

[0256] 113. B. A. Barad, et al. Nat. Methods. 2015. 12, 943-946.

[0257] 114. D. Liebschner, et al. Acta Crystallogr D Struct Biol. 2019. 75, 861-877.

[0258] 115. J. Abramson, et al. Nature. 2024. doi: 10.1038 / s41586-024-07487-w.

[0259] Example 2: Assessing Immunogenicity

[0260] Functional validation as an immunogen of the prefusion stabilized gB embodiments described herein can be assessed in both animal models and in vitro assays. Animal models can be employed to evaluate the immunogenicity of a composition (see, e.g., Reguraman N, etal. Assessing the Efficacy of VLP-Based Vaccine against Epstein-Barr Virus Using a Rabbit Model. Vaccines (Basel). 2023 Feb 24; 11 (3):540. doi: 10.3390 / vaccinesl 1030540.). Huang and Yasuda have described animal models for EBV (Front. Immunol., 23 February 2021 , Sec. B Cell Biology, Volume 12 - 2021).

[0261] The prefusion stabilized gB embodiments described herein can also be used to isolate gB-directed B-cells and antibodies from the blood of previously infected or vaccinated people and animals; these antibodies, or sera depleted for gB-directed antibodies, can then be used with in vitro fusion assays to establish the capacity of the prefusion stabilized gB embodiments described herein to elicit neutralizing antibodies (e.g., L. Zhong, et al. A cocktail nanovaccine targeting key entry glycoproteins elicits high neutralizing antibody levels against EBV infection. Nat Commun. 2024 Jun 21 ; 15(1 ):5310). In vitro fusion assays are known (Fan Q, et al. Natural Selection of Glycoprotein B Mutations That Rescue the Small-Plaque Phenotype of a Fusion- Impaired Herpes Simplex Virus Mutant. mBio. 2018 Oct 16;9(5):e01948-18. doi: 10.1128 / mBio.01948-18.).

[0262] Throughout this application various publications are referenced. The disclosures of these publications in their entireties are hereby incorporated by reference into this application in order to describe more fully the state of the art to which this invention pertains.

[0263] Those skilled in the art will appreciate that the conceptions and specific embodiments disclosed in the foregoing description may be readily utilized as a basis for modifying or designing other embodiments for carrying out the same purposes of the present invention. Those skilled in the art will also appreciate that such equivalent embodiments do not depart from the spirit and scope of the invention as set forth in the appended claims.

Claims

1. What is claimed is:

1. A Herpesviridae glycoprotein B (gB) polypeptide comprising a modified ectodomain, wherein the ectodomain comprises at least one of the following modifications:(a) deletion of residues 402-452 or 394-452 of SEQ ID NO: 1 ;(b) a disulfide bond between Q527C-E634C;(c) an A293P mutation;(d) oP3H trimer; and / or(e) a I5350A modified to harbor an extendable link to DI (l5350A2h or l5350A2hN).

2. The gB polypeptide of claim 1 , wherein the modification comprises the deletion of (a) and further comprises insertion of a GSPPGSPP linker (SEQ ID NO: 51).

3. The gB polypeptide of claim 1 , wherein the transmembrane and / or cytoplasmic domains have been deleted.

4. The gB polypeptide of claim 1 , wherein the gB is derived from an Epstein Barr Virus (EBV), a Human Cytomegalovirus (CMV), a Human Herpesvirus 6 (HHV6), a Herpes Simplex Virus 1 (HSV1), or a Varicella Zoster Virus (VZV).

5. The gB polypeptide of claim 1 , further comprising a mutation of the fusion loop as indicated in Table 1 (SEQ ID NOs: 25-36, 39, 41).

6. The gB polypeptide of claim 1 , comprising an amino acid sequence having at least 90% sequence identity to an amino acid sequence selected from SEQ ID NOs: 25-43.

7. The gB polypeptide of claim 1 , comprising an amino acid sequence selected from SEQ ID NOs: 25-43.

8. The gB polypeptide of any one of claims 1-7, wherein the polypeptide is in soluble form.

9. A composition comprising the gB polypeptide of any one of claims 1-8.

10. A nucleic acid molecule that encodes the gB polypeptide of any one of claims 1 -7.

11. The nucleic acid molecule of claim 10, wherein the molecule is an mRNA or DNA.

12. The nucleic acid molecule of claim 10, wherein the nucleic acid molecule comprises a modified nucleotide.

13. The nucleic acid molecule of claim 12, wherein the modified nucleotide comprises a methylpseudouridine.

14. A composition comprising the nucleic acid molecule of any one of claims 10 to 13.

15. A method of eliciting an immune response to a Herpesviridae family virus in a mammalian subject, the method comprising administering a composition of claim 9 or 14 to the subject.

16. A method of isolating gB-directed B-cells and / or antibodies from a blood sample obtained from a mammalian subject, the method comprising contacting the blood sample with the gB polypeptide of any one of claims 1 -8 and isolating gB-directed B-cells and / or antibodies that specifically recognize and bind the gB polypeptide.

17. The method of claim 16, wherein the subject has been previously infected or immunized with a Herpesviridae gB polypeptide.

18. The method of claim 16 or 17, wherein the isolating comprises ELISPOT assay, magnetic cell separation, affinity chromatography, fluorescence-activated cell sorting (FACS), biolayer interferometry (BLI), surface plasmon resonance (SPR), mass photometry (MP), or enzyme-linked immunosorbent assay (ELISA).

19. The method of any one of claims 16-18, wherein the gB-directed B-cells and / or antibodies are directed against a gB polypeptide derived from Epstein Barr Virus (EBV), a Human Cytomegalovirus (CMV), a Human Herpesvirus 6 (HHV6), a Herpes Simplex Virus 1 (HSV1), or a Varicella Zoster Virus (VZV).

Images