Background
Dengue viruses (DENV) are mosquito-borne flaviviruses responsible for dengue fever and dengue hemorrhagic fever [
1,
2]. DENVs are endemic in over 125 countries and its global distribution is expected to expand due to urbanization and other environmental factors that favor the mosquito vector [
3]. The DENV complex consists of 4 antigenically distinctive serotypes (DENV1–4). Infectious DENV particles contain a positive stranded RNA genome and capsid proteins surrounded by a lipid envelope, which has two membrane glycoproteins designated as pre-membrane (prM) and envelope (E). The outer surface of the mature, infectious virus has a smooth surface covered by E protein homodimers [
4]. On the viral surface, units of three E protein homodimers assemble into raft-like structures and 30 rafts are packed tightly to create a protein coat with icosahedral symmetry. The E protein mediates viral attachment and entry into cells. Additionally, the E protein is the major target of neutralizing and protective human antibodies. Recent studies have demonstrated that most human antibodies that strongly neutralize DENVs bind to quaternary structure E protein epitopes containing regions from 2 or more proteins packed on the viral surface [
5‐
8]. In other words, many human antibodies target E protein epitopes displayed on the viral surface but not on individual E protein subunits. Soluble E protein subunit antigens are, therefore, considered to be poor vaccine immunogens because they lack higher order protein structures and epitopes displayed on the intact virus. When flavivirus prM and E proteins are co-expressed in cells, these proteins assemble into virus-like particles (VLPs) that are secreted from cells. Flavivirus VLPs have been produced in a wide variety of expression platforms such as plants, insect cells, bacteria, yeast and mammalian cells. Flavivirus VLPs, which do not contain a viral nucleocapsid or genomic RNA, are smaller than intact virions [
9‐
12]. VLPs, often, share structural similarities and physicochemical features of the native virus, but are superior in terms of safety, and ease of production and purification [
10]. VLPs have proven to be effective vaccine antigens in preclinical or early stage clinical studies with different enveloped viral pathogens such as West Nile virus, Tick-borne encephalitis virus, Japanese encephalitis virus, Zika virus, Chikungunya virus, Influenza virus and Dengue virus [
10‐
17]. For DENV VLPs, no data is available on the display and presentation of epitopes of varying complexity recognized by human neutralizing antibodies. Some structural information is available for tick-borne encephalitis virus (TBEV) VLPs [
12]. The low-resolution cryo-EM image based reconstruction of TBEV VLPs indicates that the number and organization of E proteins on the VLPs are different from the infectious virus. Even though TBEV VLPs are heterogeneous in size, investigators analyzed a uniform population of particles that were smaller than virions and concluded that VLPs contain 30 E-homodimers assembled in a
T = 1 icosahedral lattice [
12]. It is unclear if quaternary structure antibody epitopes, especially those containing residues from different adjacent E homo-dimers, are displayed similarly on VLPs and the larger infectious virions [
12].
In this study we describe in detail the properties of epitopes present on DENV VLPs of all 4 serotypes using a large panel of well-defined human mAbs and immune sera from dengue patients. We directly compare the antigenic features of DENV VLPs to whole virions and explore their use as tools in serologic assays. Our results show the equivalence between viruses and VLPs in a setting that is particularly relevant for flavivirus vaccine development, diagnostics and understanding the difference between virus and VLPs.
Discussion
The immunogenic properties of DENV VLPs have been analyzed in animal models by several groups. However, little is known about E-protein organization on the surface of VLPs [
10,
11]. VLP epitope display has not yet been directly compared to that of virus particles and outside the use as vaccine antigens, VLP applicability as serological tools has been underappreciated. The terms subviral particle and VLP are often used to describe the same particle format, but can be considered different. In this study we refer to particles that do not contain a nucleocapsid and are produced after the expression of prM-E as VLPs. In our opinion, the term subviral particle is better used to describe the formation of smaller spherical particles obtained as a side product of wildtype infections.
DENV1–4 VLPs produced in mammalian cells efficiently display epitopes that have mapped on virus particles. TS and CR monoclonal antibodies isolated from different DENV patients with varying infection history bound to the VLPs indicating that native epitopes are present on the VLPs. In the low resolution cryo-EM structural data available on flavivirus virus-like particles or subviral particles, 30 E-dimers are assembled in a
T = 1 icosahedral lattice, different from the 90 E-dimers found on mature virus particles organized in a
T = 3 symmetry [
12]. Our analysis here demonstrates that overall differences in arrangement of E-dimers on VLPs versus virions and the absence or presence of a nucleocapsid do not alter the display of most quaternary epitopes targeted by human antibodies.
The DENV VLPs were found to have a similar size distribution as the TBEV subviral particles. With particle diameters of ~ 29–34 nm, the VLPs are considerably smaller in size than natural virus particles (~ 50 nm) [
4,
12,
22]. The minor VLP size differences between serotypes could be attributed to maturation dissimilarities, however, as the highly concentrated VLPs are not perfect spheres and have structural irregularities, the measurements of diameters might be error prone. Future structural studies are needed to confirm the observed differences in DENV VLP diameter.
Human mAb 1F4 is DENV1 specific and although it only binds the virus and not the E-monomer, its footprint has been mapped within one E-monomer and not across neighboring E-proteins within the homodimer [
7]. Despite differences in size and prM content, 1F4 bound equally well to DENV1 VLPs and purified virus. Similar results were found for 2D22 (DENV2), 5 J7 (DENV3) and 5H2 (DENV4). 2D22 is a potent neutralizing antibody and binds epitopes that span across the E-protein dimer, blocking envelope reorganization necessary for viral fusion [
5]. DENV3 specific mAb 5 J7 has a footprint that spans across neighboring E-dimers including residues at the EDI/EDII hinge region [
23]. Binding of 5 J7 to DENV3 VLPs strongly indicates that E-dimers interact and form poly-dimer structures similar to the rafts found on virion surfaces. For long, 5H2 was the only DENV4 specific highly neutralizing mAb of which the binding footprint has been mapped. This chimpanzee derived mAb docks within EDI of E [
24]. It does not bind monomeric E-proteins, but efficiently binds both DENV4 virus and VLPs. Recent studies have identified new DENV4 TS mAbs that bind E-protein in ranging complexity [
8]. DV4 126 and DV4 131 are highly neutralizing and their quaternary footprint was mapped to the EDI/EDII hinge region. Just like 5 J7, these mAbs do not bind monomeric E-proteins, but recognize virus and VLPs.
We used 1F4, 2D22, 5 J7 and 5H2 to normalize the binding of all the other mAbs to both virus and VLP, allowing for a relative comparison between epitope availability on virus vs VLPs. Relative to these mAbs we see that the EDIII binding mAbs 12C1.5 and 3H5 bind better to VLPs, indicating that the EDIII might be more exposed on VLPs. The quaternary CR EDE epitopes are relatively better presented on DENV2–3 virus particles. In particular EDE1-C10 binding has a strong preference for virus particles, due to the absence of binding to VLPs of all serotypes. It remains unclear why C10 does not bind VLPs, while the other EDE mAbs, that share a very similar footprint, bind efficiently.
The maturity of virus particles and VLPs affects epitope display as exemplified by the fusion loop binding CR mAb 4G2. During virus replication, the pr peptide associates with the fusion loop in E, shielding it from premature low-pH induced fusion with host cell membranes [
25]. Our analysis indicates that the fusion loop is efficiently exposed on DENV4 VLPs, which is in accordance to the undetectable prM levels in DENV4 VLPs. Differences in prM levels between DENV2–3 VLPs and virus might have similar steric hindrance effects on the EDE mAbs. However, this requires further structural analysis. In general, epitopes present on the fusion loop domain, EDIII and DI/DII/DIII spanning regions are all represented on DENV VLP surfaces.
The footprints of many mAbs have been mapped on dimers generated from soluble monomeric E proteins. Soluble E monomers crystallize into dimers and do not contain prM. Even though it is assumed that the crystallized dimers are structurally similar to E-dimers found on virus particles, they might differ when part of complex multiprotein complexes on VLP envelopes. Future studies should focus on DENV VLP protein structures to answer these questions.
The VLPs and virus particles were captured by mAbs 4G2 and 1 M7, suggesting a bias for fusion loop exposed particles. However, using other capturing mAbs, the bias would be skewed towards particles displaying that specific epitope. Using other well-characterized mAbs, we have shown that DENV VLPs display an epitope landscape very similar to that found on virus particles. Depletion studies using VLPs as a depletion antigen translate this finding to polyclonal sera of DENV infected patients. Secondary infected serum was efficiently depleted from all TS and CR (non)-neutralizing antibodies. Serum depletions are usually performed using infectious purified virus antigen conjugated to microbeads. However the use of VLPs has practical benefits over the use of purified antigen. Issues of infectious virus production and purification and virus leaching into the serum during handling are not present when using VLPs. Additionally, VLPs are more easily genetically adapted and are therefore valuable tools in flavivirus pAb and mAb mapping and serology.