Crystal Structure of Venezuelan Hemorrhagic Fever Virus Fusion Glycoprotein Reveals a Class 1 Postfusion Architecture with Extensive Glycosylation

Guanarito virus (GTOV) is an emergent and deadly pathogen. We present the crystal structure of the glycosylated GTOV fusion glycoprotein to 4.1-Å resolution in the postfusion conformation. Our structure reveals a classical six-helix bundle and presents direct verification that New World arenaviruses exhibit class I viral membrane fusion machinery. The structure provides visualization of an N-linked glycocalyx coat, and consideration of glycan dynamics reveals extensive coverage of the underlying protein surface, following virus-host membrane fusion.

The structure was solved by molecular replacement with Phaser (31) using aglycosylated LCMV GP2 as a search model (PDB accession number 3MKO [17]). Model building was performed with Coot (32). The crystal contained protein-glycan and glycan-glycan lattice contacts, which facilitated visualization of carbohydrate chains from three out of the five N-linked glycosylation sites on GP2. The two remaining N-linked sites, although likely to be at least partially occupied (Fig. 1C), lacked surrounding stabilizing environments and were thus not visible in the crystal structure. Oligomannose-type glycans were built based on carbohydrate from PDB accession number 2WAH (33). Model building was facilitated by map sharpening (35). Refinement in Buster (36) used local structural similarity restraints (LSSR) to the high-resolution 3MKO structure, grouped B-factor refinement (grouped by chain), tensor libration screw (TLS) modeling, and local 3-fold noncrystallographic symmetry restraints ( Table 1). The final protein structure was validated using MolProbity (34). of the GTOV structure. Arenaviruses contain a bisegmented, ambisense RNA genome. The long RNA segment encodes the RNA polymerase (L) and matrix protein (Z). The short segment encodes the nucleoprotein (NP) and a glycoprotein precursor (GPC). Proteolytic cleavage of GPC by the cellular proprotein convertase site 1 protease (SK-1/S1P) (45) yields three products: a stable signal peptide (ssp) required for maturation (46)(47)(48), GP1, and GP2. These components noncovalently associate to form the GP, which further assembles into a trimeric spike on the virion surface (49). TR, transmembrane region; CR, cytoplasmic region. Each of the three GP2 subunits in the asymmetric unit consists of three regions: a 45-amino-acid N-terminal ␣-helix (residues 301 to 346), a 50-amino-acid "T region" (residues 347 to 398; named in comparison with the Old World LCMV GP2 structure [17]), and a short C-terminal ␣-helix (residues 399 to 408) ( Fig.  2A and B and 3A and B). The N-terminal ␣-helix spans the entire length of the 85-Å-long molecule and connects to the C-terminal helix through the T region, which is composed of loops and a small ␣-helix (residues 358 to 364). The N-and C-terminal helices pack closely in an antiparallel arrangement, where each protomer associates to form a trimeric coiled coil. The N-terminal part of this coiled coil contains the previously described "stutter" structure, which is common throughout class I viral fusion proteins (17). This confirms that the New World GP2 is a class I fusion glycoprotein. Our structure is in a postfusion conformation (8,37), with the N terminus (fusion loop region) and C terminus (transmembrane region) colocalized, consistent with the merger of the virion and host cell membranes (Fig. 2).
The closest relative of GTOV GP2 with a known structure is aglycosylated Old World LCMV GP2 (17). Both structures are identical in length, exhibit very similar secondary structures, and maintain approximately 50% amino acid sequence identity ( Fig.  3C and D). However, superposition of equivalent GTOV and LCMV protomers leads to a greater structural variation between the two structures (a 2.2-Å root mean square deviation [RMSD] over 96 C␣ residues) than expected between two structures with 50% sequence identity (an ϳ1.0-Å RMSD [24]). This is largely due to rigid-body translational differences in the organization of the three regions ( Fig. 3C and D). While the N-terminal helices of GTOV and LCMV independently superpose well (1.05-Å RMSD over 50 C␣ residues), the locations of many equivalent residues in the T region and C-terminal helices differ by more than 3 Å (Fig.  3D). Such structural dissimilarities may reflect intrinsic differences between Old and New World architectures.
In contrast to the well-ordered N-and C-terminal helical regions, the T region exhibits a high degree of flexibility. Comparison of the three independent GTOV protomers reveals that there are differences between loop conformations in the T region (Fig.  3B). The T region also contains a 10-amino-acid stretch (residues 365 to 375) which is fully ordered in one protomer and disordered in the second and third. Differences in T regions are likely due to differential crystal packing environments of individual protomers. Surprisingly, the T region does not correlate with marked sequence divergence (Fig. 1D) (60% of T-region residues are conserved among New World arenaviruses).
Our structure was determined in the absence of the N-terminal 50-amino-acid segment, corresponding to the bipartite fusion loop. Although it would be interesting to investigate how this loop is structured and how it might influence the flexibility of the underlying T regions, this peptide is orientated in the opposite direction, toward the target host membrane (Fig. 1). We therefore predict limited direct interactions between these regions. N-linked glycosylation plays a key role in arenavirus glycoprotein folding, maturation, cellular tropism, viral fusion, and immune evasion (38)(39)(40). As estimated by SDS-PAGE analysis, Nlinked glycosylation contributes up to 40% of the mass of recombinant GTOV GP2 (Fig. 1C) (28). The N-linked sequons are differentially occupied when recombinantly expressed, supporting the conclusion that most of these sites are, at least individually, dispensable for folding (Fig. 1C). Analysis by the NetNGlyc server (http://www.cbs.dtu.dk/services/NetNGlyc/), which is trained on a database of experimentally determined glycan occupancy data (41), supports the conclusion that GTOV GP2 likely contains partially occupied sites on the native infectious virion.
A high density of carbohydrate covers the surface of the GP2 postfusion structure. Based on the visible electron density of these sites (Fig. 2), we created a model for how N-linked glycosylation is presented by GP2 following fusion (Fig. 3E). Although we do not preclude the presence of oligomannose-or hybrid-type structures on the mature virion, this was performed using the structure of a complex glycan (42), typical of secreted glycoproteins (43), as opposed to the artificial glycosylation structure engineered in this study. Consideration of the natural conformational flexibility of these glycans (44), when not restricted by crystallographic packing, reveals that the majority of the protein surface is occluded by a dynamic glycocalyx, with only the N-and C-terminal membrane-proximal regions readily accessible (Fig. 3E).
Unlike the glycosylation sites on GP1 (12), the sites on GP2 are well conserved across New World arenaviruses, with the exception of one unique site in GTOV (Asn314) (Fig. 1D). The conservation of all but one site suggests that there is selective pressure to maintain these positions. This restriction is likely to be driven, in part, by the requirement of GP2 glycosylation sites to be solvent accessible in both prefusion (GP1-bound [18,19]) and postfusion states, a factor that does not significantly influence GP1. Because of the sheer density of glycans on GP2 and the requirement for association with GP1 (18,19), we suggest that ordering some of the glycans may be necessary for the productive formation of the GP1-GP2 heterodimer (Fig. 3E). A further consequence of this hypothesis might be that the glycans may be less processed than predicted due to steric constraints of the overall packing environment of the GP-GP2 heterotrimer. Given the influence of carbohydrate processing on arenaviral tropism and immune responses (38)(39)(40), it would be valuable to know the carbohydrate compositions of native virions. We present here the first structure of a New World arenavirus fusion glycoprotein and show that GTOV GP2 adopts an archetypal class I-type postfusion fold. The positions of the well-ordered N-and C-terminal helical regions differ from those of the equivalent Old World LCMV, and our structure thereby provides an improved template for the New World fusion glycoprotein. In our structure, we also visualize N-linked carbohydrates that form a glycocalyx and obscure much of the protein surface. On the whole, our analysis broadens the structural coverage of the mature New World glycoprotein subcomponents and provides evidence that arenaviral glycoprotein architecture can vary between clades.  Protein structure accession number. Coordinates and structure factors have been deposited in the Protein Data Bank (accession number 4C53).