Temperature-dependent folding allows stable dimerization of secretory and virus-associated E proteins of Dengue and Zika viruses in mammalian cells

Dengue and Zika are two of the most important human viral pathogens worldwide. In both cases, the envelope glycoprotein E is the main target of the antibody response. Recently, new complex quaternary epitopes were identified which are the consequence of the arrangement of the antiparallel E dimers on the viral surface. Such epitopes can be exploited to develop more efficient cross-neutralizing vaccines. Here we describe a successful covalent stabilization of E dimers from Dengue and Zika viruses in mammalian cells. Folding and dimerization of secretory E was found to be strongly dependent on temperature but independent of PrM co-expression. In addition, we found that, due to the close relationship between flaviviruses, Dengue and Zika viruses E proteins can form heterodimers and assemble into mosaic viral particles. Finally, we present new virus-free analytical platforms to study and screen antibody responses against Dengue and Zika, which allow for differentiation of epitopes restricted to specific domains, dimers and higher order arrangements of E.

carries the hydrophobic glycine-rich fusion loop (FL) that is highly conserved in related flaviviruses 11 . In addition, DII provides the surface where some of the main interactions that drive E antiparallel dimerization occur 12 . DIII is an Ig-like β-barrel domain that shows variability among the different serotypes and has been implicated in the initial interaction to cellular receptors 13 . In addition, the 4 loosely packed peptide strands that connect DI and DII form a functional domain named Hinge region, which provides the flexibility needed for E trimerization 14 . In this pH-induced conformational reorganization of the viral surface, DII drives E reorientation from a horizontal antiparallel dimer into a vertical parallel trimer exposing the FL outwards to initiate the fusion process 15 .
It has been proposed that during viral assembly the PrM protein interacts with the nascent E protein to assist its proper folding and remains bound to the fusion loop during its transit through the secretory pathway, thus preventing premature fusion of the viral particle with the ER membrane 16 . As the nascent virion enters the Golgi network, Pr is cleaved by the host-encoded furin protease and remains attached to E until the virion is secreted 17 . Upon release from the cell, the neutral pH of the extracellular environment stabilizes E dimers and promotes dissociation of Pr 18 .
As described for both DENV and ZIKV, E protein is also the main target of the human antibody response upon infection 19,20 . Neutralizing epitopes have been described on all three domains but the immune response is heavily dominated by poorly neutralizing and highly cross-reactive antibodies against epitopes on DI and DII (DI/DII) 21 . On the other hand, antibodies against DIII are generally described to be highly neutralizing and usually serotype specific 22,23 . However, the contribution of these antibodies to the overall immune response in infected individuals appears to be very limited 24 .
The symmetric organization of E proteins on the virus creates new quaternary epitopes involving two or more E molecules found only on the surface of the intact virion. Recent data indicates that these epitopes are targeted by the majority of strong neutralizing antibodies found in the sera of infected patients 25 . Recently, a new type of potently neutralizing antibodies, recognizing epitopes restricted to the interface between two E proteins in the head-to-tail dimer conformation, has been described 26 . These dimer-dependent epitopes (named E-dimer epitope, EDE) do not require higher-order structural arrangements beyond the dimeric conformation of E, and appear to be highly conserved among flaviviruses 27,28 . Antibodies against these epitopes were shown to have strong cross-neutralizing activity not only among DENV serotypes, but also against ZIKV 29,30 .
The development of an efficient vaccine against DENV and ZIKV has become a high priority. One of the main obstacles for developing viable DENV vaccines is the risk of enhancing infection through the mechanism known as antibody-dependent enhancement of infection (ADE). In ADE, opsonization of viral particles by antibodies in non-neutralizing conditions leads to an increased Fcγ receptor-mediated uptake by monocytes and macrophages, which results in an increased and more severe infection 31 . Recently, cross-reactive antibodies against conserved epitopes in DI and DII, particularly those against the FL, were shown to enhance infection of other related flaviviruses like ZIKV and YFV [32][33][34] . The discovery of complex quaternary epitopes on DENV E protein dimeric structure, some of them conserved for instance in ZIKV, and recognized by antibodies with strong neutralization activity, has opened the possibility of developing a safe universal vaccine. However, the isolation of the precise epitopes to induce strong cross-neutralizing antibodies while avoiding induction of unwanted anti-DI/DII antibodies is a complex and yet unresolved task 35 .
With the aim of understanding and characterizing the dynamics involved in E dimerization, we describe a successful covalent stabilization of E dimers of DENV and ZIKV, along with a detailed biochemical analysis of the expression, secretion and folding in mammalian cells that provide a platform to facilitate the development of E dimer-based genetic or subunit vaccines. Moreover, we demonstrate that E dimerization within the ER lumen is independent of PrM but highly dependent on temperature, and show that because of their close structural and functional characteristics, DENV and ZIKV E proteins can interact to form heterodimers and assemble into mosaic viral particles.

Results
Temperature-dependent secretion of a covalently stabilized dimeric sE. Given that some of the most potent cross-neutralizing antibodies against DENV and ZIKV recognize complex quaternary epitopes that only exist within the context of the E dimer 28 , we attempted to stabilize the dimeric conformation of recombinant E from DENV and ZIKV expressed in mammalian cells. We have previously reported a detailed analysis of the secretory profile of recombinant E derived from the different DENV serotypes in mammalian cells and shown that DENV2 E ectodomain (2sE) undergoes different folding depending on temperature 36 . Here we have further investigated the role of temperature and PrM in the production and secretion of sE from DENV and ZIKV in transfected HEK293T cells, using constructs encoding sE alone or sE and the viral proteins PrM or only M (Fig. 1a). In addition, we introduced an Ala to Cys mutation in DII (Ala259 in DENV serotypes 1, 2 and 4, Ala257 in serotype 3 and Ala264 in ZIKV) because in the dimeric structure that residue is positioned just opposite of itself in the anti-parallel monomer (Fig. 1b). We reasoned that, if E undergoes transient dimerization in the ER, covalent stabilization through a disulphide bridge should be observed.
Non-reducing western blot (WB) of intracellular and secreted material from transfected HEK-293T cells incubated at either 37 °C or 28 °C (see protocol in Supplementary Fig. S1) are shown in Fig. 1c and d, respectively. At 37 °C, 3sE was secreted both as a monomer from wild-type (wt) constructs and as a covalent dimer (cvD) from constructs with the A257C mutation (Fig. 1c). Secretion and dimerization were independent of the presence of M or PrM as sE-wt and sE-cvD were equally produced and secreted in the presence or absence of PrM. In contrast, wt and cvD of both, 2sE and ZsE were not secreted regardless of the presence of M or PrM. A smear of disulphide bonded oligomers were detected intracellularly that most likely represent folding intermediates. At 28 °C (Fig. 1d), however, ZsE followed a secretory profile similar to 3sE and was efficiently produced and secreted both as sE-wt and sE-cvD, while 2sE was only secreted as disulphide-bonded dimers (2sE-cvD). Noteworthy, in both cases, the results were independent of the presence of PrM. 3sE-cvD showed similar dimerization and secretion profiles at 28 °C and 37 °C.
PrM-independent but temperature-dependent dimerization and secretion of cvD formats were also observed at 28 °C for DENV4 (4sE) and DENV1 (1sE) even though secretion of 1sE-cvD was less efficient ( Supplementary  Fig. S1). In addition, these characteristics were also independent of the mammalian cell line used ( Supplementary  Fig. S1).
Correct folding and assembly of DENV and ZIKV sE-cvD was confirmed in ELISA and by cytofluorimetry of cells displaying dimers on the cell membrane. For ELISA, in vivo mono-biotinylated and secreted dimers of DENV 2, 3 and 4 (DENV1 sE-cvD is not secreted in sufficient amounts to perform a normalized assay) and of ZIKV were obtained from culture supernatants of transfected HEK-293T cells and immobilized on avidin-coated plates. We then tested different mAbs recognizing E conformational epitopes: 4G2 (fusion loop specific 37,38 ) and 4E5A (DENV DIII specific 39,40 ) recognize epitopes found on monomeric E, while EDE1-C10, EDE2-B7 (against E dimeric epitopes EDE1 and EDE2, respectively 27 ), 2D22 (recognizing a DENV2-specific dimeric epitope 41 ) and 5J7 (a DENV3-specific mAb that binds to 3 different E proteins on the viral surface 42 ), bind to quaternary epitopes involving more than one E molecules. In addition, the DENV1-specific mAb 1F4 that binds to DI and the hinge region of a single E molecule, but only when present on the viral surface 43 , was also included. As shown in Fig. 2a, the EDE-specific mAbs EDE1-C10 and EDE2-B7 clearly detected cvD of 2sE, 3sE and 4sE while ZsE-cvD was detected by EDE1-C10 but not by EDE2-B7. As expected, 4G2 was positive in all cases, while 2D22 was positive only on 2sE-cvD and 4E5A recognized DIII of DENV but not of ZIKV. Not surprisingly, mAbs 1F4 and 5J7 were negative for all constructs, since the former binds only to DENV1, while the latter recognizes an epitope that requires a higher order arrangement of dimers.
The same set of reagents was used for cytofluorimetric analysis of cell membrane display of sE-cvD ((m) sE-cvD), achieved by fusing the trans-membrane and cytosolic tail of the type-I trans-membrane protein MHC-Iα at the C-terminus of sE, as previously reported 36 . Transfected cells were incubated with the set of mAbs following expression of (m)sE-cvD ( Fig. 2b) or (m)sE-wt ( Supplementary Fig. S2) at both, 28 °C and 37 °C. The profiles of reactivity resemble those of ELISA with three significant differences: i) the dimer-specific antibodies were able to detect the membrane displayed (m)sE-wt of the four DENV serotypes and ZIKV, indicating that they were present in the dimeric form, ii) mAb 1F4 recognized DENV1 E on both wt and cvD formats and iii) mAb 5J7 reacted similarly with DENV3 sE-cvD and sE-wt. An additional EDE1 specific mAb (EDE1-C8) was also included showing a profile similar to EDE1-C10. In agreement with recent reports, we found EDE1-type epitopes were conserved in ZIKV, while EDE2-B7 was not able to recognize ZIKV sE-cvD. Antibodies against EDE2-type epitopes have been shown to have a significantly reduced affinity towards ZIKV as binding is altered due to a change in the orientation of the glycan on DI (N154) and the lack of N-glycosylation in DII 28,30 . This was demonstrated after analyzing a membrane displayed non-glycosylated ZIKV sE-wt mutant (Δglyc, T156I) 44 , which showed positive reactivity with mAb EDE2-B7 ( Supplementary Fig. S3). As expected, reactivity to EDE1 mAbs was not affected.
We therefore concluded that both secreted and membrane bound sE-cvD, correspond to the dimeric structure present in mature viral particles. Viral particles with covalently stabilized E-dimers. We next attempted incorporation of the full-length covalent dimeric E into complete viral particles. For that purpose we generated pseudoviral particles by packaging a West Nile virus replicon encoding a reporter EGFP (WNV-rep) 45 with a construct containing ZIKV-derived structural proteins C-PrME carrying the wild type sequence (E-wt) or the A264C cvD mutation (E-cvD). If assembled, particles with the cvD mutation should not be infective, as covalent E dimers would not allow E trimerization required for fusion of the viral envelope with endosomal membranes (Fig. 3a,b). Viral particles were obtained with both, E-wt and E-cvD mutant, as shown by RT/PCR amplification of replicon RNA from culture supernatants of cells co-transfected with E-wt or E-cvD and the WNV-rep, but not from cells transfected with the WNV-rep alone (Fig. 3c). While the production of pseudoviruses with E-wt was higher at 37 °C in some experiments (yet with no significantly different when comparing production at both temperatures), the yield of particles carrying the cvD mutation was significantly higher at 28 °C. Furthermore, the expected monomeric and dimeric forms of, respectively, E-wt and E-cvD were confirmed by non-reducing WB analysis of particles released in the culture supernatants at 28 °C (Fig. 3d). As expected, all the protein recovered in the E-cvD pseudoviral particles preparation was dimeric.
In contrast to pseudoviral particles with E-wt, those with E-cvD were not infective when tested in Vero, HuH-7 or HEK-293T cells (Fig. 3e). As previously mentioned, this could be due to impaired fusion with endosomal membranes or, alternatively, to impaired interaction with cellular receptors. In this latter case, E-cvD particles would be able to infect K562 cells when opsonized with anti-DI/DII antibodies, as Ab-mediated entry would be independent of viral receptors. As shown in Fig. 3f, while infectivity of particles packaged with E-wt was enhanced by anti-DI/DII antibodies, infectivity of those packaged with E-cvD was not, consistent with lack of E protein trimerization within endosomes due to the presence of the disulphide bond.
To our knowledge this is the first mutant pseudovirus with a dimeric covalent E protein described. The results indicate that virus particles with the E protein "frozen" in the dimeric conformation are permissive for assembly in the ER and compatible with trafficking through the secretory pathway.

Hetero-dimerization of DENV and ZIKV sE.
Since all DENV and ZIKV sE harboring the cvD mutation were capable of forming and secreting dimers, we then asked whether interactions between sE belonging to different viruses were viable. To address this, we produced sE-cvD constructs with two different tags to allow discrimination of each monomer: 3sE was tagged with BAP (Biotin Acceptor Peptide, which can be biotinylated in vivo 46,47 ) while 1sE, 2sE, 3sE (as a control), 4sE and ZsE were tagged with SV5 48 . Homo-and hetero-dimers were detected based on the different tags and the different gel migration (schematically shown in Fig. 4a). As shown in Fig. 4b, 3sE formed covalent hetero-dimers with 2sE, 4sE and ZsE that were secreted from HEK-293T cells. 1sE also formed hetero-dimers, although less efficiently (Fig. 4b, right panel).
These results were further confirmed by cytofluorimetry, after co-expression of the membrane-displayed (m) sE-cvD of each DENV serotype and ZIKV with the SV5-tagged secretory version of 3sE-cvD. Cells become positive for SV5 only when the secretory protein forms a hetero-dimer with the membrane-bound version (Fig. 4c). As expected, anti-SV5 staining was not observed when the secretory construct was expressed alone. Interestingly, although less efficiently in particular for DENV serotypes 1 and 4, hetero-dimers were also detected with the sE-wt proteins, while for ZIKV it was almost undetectable (Fig. 4d). This result confirms that hetero-dimerization  and -cvD (b) constructs and the WNV replicon used to produce pseudoviral particles, tested for infectivity (EGFP expression) on Vero cells. Images correspond to representative experiments (Bar, 30 μm). (c) Yield of pseudovirus determined by RT/PCR of the replicon-derived RNA, on samples obtained with ZIKV E-wt packaging construct with (for each temperature, n = 7; t = 1,863 df = 12) and without (for each temperature, n = 6) replicon, or ZIKV E-cvD packaging construct with (for each temperature, n = 6; t = 4,896 df = 10) and without replicon (for each temperature, n = 6), supernatant of cells transfected only with the WNV replicon was used as control (for each temperature, n = 6), data is represented as mean ± s.d. (d) Left panel, non-reducing Western blot of E protein from culture supernatants of cells transfected with WNV replicon and ZIKV E-wt or E-cvD packaging constructs, developed with mAb 4G2. Right panel, western blot of the corresponding sE SV5-tagged proteins developed with anti-SV5; migration differences correspond to the effect of the SV5 tag. (e) Infectivity of pseudoviral particles produced with ZIKV E-wt or E-cvD packaging constructs on Vero (for both preparations, n = 5), HuH-7 (for both preparations, n = 5) and HEK-293T cells (for both preparations, n = 5), determined as EGFP positive cells at 48 h post-infection (All cells were infected using 2 × 10 4 viral particles as quantified by RT/PCR), data is represented as mean ± s.d. *, undetected. (f) Infectivity of same pseudoviral particles as in (e) on K562 cells in the absence of serum (for both preparations, n = 5) or in the presence of an anti-DI/DII serum (for both preparations, n = 5) or a negative control pre-immune serum (for both preparations, n = 5), data is represented as mean ± s.d. *, undetected. takes place as a consequence of spontaneous interactions between different sE proteins, and is not imposed by the cvD mutation.

E mosaic viruses.
The hetero-dimerization results raised the possibility for distinct E proteins to assemble into viral particles with E mosaic structures. We thus investigated this further, asking whether it was possible to obtain pseudoviruses with the WNV-rep packaged with two different E proteins derived from ZIKV and DENV. Infective pseudoviral particles (revealed by Vero cells infection) were obtained with a full ZIKV construct (C Z -PrME Z ) but not with a chimeric one containing ZIKV C protein and DENV2 PrME (C Z -PrME D2 ) (Fig. 5a, panels 1, 2). Transfection of the ZIKV C protein in trans with respect to DENV2 PrME did not produce  infective particles, nor did transfection of the replicon construct alone (Fig. 5a, panels 3 and 5). In contrast, upon co-transfection of ZIKV C-PrME with DENV2 PrME, infective particles were produced (Fig. 5a, panel 4) although with a reduced efficiency compared to those obtained with only ZIKV proteins (Fig. 5b).
To test whether mosaic viral particles were produced, we used an ADE assay in which infection was mediated by antibodies specifically recognizing DENV2 E protein. By using antibodies highly specific for DENV2 (such as those directed against DIII 20 ) infection should be enhanced for mosaic particles but not for those with only ZIKV E protein (Fig. 5c). As a control, antibodies against domains DI/DII cross-reacting with both DENV and ZIKV E proteins should produce ADE on both types of particles 20 (Fig. 5c). As shown in Fig. 5d, particles produced from cells co-expressing the two E proteins were infective in K562 cells when opsonized with the DENV specific anti-2DIII serum, while the ones with only ZIKV E protein were not. Conversely, the cross-reacting anti-DI/ DII antibodies produced ADE on both type of particles, while, as expected, a negative control serum did not (Fig. 5d). Further confirmation was obtained with mAb 4E5A, which is specific for DIII of DENV. As shown in Fig. 5e, 4E5A promoted infection of K562 cells of mosaic particles containing DENV E but not of control particles produced with only ZIKV protein. As expected, mAb 4G2 that recognizes the highly conserved FL, induced ADE on both types of particles. Thus, while DENV2 PrM-E protein alone was not able to package WNV-rep, it did get in part incorporated when co-expressed with the ZIKV proteins C-PrME, producing particles with a mosaic E composition.

Discussion
Given the importance of complex quaternary epitopes for inducing potent neutralizing antibodies and based on our previous work regarding the secretory profile of DENV sE from mammalian cells 36 , we set out to analyze dimerization of DENV and ZIKV E protein.
Although each of the proteins tested showed a characteristic secretory behavior, we found in all cases that intracellular dimerization of newly synthesized sE was highly dependent on temperature but independent of the presence of PrM.
At 28 °C, dimerization was highly favored for all DENV serotypes and ZIKV, resulting in complexes that could traffic and be secreted from mammalian cells. We could address this biochemical property of E because of the design of the Ala-Cys mutation in the alpha helix B (αB) of DII on the inner surface of the E monomer. In the dimeric configuration, the side chain of the inserted cysteine expands outwards, directly facing itself on the antiparallel monomer, which resulted in dimers covalently stabilized by a disulphide bond. This mutation was permissive for dimerization and secretion of all tested proteins. In addition, the mAbs included in the analysis allowed us to probe conformation of the structural domains as well as higher order interactions that mimic the quaternary structures found on the mature viral particles.
We showed that incubation at low temperatures was essential not only for secretion of the stabilized E dimers, but also for proper E folding. This observation is in line with recent reports showing that dengue virions suffer structural reconfiguration at temperatures above 33 °C 49,50 . Our data represent, to our knowledge, the first report extending this temperature-dependent behavior to the secretory E protein from DENV serotypes and ZIKV. This information could be relevant when designing subunit vaccines.
We have previously reported that DENV serotypes showing sE poor secretory phenotypes were misfolded at 37 °C but properly folded at 28 °C 36 . Our data indicate that, while proper folding is heavily influenced by temperature, secretion of recombinant E proteins depends on their capacity to form relatively stable dimers as they transit through the ER and the secretory pathway. In fact, while for DENV3 and DENV4 sE secretion was observed for both wt and cvD forms regardless of the temperature, for DENV1 and ZIKV induction of proper folding at 28 °C was required and sufficient to achieve secretion of the wt protein. In contrast, secretion of DENV2 sE was only possible when the dimer was stabilized by the disulphide bond (despite being properly folded at 28 °C 36 ) suggesting that wt dimers are relatively unstable and only allowed to traffic through the secretory compartment when covalently stabilized. This is consistent with recent structural data indicating that the surface of DENV2, and E dimeric interactions by consequence, is significantly more dynamic when compared to the other DENV serotypes 41 , which might explain the low stability of the 2sE-wt dimers.
Previous reports have shown enhanced production and stability of DENV and ZIKV virions at relatively low temperatures 51,52 . However, it is difficult to extrapolate our observations on DENV and ZIKV E proteins and speculate on the effect of temperature on the dynamics of the viral cycle. Interactions with the ER membrane, as well as between viral and host proteins are likely to play pivotal roles on the propagation of the virus within the arthropod vector (28 °C) or the human host (37 °C) in the context of infection, and further studies are needed to understand the role of temperature in these scenarios.
The membrane display experiments further confirmed the role of temperature in proper folding of sE. sE-wt anchoring to the cell membrane favored stabilization of E dimeric conformations, most likely by keeping monomers in close proximity and thus resembling the dynamics that take place on the viral surface. Under these conditions (m)sE-wt constructs showed the same temperature-dependent behavior and structure as (m)sE-cvD constructs. These results indicate that dimerization of sE-wt is structurally equivalent to that of sE-cvD, and that the Ala-Cys mutation stabilizes the spontaneous interactions between E monomers.
With our set of reagents (both in ELISA and cell membrane display assays) we were able to discriminate the reactivity of antibodies based on different levels of structural complexity, from single domain specificity, like 4E5A on DIII or 4G2 on DII, to complex quaternary serotype-specific requirements, like 2D22. Moreover, reactivity of DENV3-specific mAb 5J7, which requires three different E molecules of two adjacent E dimers on the raft arrangement on the viral surface 41 , indicates that both (m)sE-cvD and (m)sE-wt are able to mimic complex structural arrangements beyond the dimeric E conformation. It is possible that assembly of quaternary epitopes is induced by the antibody upon binding, thus stabilizing the dimeric conformation (or higher structural arrangements) of (m)sE-wt proteins. This kind of antibody-mediated stabilization has been previously described for DENV sE proteins 27 .
Scientific RepoRts | 7: 966 | DOI:10.1038/s41598-017-01097-5 Surprisingly, while temperature was found to be essential for sE dimerization and secretion, this was not the case for co-expression of PrM proteins. Many reports have established that PrM are important to assist folding of protein E in a chaperone-like manner 16,53 . Even though there is compelling evidence highlighting the importance of PrM in preventing premature inactivation of E in the context of the whole viral particle 54 , the data regarding its chaperone-like function for the production and secretion of soluble E proteins are less convincing. In fact, previous studies have shown secretion of properly folded E in absence of PrM 36,55,56 , and our results indicate that this is also the case for soluble E dimers. This is important when evaluating E-based subunit or genetic vaccines against DENV or ZIKV, since most of them include PrM on the basis that it is required to achieve proper E folding, despite the fact that it has been proven that anti-Pr antibodies are rarely neutralizing and mostly enhance infection through ADE 57 . As proper quaternary E structures were obtained in absence of PrM, our data question its chaperon-like function and provide further evidence supporting its exclusion from future vaccine formulations.
The ZIKV E-cvD protein was capable of packaging the WNV replicon to form and secrete pseudoviral particles in which all E protein was present in the covalent dimeric form. Interestingly, our results show that, contrary to wild type pseudoviruses but in agreement with the results with the recombinant sE, production of particles with covalently dimerized E was significantly higher at 28 °C, further supporting the role of temperature in E-cvD folding. In addition, this shows that the Ala-Cys mutation and the consequent disulphide bond are compatible with the higher order arrangements present on the viral surface. The data indicate that E dimerization occurs shortly after translation and that transport and secretion of full particles with covalently stabilized dimers are allowed; it also suggests that the "spiky" arrangement of E on newly assembled immature particles within the ER lumen is not strictly required 58,59 .
As expected, E-cvD particles were not infective, highlighting the need of E trimerization within endosomes at the low pH conditions that promote fusion of the viral envelope with endosomal membranes 60 . In addition, our data support the idea that viral neutralization can be induced by interfering with viral surface dynamics, a mechanism that has been suggested to explain the strong neutralization showed by antibodies targeting the hinge region or the complex quaternary epitopes 41,42 , recently demonstrated for ZIKV neutralisation by mAb EDE1-C10 61 .
The ability of sE proteins from different DENV serotypes and ZIKV to form hetero-dimers was surprising. We confirmed this initial observation in the context of pseudoviruses assembled into phenotypically mixed mosaic particles carrying ZIKV and DENV2 E proteins. Although we lack the means to determine the precise structural configuration and composition of these particles, their mosaic nature was shown by promoting ADE on K562 cells with an anti-DENV2 specific polyclonal serum and with the DENV-specific mAb 4E5A.
Phenotypically mixed viruses have been previously produced for Influenza A/B viruses 62 . Here we show for the first time that mosaic particles with E proteins from two different flaviviruses can assemble into infective particles. Although single cell flaviviral co-infection has not been reported, our data suggest that in such scenario particles with enhanced infectivity in the presence of antibodies specific for either of the two different E proteins could exist.
Even though antibodies are important to neutralize the virus, the specificity required for efficient protection is still debated. Latest studies indicate that antibodies against DIII and complex quaternary epitopes of E, but not those against DI or DII, are able to strongly neutralize the infection with low risk of ADE, making these antigens preferable for vaccine developments 20,30,34 . The ELISA and cytofluorimetry-based assays presented here provide efficient tools not only to study the binding properties of antibodies, but also to screen for those able to recognize dimeric and more complex epitopes without the safety concerns of using whole viral preparations. Moreover, our covalent stabilized E dimers represent a first step to exploit the immunogenic potential of quaternary epitopes and formulate efficient cross-neutralizing vaccines. Constructs tagged with BAP (biotin acceptor peptide, GLNDIFEAQKIEWHE) 46, 47 , or targeted for membrane display (with the transmembrane and cytoplasmic domains of the human MHC-Iα chain), were obtained by cloning into the corresponding vectors as previously reported 36 .

Materials and Methods
For pseudovirus production, codon optimized sequences coding for C-PrM-E genes of ZIKV Mr766, and the chimera containing the C gene from ZIKV Mr766 and the PrM-E form DENV2 New Guinea C strain (C Z -PrME D2 ), were obtained as synthetic fragments and cloned into pVAX vectors. Plasmids expressing only C from ZIKV or PrME from DENV2 were obtained by deletion the PrME and C coding regions, respectively.
Covalent E dimers were obtained by site-directed mutagenesis (Ala to Cys) of Ala259 for DENV1, 2 and 4, Ala257 for DENV3 and Ala264 for ZIKV (QuikChange XL Site-Directed Mutagenesis Kit, Agilent Technologies, La Jolla, CA, USA) following manufacturer instructions.
Monoclonal antibodies and mouse sera. Purified human mAbs 2D22 41 , 1F4 43 and 5J7 42 as well as mouse mAb 4G2 37 , were kindly provided by Prof. Aravinda de Silva (University of North Carolina, Chapel Hill, NC, USA). Mouse mAb 4E5A 40 was kindly provided by Prof. Jonathan R. Lai (Albert Einstein College of Medicine, NY, USA). mAbs EDE1-C8 (clone 752-2 C8), EDE1-C10 (clone 753(3) C10) and EDE2-B7 (clone 747 B7) were constructed as human IgG antibodies from previously deposited sequences 27 and used as culture supernatants from transfected cells. mAb anti-SV5 48 was obtained as previously described 64 . Two groups of six 5-weeks old, female Balb/c mice (Harlan, Milan, Italy) were immunized three times at two weeks intervals by Gene Gun technology (Bio-Rad, Hercules, CA, USA) using 1 μm gold particles coated with 1 μg of plasmid DNA encoding either the domain DIII of DENV2 E protein or domains DI/DII of DENV4 E protein as previously described 56 . Blood samples were collected at day 90 by sub-mandibular puncture and pooled to obtained anti-2DIII and anti-4DI/ DII sera, respectively. Pooled sera from non-vaccinated animals were used as negative control (Pre-immune sera). All animal procedures described in this study were approved by the ICGEB Animal Welfare Board and the Italian Ministry of Health (Ministero della Salute) (protocol DGSAF0024706) and conducted in accordance to international and institutional guidelines for animal experimentation, and in compliance to laws and policies established in the legislation D. L.vo 26/2014 of the Italian Government.
Expression of secretory E proteins. Standard calcium phosphate transfections were performed on HEK-293T cells as previously described 65 . For Vero and HeLa cells, transfections were done with Lipofectamine 3000 (Life Technologies) following manufacturer's protocols. After overnight incubation at 37 °C, culture medium was replaced with a serum-free medium and cells incubated for another 24 h at 37 °C or 28 °C; when required media was supplemented with 100 μM biotin (Sigma-Aldrich, St. Louis, MO, USA). Free biotin was removed from cultures supernatants by dialysis against PBS. For hetero-dimerization assays SV5-tagged and BAP-tagged constructs were co-transfected in a 2:1 ratio. In all cases, cellular extracts were prepared in 100 μl of TNN lysis buffer (100 mM Tris-HCl, pH 8, 250 mM NaCl, 0.5% NP-40) at 4 °C, supplemented with Protease Inhibitor Cocktail (Sigma-Aldrich).
Production of Pseudoviral particles. Pseudoviral particles were produced by co-transfection of a DNA-launched WNV replicon expressing EGFP 45 (kindly provided by Dr. Theodore Pierson, National Institute of Allergies and Infectious Diseases, MD, USA) with plasmids expressing the structural genes of ZIKV or DENV2 in trans, as previously described 34 . Briefly, HEK-293T cells were transfected with linear polyethylenimine (MW 25,000, PEI, Polysciences, Warrington, PA, USA). After 16 h, culture medium was replaced with a DMEM-7% FCS and incubation was extended for another 24 h at 37 °C or 28 °C. Pseudoviruses were harvested, clarified from cell debris by centrifugation (10 min at 10,000 rpm) and stored at −20 °C until use. For western blot analysis, particles were concentrated by ultracentrifugation at 40,000 g for 24 h. Quantification of pseudoviruses by Real-Time PCR. WNV Replicon-derived RNA was isolated from 100 μl of pseudoviral preparations using the RNAzol-BEE solution (Tel-Test, Friendswood, TX, USA) and then treated with DNaseI (Promega, Madison, WI, USA) following manufacturer's instructions. For cDNA synthesis, reverse transcription was performed using random hexamers (Sigma) and M-MLV Reverse transcriptase (Life Technologies) according to the manufacturer's protocol. The product of retrotranscription was used as a template for quantitative real-time PCR using previously reported 3′UTR specific primers 66 and EvaGreen Technology (Bio-Rad, Hercules, CA, USA) on a C1000 thermal cycler using CFX96 ™ Real-Time Detection System (Bio-Rad).
Quantification of the relative number of viral RNA genomic copies was calculated by interpolating in a standard curve made with a quantified WNV-rep plasmid (r 2 = 0.9972). This method was used in pseudoviral infection assays to determine the multiplicity of infection (MOI).