Lassa virus-like particles were generated to contain the major immunological determinants of the virus, resembled native virions structurally, and were immunogenic in mice. Plasmid vectors well suited for high level expression of recombinant proteins in mammalian cells through combination of rational design and proven genetic elements have resulted in high yields of LASV VLP. These vectors afford the possibility of developing a VLP-based vaccine candidate in mammalian cell systems at low cost per dose, using transient expression technologies. Despite incorporation of all LASV proteins into VLP, both glycoproteins were present at significantly higher levels in most sucrose density fractions than either NP or Z (Figure 1). Incorporation of high levels of both glycoproteins in VLP may be beneficial in a vaccine platform, as these viral components alone have been shown to confer full protection against challenge with lethal doses of live LASV in non-human primates [17–21]. Yet, despite the high levels of glycoprotein incorporation into LASV VLP, addition of the nucleoprotein may be of critical importance in establishing more robust and long lived immunity against Lassa virus [19, 22]. Previous studies have demonstrated physical interaction between the glycoprotein complex, the Z matrix, and nucleoproteins during viral biogenesis [23–25]. Thus, these natural interactions are greatly beneficial since they result in the generation of VLP that package all viral immunogenic and protective determinants from a single set of transiently transfected recombinant LASV genes. In these studies we employed the human endothelial kidney cell line HEK-293T/17 for its high levels of transfectability, expression of recombinant proteins from human cytomegalovirus (hCMV) promoter driven gene constructs, and resulting yields of LASV VLP. During the course of this work, we have also established the value of using HEK-293T/17 as an indicator cell line. The profound morphological changes manifested by the cell line upon expression of LASV Z matrix protein is a good indicator of transfection efficiency and overall production levels of resulting VLP (Figure 2). Despite significant adverse metabolic effects on cells expressing LASV proteins and generating budding VLP, culture viability remained high (mean = 70%) at the time of harvest. This desirable aspect of mammalian cell culture-based production is beneficial in downstream purification processes, by reducing host cell components that must be eliminated from the final purified product, namely the cellular proteins, DNA, RNA, and lipids. Other expression platforms cannot be easily employed in the generation of LASV VLP where the glycoprotein complex precursor is used to incorporate processed GP1 and GP2. Truncated versions of the GPC precursor lacking the transmembrane domain have been generated in E. coli (unpublished data from the Viral Hemorrhagic Fever Research Consortium) and in baculovirus expression systems . In E. coli, the protein is neither glycosylated nor cleaved into GP1 and GP2 subunits. In insect cells, the protein is glycosylated but is not cleaved . Both expression systems lack the critical SKI-1/S1P subtilase responsible for co-translational processing of the LASV GPC precursor in mammalian cells . Despite the possibility of co-expressing the subtilase in heterologous systems to facilitate processing of GPC precursor, the glycosylation profile of GP1 and GP2 subunits may play a critical role in the structure and function of each protein in vivo. Thus, a mammalian expression system remains a highly attractive platform for the development of an arenaviral VLP-based vaccine.
We have determined in these studies that LASV VLP contain, in addition to the intended viral polypeptides, a plethora of host cell membrane proteins, presumably acquired during budding from the cell membrane or other intracellular lipid bilayer containing structures, such as the Golgi apparatus. A significant portion of the viral envelope protein content is made up of host cell glycoproteins, as determined by a broad glycan binding analysis performed on sucrose sedimented fractions (Figure 3B,C). The host cell glycoprotein composition varies along the gradient spectrum (Figure 3B). A similar pattern of cellular glycoproteins incorporated into LASV VLP was detected in purified particles generated from expression of Z alone or in combination with GPC and NP (Figure 3C). In Z+GPC or Z+GPC+NP VLP, a diffuse lectin binding pattern could be detected between 38 and 42 kDa which was absent from VLP that did not express the glycoprotein complex. This pattern was detected in addition to a prominent ~ 48 kDa cellular glycoprotein of unknown identity present in all VLP formats (Figure 3 C). The majority of detected cellular glycoproteins incorporated into LASV VLP ranged from 30 to greater than 220 kDa in mass. Recently, Moerdyk-Schauwecker et al. 2009  characterized the spectrum of mammalian host cell proteins incorporated into vesicular stomatitis virus (VSV), an enveloped virus, during viral biogenesis. In total, 64 proteins of host cell origin were identified via a proteomics approach coupled with mass spectrometry (MS). Of the 64 host cell proteins identified in these studies, 10 were glycoproteins . Although a similar study has not been performed for any member of the arenaviridae, it is likely that some common host cell proteins are packaged among a wide array of viral classes, and some of these proteins may even play functional roles during viral infection and replication. Characterization of the host cell protein profile in LASV VLP will be paramount in gaining regulatory clearance of an arenaviral pseudoparticle-based vaccine. The immunological and functional role of such proteins must be known in order to avert untoward side effects, such as autoimmunity and physiological disregulations.
We had previously characterized the gross glycosylation profile of LASV GP1 in the context of a soluble isoform (sGP1) of this viral protein . In the present studies, we characterized LASV VLP-associated GP1 and GP2 glycosylation patterns. Glycoprotein 1 associated with VLP generated essentially the same glycosylation pattern as sGP1, with only partial deglycosylation by Endo H, and insignificant processing by neuraminidase (Figure 4A). These results point to a heterogeneous array of glycans on the surface of GP1 that include some high mannose and branched oligosaccharides. Glycoprotein 2 displayed a more heterogeneous glycan array with a highly homogeneous high mannose and hybrid oligosaccharide content that accounted for approximately 8 kDa of the fully processed mass of the protein, based on the detection of a relatively sharp 30 kDa species upon treatment with Endo H (Figure 4B, lane 3). The remaining 7 kDa of glycan content could be removed by treatment of the protein with PNGase F, but not with neuraminidase (Figure 4B, lanes 2 and 4). A similar micro- and macroheterogeneity in both GP1 and GP2 N-linked glycosylation has not been characterized in native Lassa virions.
Through these studies, we have established that GP1 incorporated into LASV VLP is highly resistant to proteolytic digestion by trypsin (Figure 7A, lanes 4 and 5), despite 13 predicted trypsin recognition sites on the polypeptide backbone (ExPASy proteomics server tools, PeptideCutter ). Similarly, GP2 is resistant to digestion with trypsin, albeit to a lesser extent than GP1, even after solubilization of the pseudoparticle envelope with Triton X-100 (Figure 7B, lanes 4 and 5). The PeptideCutter tool in ExPASy predicted 25 recognition sites with high confidence in the GP2 polypeptide backbone. However, since the glycoprotein complex spike is the viral antigen most readily accessible to the innate immune system and to circulating serum proteases, it is likely that this molecule evolutionarily developed a significant level of proteolytic resistance in the structure-function relationship. It is of paramount importance to the virus that the critical components required for binding and fusion to permissive host cells be preserved. The specific glycosylation patterns on GP1 and GP2 may play a functional role in the observed resistance to proteolytic degradation. In the studies by Schlie et al. 2010 , Proteinase K protection assays performed on GP VLP also revealed partial resistance of the GP2 component against degradation by the protease, although solubilization with Triton X-100 in conjunction with protease resulted in complete digestion of the protein. Glycosylation of arenaviral glycoproteins is critical for protein stability, as unglycosylated GP1 and GP2 generated in E. coli are insoluble and require detergents, zwitterions, and reducing agents to remain in solution [11, 30], and deglycosylating mammalian cell generated GP1 generally produces similar results (unpublished data).
To characterize the structural compartmentalization of viral proteins in LASV we performed trypsin protection assays in the absence or presence of the anionic detergent Triton X-100 (Figure 7). In the absence of detergent, trypsin completely digested non-reduceable GP1 trimer, partially degraded unprocessed GPC, but had no effect of monomeric GP1 (Figure 7A, lane 4). A similar digestion pattern was obtained for GP2 (Figure 7B, lane 4). The addition of detergent to the reaction enhanced digestion of unprocessed GPC and had a minor effect on sensitivity of GP1 to the protease (Figure 7A, lane 5). Dissolution of the envelope by detergent resulted in more pronounced degradation of GP2 by trypsin, although a significant portion of the monomer could be detected (Figure 7B, lane 5). Only treatment of LASV VLP with Triton X-100 resulted in proteolytic degradation of both Z matrix and NP proteins. These results strongly support the model of a LASV VLP containing glycoprotein spikes on the surface of a lipid envelope with an internal matrix of Z protein containing the nucleoprotein component. We have shown that the viral proteins NP, Z, GP1 and GP2 can be co-expressed in VLP. Protein-protein associations appear to be an important aspect to the formation of VLP. Schlie et al. 2010  reported that a co-localization of NP, Z, and GP occurs near the nucleus. Similarly, Eichler et at. 2004  demonstrated that NP and Z co-localize in the cell. They also demonstrated that NP could be precipitated using an antiserum against Z and vice versa. Furthermore, Schlie et al. 2010  determined that NP did not influence the interaction of GP and Z, nor could an interaction between NP and GP be detected in the absence of Z in co-localization and immunoprecipitation experiments. However, pull down experiments performed by Schlie et al. 2010  demonstrated an association between Z and GP and Z and NP. Strecker et al. 2006  reported that Z myristoylation is important for binding to lipid membranes. Flotation experiments using wild-type Z protein and a mutant of Z at the myristoylation site showed that the mutant remains localized in the cytosol, whereas the wild-type associated with the membrane. Thus, the interactions between Z and the membrane and with GP and NP result in VLP formation with relevant proteins incorporated in virions.
Another structural component of native LASV virions are host cell ribosomes that are packaged during virion assembly, presumably for enhanced viral mRNA translation in the early stages of cellular infection. To determine whether LASV VLP containing any combination of Z matrix, GPC, and NP proteins mediated the ability to package cellular ribosomes, total RNA was isolated from pseudoparticles and analyzed by denaturing RNA gel electrophoresis (Figure 5). RNA was also isolated from the corresponding transfected cells and analyzed alongside VLP RNA. All VLP formats analyzed in these studies did not contain significant levels of the 28S and 18S ribosomal RNA species known to be critical components of mammalian ribosomes (Figure 5, lanes 2, 4, 6, 8, 10). In some analyses, RNA was purified from 1 mg of total purified VLP, and the entire purified nucleic acid fraction was analyzed by gel electrophoresis without distinct ribosomal RNA bands visible (data not shown). Despite the lack of rRNA detection in LASV VLP, all pseudoparticle formats analyzed in these studies contained significant levels of low molecular weight RNA species ~ 75 - 200 nt, that co-migrated with cellular 5S (120 nt) and 5.8S (160 nt) rRNA, and transfer RNAs (75 - 95 nt). It is reasonable to assume that in native VLP the incorporation of host cell ribosomes would result in the co-packaging of critical tRNAs for translation of viral mRNAs. Although in these studies the exact nature of the packaged RNA species was not characterized in detail, the results suggest that multiple RNA species of ribosomal origin are incorporated into VLP. To confirm that ribonucleoproteins were not incorporated into virions, we performed western blot analysis on VLP proteins using antibodies raised against U1 snRNP 70, La/SSB, and Ro/SSA. No ribonucleoproteins could be detected in pseudoparticles (data not shown). These studies also point to a critical presence of viral RNA polymerase and genomic RNA segments during replication for subsequent incorporation of host cell ribosomes into nascent viral particles. The lack of detectable ribosomes in LASV VLP represents a regulatory advantage for this platform as a vaccine candidate. Administration of pseudoparticles containing autologous ribosomes to vaccinees has potential to result in untoward immunological affects.
Despite the lack of detectable 28S and 18S rRNA in LASV VLP comprised of any combination of LASV proteins analyzed in these studies, pseudoparticles that contained GPC and/or NP in addition to Z matrix protein were morphologically similar to native virions (Figure 6B,C,D). These VLP were electron dense particles with punctuate inclusions and appeared to associate with highly electron dense subcellular organelles in the cytoplasm, possibly ribosomes despite their lack of incorporation into the pseudoparticle (Figure 6C,D). The size of mammalian ribosomes is approximately 20 nm, in line with the size of the particles associated with nascent LASV VLP imaged in these studies (Figure 6D). However, these subcellular structures could not be detected in VLP budding from the surface of cells transfected with Z matrix protein alone (Figure 6A), which appeared empty and containing only an envelope structure, as shown here and reported by others .
For immunizations, LASV VLP comprised of Z+GPC or Z+GPC+NP were formulated in PBS and used to immunize BALB/c mice, in a prime + 2 boosts schedule, 3 weeks apart, in the absence of an adjuvant, and administered by i.p. injection. After a single immunization some animals showed a low level IgG response to individual LASV antigens, with increasing mean antibody titers after each subsequent boost (Table 1). ELISA analysis of terminal IgG titers showed a clear difference in the response levels against GP1, and whole VLP between Z+GPC and Z+GPC+NP pseudoparticles (p = 0.004 and 0.026, respectively) (Figure 8A,B). VLP containing all three proteins induced a significantly higher response to the glycoprotein components compared to Z+GPC VLP, with a 15 fold overall increase in titer against both GP1 and GP2, despite a not quite significant statistical difference in the GP2 titers (p = 0.092). Likewise, the titers against whole Z+GPC+NP VLP were nearly 3 fold higher than to Z+GPC pseudoparticles (Figure 8A,B).
Lastly, we attempted to use LASV VLP as a diagnostic tool for the detection of viral protein-specific IgM and IgG in the serum of convalescent subjects, patients from the Lassa ward, contacts from patients who succumbed to Lassa fever, and individuals not known to have had the febrile illness at any given time in their lives. The LASV antigen binding profile of these sera was extensively characterized using highly sensitive and specific recombinant protein-based diagnostics under development by the Viral Hemorrhagic Fever Research Consortium. The overall poor level of correlation observed in human serum IgM (r = 0.3297; r2 = 0.1087) and IgG (r = 0.6284; r2 = 0.3949) binding profiles between LASV VLP and recombinant proteins in these studies was not surprising. Recombinant LASV proteins currently employed in diagnostic assays are generated in bacterial or mammalian cell systems, as outlined in Branco et al., 2008 , and Illick et al., 2008 . Individually produced, purified, and characterized proteins are used alone or in combination to coat high protein binding ELISA plates for determination of serum IgM and IgG binding profiles. Thus, it would be expected that protein-protein interactions known to play a role during viral biogenesis and in the formation of LASV VLP result in presentation of different epitopes and conformations than in counterparts generated as individual polypeptides. The known interactions between Z, GPC, and NP proteins in a VLP format likely mask the presentation of relevant epitopes to which a given individual may have generated IgM and IgG. As a result, native presentation of antigens in the context of a VLP, even in the presence of low levels of the membrane solubilizing detergent Tween 20, will likely not result in disruption of protein interactions necessary for the detection of epitope-specific serum antibodies. This is supported by the fact that all five NP-specific mAbs used in this analysis detected and captured recombinantly expressed NP in solution (Figure 9F), albeit at different levels. In combination, these results strongly suggest that LASV proteins in the context of a VLP display epitopes that possibly mimic native conformation and presentation. These observations further support the use of LASV VLP as a vaccine platform by supplying a quasi-native antigen, thus allowing the innate and adaptive immune systems to preferentially target epitopes relevant for immune protection against the virus. In addition, the use of pseudoparticles in clinical assays may offer advantages over the use of recombinantly expressed individual proteins. Immune responses to LASV VLP may be directed against epitopes that are best or exclusively displayed in the context of a quasi native particle containing proteins assembled in a manner similar to functional viral biogenesis.
VLP have gained significant momentum in the past decade as premier vaccine platforms. The approval of Merck & Co., Inc.'s Gardasil(r) (Human Papillomavirus Quadrivalent [Types 6, 11, 16, and 18] Vaccine, Recombinant) by regulatory agencies heralded a new era in vaccinology, demonstrating that VLP are immunogenic, safe, and well tolerated in humans, and confer nearly complete protective immunity against homologous viral strains in canine models [32–38]. ENGERIX-B [Hepatitis B Vaccine (Recombinant)] is a recombinant VLP-like hepatitis B vaccine developed and manufactured by GlaxoSmithKline plc. These "Dane" particles, generated in yeast strains, are comprised of HbsAg and yeast phospholipids, and are subsequently harvested by gradient centrifugation and properly disulfide-linked in vitro . These particles are highly immunogenic, safe, well tolerated, and very efficacious.
VLP-based vaccine candidates have also been developed and tested for their efficacy in preventing a wide array of viral conditions, such as Influenza [40–44], Ebola [45, 46], Marburg [45, 47], West Nile virus , Dengue , Respiratory Syncytial Virus (RSV) , HIV [51–56], and Hepatitis C virus [57–59], and the most recently reported case of Chikungunya . VLP platforms currently being evaluated toward clinical licensure include Novavax's trivalent seasonal influenza vaccine. In recent Phase II clinical trials the vaccine was well tolerated and safe in adults age 60 and older and in healthy volunteers 18 to 48 years of age [61, 62]. Thus, it is reasonable to employ similar strategies to develop a vaccine platform based on VLP that contain all the relevant immunological determinants that are known to confer protective immunity against this viral hemorrhagic fever. Studies are currently ongoing to determine the in vivo efficacy of LASV VLP in relevant in vivo models.