Hantavirus structure – molecular interactions behind the scene

Viruses of the genus Hantavirus, carried and transmitted by rodents and insectivores, are the exception in the vector-borne virus family Bunyaviridae, since viruses of the other genera are transmitted via arthropods. The single-stranded, negative-sense, RNA genome of hantaviruses is trisegmented into small, medium and large (S, M and L) segments. The segments, respectively, encode three structural proteins: nucleocapsid (N) protein, two glycoproteins Gn and Gc and an RNA-dependent RNA-polymerase. The genome segments, encapsidated by the N protein to form ribonucleoproteins, are enclosed inside a lipid envelope that is decorated by spikes composed of Gn and Gc. The virion displays round or pleomorphic morphology with a diameter of roughly 120– 160 nm depending on the detection method. This review focuses on the structural components of hantaviruses, their interactions, the mechanisms behind virion assembly and the interactions that maintain virion integrity. We attempt to summarize recent results on the virion structure and to suggest mechanisms on how the assembly is driven. We also compare hantaviruses to other bunyaviruses with known structure.


Discovery and history of hantaviruses
The oldest clinical records of hantavirus-induced haemorrhagic fever with renal syndrome (HFRS) are found in Russia as early as 1913; however, Chinese records from AD 960 describe a similar disease (Johnson, 2001).The causative agent of HFRS, isolated in 1976(Lee et al., 1978), was named Hantaan virus (HTNV) after the Hantan River in South Korea (Lee et al., 1982).HTNV was chosen as the prototype strain of the genus Hantavirus in the family Bunyaviridae (Schmaljohn & Dalrymple, 1983).Nephropathia epidemica (NE) described in Sweden and Finland was soon found to be related to HFRS (Svedmyr et al., 1979), and the causative agent was identified as Puumala virus (PUUV) (Brummer-Korvenkontio et al., 1980).In 1993 a hantavirus, later named Sin Nombre (SNV, nameless in Spanish), was identified as the causative agent of a deadly epidemic in Northern America (Feldmann et al., 1993;Nichol et al., 1993).The disease associated with SNV infection was termed hantavirus pulmonary syndrome (HPS) (Hughes et al., 1993).However, due to the associated cardiopulmonary manifestations this disease is today generally referred to as hantavirus cardiopulmonary syndrome (HCPS) (Hallin et al., 1996).It is estimated that globally around 150 000 people get infected by hantavirus each year (Jonsson et al., 2010).Hantavirus-induced diseases have typical case fatality rates of 0.1 % for NE, from 2 up to 15 % for HFRS and around 40 % or more for HCPS (Jonsson et al., 2010).

Host and transmission of hantavirus
Hantaviruses are carried by rodents or insectivores and according to the International Committee on Taxonomy of Viruses each hantavirus has a different primary reservoir species or subspecies (King et al., 2011).Hantaviruses are divided into four distinct groups based on their host reservoirs.The first group contains HFRS-causing viruses such as HTNV, Seoul (SEOV) and Dobrava-Belgrade virus (DOBV), harboured by Old World rats and mice (Kru ¨ger et al., 2001).The viruses of the second group, carried by voles and lemmings, cause either NE (PUUV) or are considered apathogenic like Tula (TULV), Prospect Hill virus (PHV), Khabarovsk and Topografov viruses (Olsson et al., 2010).The third group of hantaviruses, carried by New World rats and mice includes SNV, Andes (ANDV), New York, Black Creek Canal and Laguna Negra viruses that cause HCPS when transmitted to man (Maes et al., 2009).The fourth group is the most recent addition to hantaviruses, even though the prototype Thottapalayam virus was already isolated in 1964 (Carey et al., 1971).The members of this group are carried by shrews (Song et al., 2007) and, currently, no disease or symptoms have been associated with insectivore-borne hantaviruses.

Hantavirus particle and its components
The literature describes hantavirus virion as a round, pleiomorphic or tubular particle with a broad size variation ranging from 70 to 350 nm in diameter/length (Battisti et al., 2011;Huiskonen et al., 2010;Martin et al., 1985).In both electron microscopy (EM) sections of infected cells and in cryo-EM the virions are reported to display mainly round or roundish morphology (Goldsmith et al., 1995;Huiskonen et al., 2010), suggesting the virions to be sensitive to mechanical stress and arguing the observed pleiomorphism to be partially artefactual.The virion consists of genome segments [encapsidated by the nucleocapsid (N) protein to form ribonucleoprotein, RNP], that are together with RNAdependent RNA-polymerase (RdRp) enclosed inside a lipid envelope decorated by spikes formed of glycoproteins Gn and Gc (Hussein et al., 2011;Spiropoulou, 2011).The virion envelope, 5 nm by thickness, holds the spike assemblies that protrude approximately 10 nm from the membrane and display a fourfold rotational symmetry (Battisti et al., 2011;Huiskonen et al., 2010).A schematic presentation of hantavirus particle is shown in Fig. 1(a).

The genome
The negative-stranded RNA genome of hantaviruses is segmented in S (small, 1.8-2.1 kb), M (medium, 3.7-3.8kb) and L (large 6.5-6.6 kb) segments, a characteristic of the members of the family Bunyaviridae (Plyusnin et al., 1996).Each segment of the viral RNA (vRNA) contains an ORF that is flanked at 39-and 59-ends by an NTR (Jonsson & Schmaljohn, 2001).The genome of hantaviruses is highly conserved at the very terminal nucleotides of the 39-and 59-ends of each segment, as initially shown for HTNV (Schmaljohn & Dalrymple, 1983), and later reported for SNV and PUUV (Chizhikov et al., 1995;Piiparinen et al., 1997).The 39-and 59-ends of the segments harbour the capacity to form a panhandle structure (through complementary nucleotides), which is thought to act as the viral promoter (Jonsson & Schmaljohn, 2001).The promoter function of panhandle is supported by observations that terminal deletions in each segment correlate with a decreased replication level in persistently infected cell cultures (Meyer & Schmaljohn, 2000).

The proteome
The genome segments S, M and L, respectively, encode the N protein, glycoprotein precursor (GPC) and RdRp (Hussein et al., 2011).The S segments of hantaviruses carried by members of the family Cricetidae rodents also contain an alternative ORF that encodes a non-structural protein (NSs) (Plyusnin, 2002).Individual hantavirus proteins are described in detail below.

N protein
The S segment-encoded N protein is a non-glycosylated protein with a molecular mass of approximately 50 kDa (Hussein et al., 2011).Both the N protein and RdRp are required for replication (Flick et al., 2003).Thus, to provide a sufficient pool of the N protein, mRNA transcription and translation are believed to precede the initiation of replication (Patterson & Kolakofsky, 1984) and the concentration of free N protein is thought to mediate the switch from mRNA synthesis towards replication (Jonsson & Schmaljohn, 2001).Furthermore, trimers of the N protein bind to the conserved panhandle structures present only in complementary RNA (cRNA) and vRNA, leaving the mRNA naked (Hussein et al., 2011).Recent reports of the N protein being responsible for sequestering the primers carrying a 59-cap, a crucial prerequisite for efficient translation in eukaryotic cells, from cellular processing (P) bodies (Mir et al., 2008) suggest even more direct involvement of N protein in virus replication.Immunofluorescence staining of hantavirus-infected cells by N protein antibodies or acute phase HFRS-patient sera is characterized by granular staining pattern (Kallio-Kokko et al., 2001;Lee et al., 1978;Yanagihara et al., 1985).This staining is suggested to be either due to aggregation of the N protein to inclusion bodies (Goldsmith et al., 1995) or due to the accumulation into P bodies (Mir et al., 2008).The N protein also localizes to the perinuclear region and is membrane associated, albeit devoid of transmembrane (TM) helices (Ravkov & Compans, 2001).

NSs
The S segment of hantaviruses carried by Arvicolinae and Sigmodontinae rodents contains an additional ORF that overlaps the N-protein-encoding ORF (+1 nt) and encodes a putative NSs (Plyusnin & Morzunov, 2001).NSs has been detected in PUUV-and TULV-infected cells, localized to the perinuclear area (Ja ¨a ¨skela ¨inen et al., 2007;Virtanen et al., 2010).NSs is reported to inhibit the expression of beta interferon (IFN-b) and to interfere with the activities of nuclear factor-kB and IFN regulatory factor 3 (IRF-3) (Ja ¨a ¨skela ¨inen et al., 2007).This immunemodulatory function is supported by the observation that TULV with the ORF for NSs survives for more consequent passages in IFN-competent cell culture than TULV without the ORF (Ja ¨a ¨skela ¨inen et al., 2008).

GPC, glycoproteins Gn and Gc
The GPC encoded by the genome M segment is a polyprotein of 1133-1158 aa residues (Spiropoulou, 2011).Cotranslational cleavage of GPC C-terminally to a conserved WAASA sequence by the cellular signal peptidase complex (Lo ¨ber et al., 2001) gives rise to glycoproteins Gn and Gc (Nand C-terminal portions of GPC, respectively, previously known as G1 and G2) (Spiropoulou, 2011).In fact, fulllength GPC has not been observed during infection, neither in recombinant expression nor in virions (Elliott et al., 1984;Pensiero & Hay, 1992;Spiropoulou et al., 2003).The Cterminal TM helix of Gn acts as a signal sequence for Gc and harbours the conserved WAASA sequence (Spiropoulou, 2011).In the case of HTNV, only a single cleavage has been reported to occur at residue 648 of GPC ( 644 WAASA 648 ) and the resulting Gn contains at least residues 588-614 with no evidence of a secondary cleavage (Lo ¨ber et al., 2001;Schmaljohn et al., 1987).
Folding and maturation of Gn and Gc.Virus-neutralizing Gn-and Gc-specific antibodies (Hooper et al., 1999;Schmaljohn et al., 1990) make an ideal tool for the study of glycoprotein folding, since they recognize the mature glycoproteins of the virion.Another approach to study folding is to characterize the trimming of glycoprotein sugar moieties during maturation.Both immature and mature glycoproteins of hantaviruses are endoglycosidase H (endo H) sensitive, indicating a high-mannose type of glycosylation (Schmaljohn et al., 1986), thus disabling this approach in the study of folding.
During infection the glycoproteins Gn and Gc of hantaviruses are found in the Golgi complex and coexpression is considered as a prerequisite for localization to the Golgi complex (Pensiero & Hay, 1992;Ruusala et al., 1992).It is clear that Golgi transport of Gc relies on the expression of Gn, but it is less evident whether Gn is able to traffic there when expressed alone (Spiropoulou, 2011).The Gn proteins of other family Bunyaviridae viruses appear to be capable of localizing to the Golgi complex when expressed without Gc (Bupp et al., 1996;Haferkamp et al., 2005;Kikkert et al., 2001;Lappin et al., 1994;Ro ¨nnholm, 1992), and a Golgi targeting motif has been mapped to the N-terminal part of Gn cytoplasmic tail (Gn-CT) in the case of phleboviruses (Andersson et al., 1997;Gerrard & Nichol, 2002).Whether the observed differences in the localization of Gn are due to the expression system or the cell line used, or whether hantaviruses truly are the only genus in the family, whose Gn protein is incapable of Golgi localization, remains to be determined.Interestingly, the glycoproteins of SNV were found to be transported to the plasma membrane (Ogino et al., 2004;Spiropoulou et al., 2003).Spiropoulou et al. suggested this phenomenon reflected a difference between the Old and New World hantaviruses, while Ogino et al. explained the transport to the plasma membrane was due to the different cell lines used.It would seem reasonable that both Old and New World hantaviruses would bud at the same cellular location, presumably at cis-Golgi as the other members of the family Bunyaviridae.This remains as an open question that requires further studies.
The maturing Gn and Gc are decorated by N-linked glycans that are thought to be involved in protein folding, receptor binding, membrane fusion and viral morphogenesis.The GPC of HTNV contains five N-glycosylation sites on Gn (asparagine, N, residues 134, 235, 347, 399 and 609) and one on Gc (N928) (Schmaljohn et al., 1987).Of these sites N134, N347, N399 and N928 are conserved among hantaviruses (as depicted in Fig. 1b), and are occupied by high-mannose type glycans in the glycoproteins of HTNV and PUUV (Johansson et al., 2004;Shi & Elliott, 2004).There is also variation in the glycosylation pattern between hantaviruses, for instance HTNV Gn contains an additional N-glycosylation site at N235 (Shi & Elliott, 2004) and PUUV Gc an O-glycosylation site at threonine 985 (Johansson et al., 2004).Mutations of N-glycosylation sites affects the ability of Gn and Gc to form complexes, and thus also influences the intracellular localization and the formation of epitopes (Shi & Elliott, 2004).The trimming of glycans occurs at medial-and trans-Golgi compartments, and thus the existence of high-mannose type Nglycans in the glycoproteins indicates that they are only able to transit to the cis-Golgi (Shi & Elliott, 2004).
Hantavirus-infected cells can be induced to form syncytia under acidic conditions (at pH values below 6.3) (Arikawa et al., 1985).The ability to induce syncytia varies between hantavirus strains and low-pH treatment of PUUV-, HTNV-and SEOV-infected cells, respectively, yields small, medium-sized and large syncytia (McCaughey et al., 1999).The fusogenicity is shown to reside in GPC, since its expression transforms cells prone to low-pH syncytiuminduction (Ogino et al., 2004).The formation of syncytia can be prevented by glycoprotein antibodies with epitopes either in Gc or in the Gn-Gc complex (Ogino et al., 2004), suggesting Gc to harbour the fusogenicity.Bioinformatics studies have further suggested the Gc of hantaviruses (and other members of the family Bunyaviridae) to be a class II viral fusion protein (Garry & Garry, 2004;Tischler et al., 2005).Tischler et al. initially demonstrated that the putative fusion loop of Gc is able to bind artificial membranes and later confirmed the existence and functionality of the predicted fusion loop in Gc by utilizing lentiviral vector system to express the GPC (Cifuentes-Mun ˜oz et al., 2010, 2011).The fusion loop of class II viral fusion proteins is in the pre-fusion conformation hidden by a tight interaction between either homo-or heterodimers of glycoproteins (White et al., 2008).At acidic pH the fusion loop of class II fusion protein is exposed by dissociation of the homo-or heterodimeric complex (White et al., 2008).The conformation of the fusion protein is altered to allow the attachment of the fusion loop to the target membrane, followed by further changes in conformation and by formation of a tight homotrimeric complex (White et al., 2008).The low-pH exposure should cause irreversible changes in the conformation of a class II fusion protein (White et al., 2008).We did not observe complete inactivation of TULV after low-pH treatment, an observation supported by studies of Higa et al. (Hepojoki et al., 2010a;Higa et al., 2012).In both studies the reduction in virus titre could be explained by aggregation of virions, as reported in the case of phleboviruses (Overby et al., 2008).Since class III viral fusion proteins are capable of reversible conformation changes (White et al., 2008), it is possible that the complete change in conformation of Gc may not be achieved by low-pH alone, but would require the presence of additional factors.Such factors could include binding to cellular receptor, shuffling of disulphide bonds by protein disulphide isomerase or presence of target membrane (Hepojoki et al., 2010a;Krey et al., 2005;Sanders, 2000).The Gc of hantaviruses contains a conserved CxxC motif (Hepojoki et al., 2010a;Strandin et al., 2011b), which is found in protein disulphide isomerases (Sanders, 2000).This motif could release the fusogenic activity of Gc during virus entry and account for the observed stability of Gc at low pH.We demonstrated a significant reduction in virus infectivity after treatment of virions with thiol-reactive compounds in support of the disulphide isomerase and/or active thiol shuffling hypothesis (Strandin et al., 2011b).
Cytoplasmic (or intraviral) tails (CTs) of Gn and Gc.Both glycoproteins possess a CT; the one in Gn is ~110 aa and the one in Gc ~10 aa residues (Spiropoulou, 2011), and recent evidence suggests that the Gn-CT acts as a surrogate matrix protein of hantaviruses (Hepojoki et al., 2010b;Strandin et al., 2011a).The Gn-CT contains a zinc finger (ZF) domain of approximately 50 residues, shown by NMR analysis to fold as a tandem CCHC ZF (Estrada et al., 2009).The ZF domain is conserved among pathogenic and apathogenic hantaviruses, and it is claimed to be the only domain with a clearly defined fold in Gn-CT (Estrada et al., 2011).However, the function of this domain is not known.Cytotoxicity has prevented the expression of the full Gn-CT in bacterial cells (Estrada et al., 2011), and thus the structure for complete Gn-CT is unfortunately not yet available.The cytotoxicity might be due to Gn-CTs nonspecific ability to bind nucleic acids (Strandin et al., 2011a), which could lead to intervention of bacterial transcription and translation.We have been able to express the Gn-CTs of PUUV and TULV as glutathione S-transferase-fusion proteins in bacteria (Strandin et al., 2011a), and perhaps such a construct weakens the cytotoxicity.The C-terminal TM helix of Gn-CT acts as a signal sequence for Gc and is termed a degradation signal or degron that destines the Gn-CT to proteolytic degradation (Geimonen et al., 2003a).The degron was reported to be present only in the Gn-CT of pathogenic hantaviruses (Sen et al., 2007), but a similar domain was found also from the apathogenic TULV and PHV (Wang et al., 2009).The degron is most likely an artefact caused by Gn-CT aggregation in cells due to lack of the N-terminal TM helix and a signal sequence required for targeting the protein to endoplasmic reticulum (ER).The C terminus of Gn-CT contains a highly conserved YxxL motif, and HCPS-causing hantaviruses are claimed to have two such motifs that form an immunoreceptor tyrosine-based activation motif (ITAM) (Geimonen et al., 2003b).The second YxxL motif is by prediction buried inside the C-terminal TM of Gn-CT and it is thus questionable whether the ITAM could be functional in native Gn.Thus, YxxL of hantaviruses is likely to serve some other function, such as acting as a trafficking or targeting signal.Several studies have suggested Gn-CT to also regulate cellular responses to infection (Alff et al., 2006;Matthys et al., 2011).

RdRp (or L protein)
The largest protein encoded by hantaviruses is the RdRp derived from the L segment (Spiropoulou, 2011).By mobility in SDS-PAGE the molecular mass of RdRp has been estimated to be ~250 kDa (Kukkonen et al., 2004).Since both the N protein and RdRp of hantaviruses are associated with perinuclear membranes, the transcription and replication events are believed to occur on membranes (Kukkonen et al., 2004;Ravkov & Compans, 2001).The RdRp of hantaviruses is capable of recombining homologous RNA sequences and thus enables virus evolution via superinfection (Plyusnin et al., 2002).The RdRp of hantaviruses requires a divalent cation (either Mg 2+ or Mn 2+ ) for activity with a preference for Mn 2+ (Schmaljohn & Dalrymple, 1983).

Entry of hantaviruses
Several putative receptors have been described for hantaviruses and include decay-accelerating factor, complement receptor gC1qR, an unknown 70 kDa protein, and b1-and b3-integrins (Buranda et al., 2010;Choi et al., 2008;Gavrilovskaya et al., 1998;Krautkra ¨mer & Zeier, 2008;Mou et al., 2006).Some lectins also promote the entry of hantaviruses (Ogino et al., 1999) and thus it is likely that the glycosylation of Gn and Gc plays a role in the attachment to host cells.The entry occurs preferably from the apical side of polarized epithelial cells (Ravkov et al., 1997), but the virus can also pass through the basolateral side (Rowe & Pekosz, 2006).After attaching to the cell surface, the virus is internalized in clathrin-coated vesicles as reported for HTNV (Jin et al., 2002).However, ANDV was reported not to use clathrin-mediated endocytosis (Ramanathan & Jonsson, 2008) and thus different routes may be involved in the entry, similarly to the observations made with Uukuniemi phlebovirus (Lozach et al., 2010).After internalization viruses enter the early endosome and hantaviruses are reported to require low pH for successful infection (Jin et al., 2002), suggesting that fusion occurs either at the early or late endosome (White et al., 2008).The mechanism of RNP transport to the site of replication is not known; however, it could be driven by interactions of the N protein with microtubules and actin (Ramanathan et al., 2007;Ramanathan & Jonsson, 2008;Ravkov et al., 1998).

Oligomerization of N protein and packaging of RNA
The N-terminal domain of roughly 80 residues of the N protein folds as an antiparallel coiled-coil when expressed alone (Boudko et al., 2007).This domain along with the Cterminal residues mediates the oligomerization of the N protein (Alfadhli et al., 2001;Kaukinen et al., 2003).The coiled coiling in the N-terminal domain is either intra- (Boudko et al., 2007) or intermolecular (Alfadhli et al., 2002) depending on the concentration.The oligomerization of N protein to trimers was suggested to be mediated by head-to-head and tail-to-tail interactions with both the N-and C-terminal residues contributing to the oligomerization (Kaukinen et al., 2003).As the conformation of the N protein is altered as a result of binding to either mRNA caps or vRNA (Mir et al., 2010), it could be speculated that the interaction with nucleic acids would also contribute to the oligomerization of the N protein.
Interaction with RNA and formation of RNP complexes.Even though the N protein has the ability to bind various different RNAs: unspecific RNA, tRNA, vRNA, cRNA and mRNA (Go ¨tt et al., 1993), it has the strongest affinity for the 59 terminus of the vRNA (Mir & Panganiban, 2004;Severson et al., 1999).The reported RNA-binding sites are dispersed throughout the N protein, and amino acids that contribute to binding are found at the N-terminal half (Severson et al., 2005), at residues 175-217 in the case of HTNV (Xu et al., 2002), and at the C-terminal region (Go ¨tt et al., 1993).Formation of the RNP complex is initiated by formation of a trimeric N-protein complex (Kaukinen et al., 2001), which recognizes the vRNA panhandle in a genus-specific manner (Mir et al., 2006).The conformation of N protein is altered as a result of binding to the panhandle, which probably drives oligomerization further (Hussein et al., 2011).Mature RNPs appear as long helical filaments in EM (Goldsmith et al., 1995) backbone of vRNA is presumably wrapped around a scaffold formed by N protein oligomers, as suggested for Rift Valley fever virus (RVFV) N protein (Raymond et al., 2010).
N protein also participates in the transcription and replication via interactions to vRNA and cRNA (Hussein et al., 2011).An even more direct role has been assigned for N protein in transcription, since it has been shown to mediate the binding and sequestering of cellular mRNA caps from P bodies, wherein the cellular mRNAs are decapped and deadenylated (Mir et al., 2008).The 59-caps of cellular mRNAs, rescued from the P bodies by N protein, are used in the viral transcription and translation processes (Mir et al., 2008(Mir et al., , 2010;;Panganiban & Mir, 2009).In addition, N protein participates directly in the translation at ribosomes via promoting the initiation by replacing the multisubunit eukaryotic initiation (eIF) 4F complex (Cheng et al., 2011;Mir & Panganiban, 2008, 2010;Panganiban & Mir, 2009).The eIF4F complex is formed of a cap-binding protein, eIF4E and eIF4A (RNA helicase) that are assembled on a scaffolding protein eIF4G (Walsh, 2010).Even though there is no significant homology between the N protein and the eIF4F complex (Mir & Panganiban, 2008), the N protein was shown to act as a translation initiation factor also for non-viral proteins (Panganiban & Mir, 2009).This suggests that the N protein is a bridge between mRNA and 43S pre-initiation complex and the N protein does indeed interact with the small ribosomal unit 40S via binding to its subcomponents, ribosomal protein S19 (RPS19) and 18S rRNA (Haque & Mir, 2010).The N protein binds simultaneously to the RPS19 and a conserved triplet repeat sequence at the 59end of vRNA, and thus presumably mainly promotes the translation of viral mRNAs (Cheng et al., 2011).
The Gn-Gc spike complex -structure and composition By now cryo-EM has been applied to study the structures of two genera, hantaviruses and phleboviruses, of the family Bunyaviridae.RVFV and Uukuniemi virus (UUKV), members of the genus Phlebovirus, are characterized with a clearly ordered icosahedral surface structure (Freiberg et al., 2008;Huiskonen et al., 2009;Overby et al., 2008), whereas the surface of HTNV and TULV is composed of subunits with fourfold symmetry (Battisti et al., 2011;Huiskonen et al., 2010).Although, not studied in such detail, the surface of Bunyamwera virus of the genus Orthobunyavirus also contains substructures (penta-and hexagonal profiles) related to icosahedral symmetry (Novoa et al., 2005).Currently, no ordered surface structure has been assigned to the viruses of the other two genera, Nairovirus and Tospovirus, of the family.
Since Gn and Gc originate from GPC and require coexpression for proper folding and localization (Spiropoulou, 2011), it seems likely that equimolar ratio of glycoproteins is found in the virion.The spikes seen on the surface of the hantavirus particle show fourfold symmetry (Huiskonen et al., 2010), and based on the interactions of Gn and Gc each spike contains four molecules of both glycoproteins (Hepojoki et al., 2010a).The fourfold symmetry is most likely maintained by homo-oligomerization of Gn, since it has been described to form SDS-stable complexes yielding several bands in SDS-PAGE (Antic et al., 1992;Hepojoki et al., 2010a;Pensiero & Hay, 1992;Spiropoulou et al., 2003).Moreover, the Gc protein of hantaviruses forms stronger homo-than hetero-oligomeric contacts (Hepojoki et al., 2010a).The interplay between glycoproteins during folding (Pensiero & Hay, 1992;Ruusala et al., 1992), and observations that Gn and Gc co-immunoprecipitate (Antic et al., 1992;Arikawa et al., 1989) and are cross-linked (Antic et al., 1992), demonstrate an interaction between them.It is evident that Gn and Gc make contacts to form the virion.However, based on our data the interactions are rather between multiple units of both glycoproteins than just one of each (Hepojoki et al., 2010a).We thus suggest that the surface of the hantavirus particle is created by Gn tetramers that interconnect by Gc dimers (one dimer at each face of the square-like spike, making the Gn : Gc ratio 1 : 1) rather than interconnected heterotetramers.Recently, the spike complex of HTNV was resolved and suggested to form a Gn-Gc heterotetramer (Battisti et al., 2011).Thus, Battisti et al. reported that the HTNV spike structure differed from the TULV-spike complex reported earlier (Huiskonen et al., 2010).It would seem possible that the observed differences between the HTNV-and TULV-spike complex are due to different sampling techniques used in the electron cryotomography rather than due to molecular differences between the two viruses.In TULV structure there is a central stalk at the fourfold symmetry axis, which is lacking in the HTNVspike complex (see Fig. 2).There is a density directly below the spike (and the stalk of TULV spike) in both structures, and it would thus seem likely that the stalk would exist in both structures.We suggest that the stalk would locate to the ectodomains of Gn, N-terminally to the first TM helix.This region is involved in multimerization of Gn (Spiropoulou et al., 2003) and thus the stalk could represent contacts between four Gn molecules we recently docked.In silico docking of a structure of a tetrameric coiled-coil to the stalk of TULV-spike complex (Hepojoki, 2011) the viral docking indicates that the stalk could accommodate such a structure.Coiled-coil interactions can provide extreme thermostability to protein complexes that are required to function under extreme conditions (Burkhard et al., 2001).Coiled-coiling is also considered to be the most frequent oligomerization motif between proteins or between subunits of proteins (Burkhard et al., 2001), thus tetramerization of Gn via such interaction could explain the observed SDS-stable Gn oligomers.We hope to provide experimental evidence to support this hypothesis in the near future.
In contrast to hantaviruses, where no classical symmetry can be applied, the surface structures of UUKV and RVFV phleboviruses display an icosahedral symmetry (Freiberg et al., 2008;Overby et al., 2008).This architecture is suggested to be maintained by contacts between heterodimeric Gn-Gc units (Huiskonen et al., 2009).Similarly to hantaviruses, the proper folding of UUKV glycoproteins requires co-expression (Persson & Pettersson, 1991), but conflicting data of the multimeric nature of Gn and Gc has been reported.Curiously one study showed that UUKV Gn and Gc form homodimers rather than heterodimers when extracted from virions (Ro ¨nka ¨et al., 1995), and in contrast to this, another group suggested the UUKV Gn and Gc exist as heterodimers when extracted from virions (Persson & Pettersson, 1991).The latter study was based on coimmunoprecipitation of Gn and Gc from density gradient fractions using Gn-and Gc-specific antisera.However, the sizes of homo-and heterodimeric complexes differ very little and the results of Ro ¨nka ¨et al. indicated that the mobility of only Gc is altered at low pH, suggesting glycoprotein complexes to be homodimeric.Together the results of Persson et al. and Ro ¨nka ¨et al. suggest that the complex between Gn and Gc is between oligomers of both proteins, as also suggested for hantaviruses.The structures of HTNV-, TULV-and RVFV-spike complexes are compared in Fig. 2. The interaction data of hantavirus glycoproteins are more easily embedded into the observed structure, where Gn tetramers would be found in the middle of the spike and Gc dimers would locate between the spikes.The interactions between phlebovirus glycoproteins are somewhat harder to include into the structure.The strong homodimeric contacts could be found in the interspike region, similarly to hantaviruses.However, since both Gn and Gc are reported to exist as homodimers when extracted from virions, it is hard to determine which of the glycoproteins would mediate this interspike contact.We suggest that the spikes of phleboviruses are connected via Gc-Gc contacts, analogously to hantaviruses.This is based on the assumption that Gc is the fusion protein of all bunyaviruses.By definition, class II fusion proteins lie parallel to the membrane and the fusion loop is covered by either hetero-or homodimeric interaction (White et al., 2008).When making connections between spikes the Gc could lie parallel to the membrane, and make both homoand hetero-contacts.The contacts of Gn protein in the RVFV could be explained more easily if only hexameric assemblies would be found, but due to icosahedral symmetry also pentameric assemblies need to be present.A hexamer could be broken into three homodimers, but one Gn molecule would be left out when a pentamer would be broken into homodimers.This sort of an arrangement of glycoproteins would make the virion contain equal amounts of both Gn and Gc.Of course it is also possible that the observed structure is maintained by contacts between Gn-Gc heterodimers, but the amount of different types of contacts required would be slightly higher.

Formation of the spike complex
The co-localization of glycoproteins to the Golgi complex suggests that the spike complex of the viruses of the family Bunyaviridae is assembled at either ER or cis-Golgi (Pettersson & Melin, 1996).The rate of phlebovirus glycoprotein folding has been studied with both UUKV and RVFV.The Gn of UUKV folds faster than Gc, as judged by either incorporation into the newly formed virions or by biochemical characteristics.Furthermore, newly formed Gn is assumed to complex with previously synthesized Gc based on faster incorporation of labelled Gn into virions (Kuismanen, 1984;Persson & Pettersson, 1991).For RVFV, the newly formed Gc is suggested to rapidly complex with Gn (Gerrard & Nichol, 2007).One could speculate that the result of fast Gn-Gc complex formation in the case of RVFV glycoproteins would be due Gn acting as a chaperone for Gc.In the case of UUKV the folding rates of Gn and Gc are roughly 10 and 60 min, respectively, and the formation of complexes containing both proteins were observed roughly after 30 min (Persson & Pettersson, 1991).Whether the rate of glycoprotein incorporation into viruses would differ between RVFV and UUKV is not known.In the case of hantaviruses the folding of both Gn and Gc requires co-expression, but the rate of glycoprotein folding has not been studied.The spike-forming glycoprotein of influenza virus (HA) folds in minutes and trimerizes in ~10 min (Gething et al., 1986).Interestingly, this is roughly the same time required for the folding of UUKV Gn protein (Persson & Pettersson, 1991).Since UUKV glycoproteins exist as homodimers when extracted from virions (Ro ¨nka ¨et al., 1995), it may be speculated that oligomerization of the glycoproteins would be required for making the Gn and Gc contacts.Clearly the spike complex of phleboviruses (see Fig. 2) is different from its hantaviral counterpart; however, the spike formation could still follow analogous routes in hantaviruses and phleboviruses.In both cases the Gn complex would form initially, thus in the case of hantaviruses the tetrameric Gn complex would precede the formation of the hetero-oligomeric complex.The Gn complex could act as a chaperone for the folding of Gc, in order to prevent the exposure of the putative fusion loop of Gc.The formation of the hetero-oligomeric complex would somehow alter the conformation of Gn (or both proteins) resulting in the transport from the ER to the Golgi, as Gn-Gc co-expression is a prerequisite for Golgi localization (Pettersson & Melin, 1996).In the case of hantaviruses this hypothetical conformation change would most likely affect the availability of Gn-CT at the cytoplasmic face of the ER membrane, and expose a targeting signal that would promote the transfer to the Golgi.The above hypotheses are depicted in Fig. 3.

Spike-RNP interaction and assembly
It is common for enveloped viruses that the capsid or RNP is connected to the envelope via a matrix protein.The members of the family Bunyaviridae lack a matrix protein and thus the glycoprotein CTs of several viruses have been shown to mediate the interaction to the RNP (Hepojoki et al., 2010b;Overby et al., 2007;Ribeiro et al., 2009;Snippe et al., 2007).In addition to interacting with the RNP, the matrix proteins of other enveloped viruses participate actively in regulation of replication (Suryanarayana et al., 1994;Watanabe et al., 1996), mediated via direct interactions to RNA (Elster et al., 1994).Curiously Gn-CTs, the putative matrix surrogates of RVFV (Piper et al., 2011), CCHF (Estrada & De Guzman, 2011) and TULV (Strandin et al., 2011a), were recently reported to bind nucleic acids.Hantavirus Gn-CT contains two classical CCHC-type ZFs, that do not participate in nucleic acid binding (Estrada et al., 2009).The interaction with DNA/RNA typically requires three such ZFs (Matthews & Sunde, 2002), and thus tetrameric Gn assembly containing eight ZFs might be able to bind nucleic acids.Currently, there is no experimental evidence to support this.
The spike-RNP interaction obviously takes place at the budding site, which is situated at the ER-Golgi intermediate compartment (ERGIC) or at the cis-Golgi based on the localization of Gn, Gc and N protein (Kuismanen et al., 1984).The association of hantavirus N protein with microtubules and actin along with localization to the ER-Golgi (Ramanathan et al., 2007;Ravkov et al., 1998) suggest that hantaviruses form a viral factory situated in the ERGIC or the Golgi, similarly to Bunyamwera virus (Fontana et al., 2008).Successful packaging of infectious progeny viruses requires contacts between the RdRp and RNP, binding of RNP to the spike complex and interactions between spikes.The RdRp and N protein of hantaviruses co-localize to the perinuclear and/or Golgi region similarly to RVFV (Brennan et al., 2011a).The N protein and RdRp of hantaviruses are suggested to interact (Cheng et al., 2011), and for RVFV this interaction has been demonstrated (Brennan et al., 2011a).Infectious virus should contain the genetic information of the three genome segments; however, there seems to be no requirement for the exact number of genome segments, since a two-segmented variant of RVFV was recently reported (Brennan et al., 2011b).How does the virus ensure packaging of multiple segments?A possible explanation would be that the RdRp exists as a homo-oligomer inside the infected cell.Oligomerization of the RdRp would help in recruiting multiple genome segments into progeny virions via cross-talk between the RdRp units.Curiously, the RdRp of RVFV exists as oligomers inside cells, and furthermore it is shown to be functional only when oligomerized (Zamoto-Niikura et al., 2009).Interestingly, the Gn-CT of phleboviruses was shown to bind the RdRp (Piper et al., 2011) and it is tempting to speculate that this interaction would be essential to initiate the packaging of infectious virions.
For hantaviruses it is known that generation of virus-like particles (VLPs) requires the co-expression of S-and Msegments (Betenbaugh et al., 1995).Also the assembly of RVFV can occur without the RdRp (Piper et al., 2011), but in contrast to hantaviruses the glycoproteins of phleboviruses are capable of forming VLPs without other viral proteins (Habjan et al., 2009;Overby et al., 2006).This is probably due to the fact that the Gn and Gc of phleboviruses can form an icosahedron (Freiberg et al., 2008;Overby et al., 2008).The incorporation of RdRp into the progeny virions is driven by interaction with the components of RNP (Brennan et al., 2011a) or by binding to Gn-CT (Piper et al., 2011).Given the interactions of Gn-CTs with nucleic acids and the ability of Gn-CT to bind both RdRp and RNP, it is tempting to speculate that Gn-CT would act as a trigger for the assembly, acting similarly to arenavirus Z protein that locks the polymerase-promoter complex (Kranzusch & Whelan, 2011).Our hypothesis for the initiation of assembly (depicted in Fig. 3) is that the formation of the hetero-oligomeric spike complex unveils the Gn-CT to the cytoplasm, thus explaining why co-expression Oligomerization of Gn 3a Fig. 3.A model for the initiation of hantavirus assembly.Steps 1-4 describe a hypothesis for the spike generation based on the assumption that folding of Gn is faster than the folding of Gc.Tetrameric Gn complex formed at step 4 acts as a chaperone for the folding (and synthesis) of Gc, steps 5-7.Repetition of steps 5-7 results in formation of a heterotetrameric spike complex.The formation of complete spike complex unveils Gn-CT to the cytoplasm, and drives the translocation of the complex to the cis-Golgi (or ER-Golgi intermediate compartment, ERGIC).At the site of replication the Gn-CT regulates transcription via contacts to RdRp, and halts the transcription when vRNA and N protein comes into the complex.Eventually at least three such complexes are clustered, through interactions between neighbouring RdRps, in order to package multiple RNPs into the progeny virion.This complex then acts as a nucleation centre for initiation of virion assembly, and more heterotetrameric spike complexes are recruited via interactions between RNP and Gn-CT (and Gc-CT).When two spikes come into contact with each other, the Gc proteins of these spikes make a new contact, Gc-Gc (for simplicity only two of four Gc molecules are shown for each spike) that interconnects the spikes.When a heterotetrameric spike complex has made contact from all four sides to neighbouring spikes, the Gn-CT becomes unavailable to the cytoplasm disrupting the Gn-CT-RNP contact (marked by white x inside the spike).This process continues and induces consecutive packaging of the RNP.Eventually an oval or roughly spherical assembly is created that pinches off from the membrane by an unknown mechanism creating the virion. of Gn and Gc is required for Golgi localization.This is supported by our observation that the CTs of both Gn and Gc are able to bind RNP (Hepojoki et al., 2010b), and further suggests that formation of a complete spike complex is a crucial prerequisite for initiation of packaging.The Gn-CTs of the spike interact with oligomers of RdRp, and recognition of the RNP-RdRp complex by Gn-CT induces a conformational change in the RdRp that halts the transcription.When neighbouring RdRps in the complex have stopped transcription, this complex acts as a nucleation centre for spike oligomerization.Such an arrangement would recruit glycoproteins, RdRp and RNPs into the progeny virions (but would not ensure packaging of different types of RNPs).The assembly would continue as driven by spike-RNP interactions that would bring pre-assembled spikes into close proximity, thus leading to formation of the spike lattice and eventually to formation of an intact virion.

Concluding remarks
The development of high resolution imaging and application of high-throughput techniques in crystallization provide an increasing number of structures for macromolecular complexes.The structural information is highly valuable, but typically needs to be complemented by in-depth characterization of the molecular interactions to attain thorough understanding.While this review attempts to put together some of the available biochemical and structural data, it also highlights some of the questions to be answered by upcoming studies.The hypotheses presented herein aim to encourage researchers in this field to solve the mechanisms behind the assembly of bunyaviruses and of the subcomponents of the virion.Currently, the surface structure of the virion is available for two genera in the family Bunyaviridae with striking differences in the overall glycoprotein organization.Future studies will hopefully also answer the very interesting question of whether the surfaces of other members in this virus family resemble either of the two known structures.In the future we also hope to see the structures of bunyavirus glycoproteins to be resolved by X-ray crystallography.Such a structure would complement nicely the currently available electron cryotomography data and determine whether the hypotheses presented herein are right or wrong.The structure-assembly of enveloped viruses is a highly complex process.In-depth knowledge of the interactions behind assembly may provide means to intervene in the formation of virions.Thus, this field also has potential applications in biomedicine, therefore making the results of such studies interesting to a broad range of researchers.

Fig. 1 .
Fig. 1.Hantavirus particle.(a) Schematic representation of the virion.(b) The GPC (ORF of M segment) is shown above, and mature Gn and Gc are shown below.The scissors in GPC indicate the cleavage sites of signal peptides, transmembrane (TM) helices are shown in black, the cytoplasmic tails (Gn-CT and Gc-CT) in dark grey and membrane in light grey.In mature proteins, the N-glycosylation sites are numbered based on HTNV GPC and an underlined label indicates a conserved Nglycosylation site.The signal sequence of Gc is included in the mature Gn; however, the faith of this TM helix is unknown.ER, Endoplasmic reticulum.

Fig. 2 .
Fig. 2. Comparison between hantavirus (TULV; Huiskonen et al., 2010) and HTNV (Battisti et al., 2011) and phlebovirus (RVFV; Huiskonen et al., 2009) spike structures.(a) The complete spike density for each virus.A hexameric complex was extracted from the whole densitogram of RVFV structure.(b) Slices through the spike density from the side.(c) Slices through the spike density from above.The HTNV density map, at 2.5 nm resolution, was kindly provided by Professor Michael Rossmann and Anthony Battisti.