ENVELOPED VIRUS-LIKE PARTICLES (eVLPs) EXPRESSING MODIFIED FORMS OF ZIKA VIRUS PROTEINS E AND NS1 PROTECT MICE FROM ZIKA VIRUS INFECTION

While Zika virus (ZIKV) infection induces mild disease in the majority of cases, it has been identified as responsible for microcephaly and severe neurological disorders in recent 2015-2016 outbreaks in South America and the Caribbean. Since then, several prophylactic vaccine strategies have been studied. Here, we describe the development of a ZIKV candidate vaccine consisting of bivalent enveloped virus-like particles (eVLPs) expressing a modified form of E and truncated NS1 (EG/NS1) proteins. In EG/NS1, the E transmembrane/cytoplasmic tail has been replaced with those domains from the VSV G protein and a β-domain of NS1 was fused in-frame to Gag from Moloney murine leukemia virus (MLV). Immunization of BALB/C mice demonstrated that bivalent EG/NS1 and monovalent EG eVLPs induced comparable levels of antibody (Ab) titers but that EG/NS1 induced much higher neutralizing activity, comparable to naturally acquired anti-ZIKV immunity. In contrast, monovalent NS1 eVLPs did not induce a significant anti-NS1 Ab response but promoted strong T cell immunity that was also elicited with EG/NS1 eVLPs. ZIKV challenge studies in C57BL/6-IFNαR−/− mice demonstrated that EG/NS1 eVLPs conferred 100% protection against clinical disease after ZIKV challenge compared to 80% protection after EG eVLP vaccination, with protection against challenge correlating with neutralizing antibody titers and overt signs of infection. Author Summary Zika virus has caused rapidly spreading epidemics with potentially severe neurological symptoms including microcephaly in new born babies. Rapid progress has been made with several candidate vaccines under clinical evaluation but no vaccine or treatment is yet available. In this context, we have produced and tested recombinant virus-like particles that incorporate one or two Zika virus proteins, E and NS1 that have been modified for optimal efficacy. Our immunogenicity studies in mice showed a synergistic effect of both proteins in the bivalent vaccine. NS1 induced a strong T cell response enhancing the neutralizing antibody production induced by the E protein. In challenge experiments, the bivalent vaccine protected 100% of mice from clinical signs of Zika virus infection. These products could be further used to explore Zika virus correlates of protection and evaluated as vaccine candidates.


51
Zika virus was first isolated in 1947 in a Rhesus sentinel monkey in Uganda, and was 52 not initially regarded as a significant human pathogen. However, large outbreaks in the 53 Pacific area including Yap Island in 2007; French Polynesia in 2013 to 2014) and the intense 54 ZIKV NS1-specific T cell immunity, as shown by induction of IFN- and IL-5 T cell 165 secretion (Figs 6A-B), was detected in all groups of mice that received NS1 eVLPs. Similar 166 responses were observed when comparing monovalent NS1 eVLPs and bivalent EG/NS1 167 eVLPs. In contrast, co-injection of EG eVLPs with NS1 eVLPs did not enhance the nAb 168 response and induced lower levels of T cell activation levels than those observed with 169 EG/NS1. These results emphasized the value of having the EG and NS1 proteins co-170 expressed within the same particles. 171 172 EG and EG/NS1 eVLPs protect mice from ZIKV challenge 173 The protective efficacy of the ZIKV eVLPs was examined in IFNR deficient mice fully 174 bred to a C57BL/6 background (C57BL/6-IFNR -/mice). These mice, like BALB/C mice are 175 able to mount protective immunity to ZIKV, but unlike the latter during primary infection, they 176 also display significant weight loss associated with other overt sign of infection such as hind 177 leg paralysis, the absence of which provides a facile read out of vaccine efficacy (30). Two 178 separated challenge experiments were conducted successively using identical protocols with 179 different groups. 180 A first challenge experiment was conducted to establish whether EG and EG/NS1 181 eVLPs could induce protective Ab immune response against ZIKV infection in the C57BL/6-182 IFNR -/mouse model. Four groups of 10 mice were injected with either saline, eVLPs 183 control expressing no ZIKV protein (Gag eVLP), EG eVLPs or EG/NS1 eVLPs. After a single 184 injection of EG or EG/NS1 eVLPs, the levels of anti-E IgG rapidly increased in the serum of 185 immunized mice peaking at 14 days after the second injection while the third injection 186 maintained Ab titers at a plateau ( Fig 7A). 187 Evaluation of individual sera demonstrated that antibody responses were induced in 8 188 of 10 mice immunized with EG eVLPs and in 10 of 10 mice immunized with EG/NS1 eVLPs 189 in Exp.1 (Fig 7B). Neutralization activity against ZIKV was evaluated by PRNT90% 27 days 190 after the third injection. The sera from mice immunized with EG and EG/NS1 eVLPs possessed significant neutralizing activity which correlated with induction of antibody 192 responses in the mice (Figs 7B-C). Of interest, the sera from both mice with undetected 193 levels of anti-E had no neutralizing activity. A similar pattern of antibody response with higher 194 levels of IgG was observed in a second experiment were a group of mice were injected with 195 NS1 eVLPs instead of saline. In this experiment, all mice EG and EG/NS1 groups develop an 196 Ab response. We further analyzed the immunoglobin isotype usage in the sera from mice 197 from Exp.1. The IgG1 to IgG2b ratio indicated a predominant Th2 response in the EG group 198 as expected from Alum adjuvant. However a significant difference was observed with 199 bivalent EG/NS1 eVLPs, which promoted a switch towards IgG2b, that is usually associated 200 with cell-mediated immunity (Fig 8). 201 After ZIKV challenge, mice were monitored daily for body weight change and 202 behavior. C57BL/6-IFNR -/mice in groups that received saline, Gag or NS1 eVLPs lost 203 body weight and developed hind leg paresis between days 8-10 after ZIKV injection (Fig 9). The second challenge experiment was aimed at determining if NS1 eVLPs used 213 alone could induce protection despite the absence of an efficient Ab response. Careful 214 observation of animal behavior showed that despite no significant protection from paralysis 215 compared to the control group (Fig 9C), some mice injected with NS1 eVLPs had milder 216 symptoms after day 10 ( Fig 9D). 217

219
Several ZIKV vaccine candidates have advanced through efficacy studies in mice and 220 non-human primates (NHP) and are either in, or poised for, Phase I clinical trials. These 221 include DNA-based vaccines expressing the ZIKV envelope protein (EP), whole killed ZIKV, 222 and recombinant viruses including recombinant DENV and YFV expressing ZIKV EP (31-223 34). Live-attenuated and replication-competent vaccines are expected to provide long-term 224 immunity and to be highly effective (35), but these vaccines are usually contraindicated in 225 immunocompromised individuals and in pregnant women because of a potential risk of 226 reversion of pathogenicity. The development of Zika vaccines may also be complicated by 227 potential issues of ADE mostly due to a high degree of homology among the members of this 228 family, especially between ZIKV and DENV (36). 229 We have developed an eVLP platform that led to the production of a CMV vaccine 230 candidate currently under Phase I clinical trial evaluation (37). In the present study we have 231 used this eVLP platform to develop a prophylactic ZIKV vaccine candidate that includes a 232 modified form of the ZIKV E protein, consisting of swapping its transmembrane/cytoplasmic 233 domain for that of the VSV-G protein as well as the β-ladder domain of the NS1 protein. The 234 present study demonstrated that co-expression of the β-ladder domain of NS1 provides a 235 synergistic effect on the neutralizing antibody response to the ZIKV envelop protein, and is 236 associated with significant T cell activation 237 In immunocompetent BALB/C mice, EG eVLPs injected in the presence of Alum 238 adjuvant induced a potent IgG response but neutralizing activity was modest. We 239 hypothesized that NS1 could induce NS1-specific T-cell activation to provide T-cell help and 240 increased neutralizing activity. When NS1 and EG were co-expressed in EG/NS1 eVLPs, a 241 synergistic effect was observed with the neutralizing IgG response to reach an average of 242 100 fold increase over the response to EG alone. Mice that received eVLPs expressing NS1 243 showed a marked increase in NS1 specific T-cell activation that correlated with the increase 244 in anti-ZIKV E IgG (r=0.9). 245 Despite the presence of several B-cell epitopes in the sequence used in our NS1 246 construct (38,39), NS1 and EG/NS1 eVLPs induced no or low Ab responses to this protein. 247 In contrast, previous studies have demonstrated that vaccine candidates expressing full 248 length NS1 promoted efficient specific anti-NS1 neutralizing Ab responses associated with 249 protection against ZIKV infection (12,(40)(41)(42). The discrepancy between our data and others 250 is likely due to alteration of NS1 presentation in two ways: i) the NS1 domain was fused with 251 Gag which can alter the tri-dimensional structure of NS1, ii) the NS1 domain is kept inside 252 the particles, reducing its accessibility to B cells. These data reveal another potential utility of 253 eVLPs for the analysis of T-cell versus B-cell responses against proteins, depending on their 254 location on or within the particles. Of note, the β-ladder of the NS1 is flexible and can interact 255 with many proteins. It is then possible that the use of a fusion protein Gag-NS1 can induce 256 alteration of the structure of the particle, and subsequent alteration of the coexpressed EG 257 protein, as suggested by the spike-like appearance of particles analyzed by cryo-EM (29). 258

As observed in BALB/C mice, EG and EG/NS1 eVLPs induced similar levels of anti-E 259
IgG in C57BL/6-IFNR -/mouse sera. However, in C57BL/6-IFNR -/mice, expression of EG 260 alone was sufficient to induce a strong neutralization antibody response equivalent to the one 261 observed in response to EG/NS1 eVLPs in 18 out of 20 mice across 2 independent studies. 262 Future studies may clarify if this difference is due to the genetic differences between 263 C57/BL6 and BALB/6, or to the IFNR pathway deficiency in C57/BL6-IFNR -/-. All mice that 264 mounted a nAb response, either in response to EG eVLPs or EG/NS1 eVLPs were protected 265 against experimental ZIKV infection. 266 These data confirm that neutralizing Ab responses are critical in protection against 267 ZIKV clinical disease and that the bivalent EG/NS1 eVLPs provide a more robust induction of 268 nAb response than monovalent EG eVLPs. In C57BL/6-IFNR -/mice, NS1 promoted a 269 switch towards a Th1 antibody response, emphasizing a strong effect on T-cell help even in 270 the presence of the Th2-adjuvant Alum.

It has been shown that an Adenovector ZIKV vaccine that induced a T-cell response 272
against the ZIKV envelop without any Ab response could protect C57BL/6 mice against ZIKV 273 infection (43). In our model, the C57BL/6-IFNR -/mice that received NS1 eVLP alone were 274 not protected from ZIKV infection in absence of an efficient nAb response. Further studies 275 would be necessary to determine the role of T cells in these IFNR deficient mice. 276 Flaviviruses possess a high degree of homology that leads to the abundant 277 production of cross-reactive Abs. While the mechanism for ADE has not been fully 278 elucidated, it is generally recognized that the presence of cross-reactive Abs with low 279 neutralizing potency are involved in ADE (15). The neutralization potency depends on the 280 affinity of the Abs and on the accessibility of the epitopes on the virions (44,45). In the 281 present study, mice immunized developed a high neutralizing response that correlated with 282 protection. Most potent nAb with low/no ADE have been described to be directed against 283

EDIII domain and quaternary epitopes from the DI/DII domain (46,47). Further structural 284
studies would determine whether the EG conformation at the surface of the eVLPs, would 285 present such epitopes in an optimal conformation for Ab production. 286 VLPs expressing vaccine antigens are typically more immunogenic than monomeric 287 recombinant forms of vaccine antigens because of repeating, array-like presentation of 288 antigens, which is a preferred means of activating a B cell response. Moreover, the 289 particulate nature of the antigen, relative to recombinant proteins is a much better means of 290 activating dendritic cell responses and further enhances immunity. While being a highly 291 immunogenic means of delivering vaccine antigens, VLP-based vaccines avoid the potential 292 safety concerns associated with whole killed or recombinant viral-based vaccines, as there is 293 no residual host DNA/RNA and no possibility for infection. 294 When glycoproteins are expressed/present on the surface of eVLPs, one terminus of 295 the protein is anchored within the lipid bilayer, which imposes structural constraints not 296 readily observed with monomeric recombinant antigens. Moreover, by altering the 297 transmembrane domain/cytoplasmic tail of the glycoprotein, and/or the core particle structure 298 with which the cytoplasmic tail may interact, we have found that a novel presentation of glycoproteins can be obtained that is associated with enrichment for antibodies with 300 neutralizing rather than just binding activity. This was demonstrated previously with eVLP 301 expression of a modified form of the CMV gB (25), similar to what we have observed with the 302 EG/NS1 ZIKV vaccine candidate presented in this study. The high degree of neutralization 303 obtained with EG/NS1 eVLPs may minimize the risk of ADE and is an attractive candidate for 304 further development as a prophylactic vaccine against ZIKV.

Construction of plasmids 316
We used the ZIKV E sequence from Suriname isolate 2015 (23) (Genbank 317 KU312312) to construct a plasmid expressing the full length unmodified ZIKV prME and a 318 plasmid expressing the ectodomain of E fused with the transmembrane and cytoplasmic 319 domain of the VSV G protein (EG) (24). The ZIKV E sequences were preceded by a portion 320 of the sequence corresponding to the last 32 aa from the Zika virus capsid, as signal peptide. 321 For the production of VLP, we used a minimal cDNA sequence encoding a Gag polyprotein 322 of murine leukemia virus (MLV) (Gag without its C-terminal Pol sequence) or Gag-NS1 or 323 Gag-TR.NS1, were Gag was fused with full length NS1 or truncated NS1 consisting of the β-324 ladder domain of NS1 (aa 180 to 353). The design of the plasmids is summarized in Fig1. All final sequences were synthetized and optimized for mammalian cell expression at Genscript 326 (NJ), prior to subcloning into our in-house phCMV-expression plasmid (24). 327 All DNA plasmids were amplified in high-efficiency competent Escherichia coli cells 328 and purified with an endotoxin-free preparation kit using standard methodologies. 329 330 Production and purification of eVLPs 331 eVLPs were produced using transient polyethylenimine transfection in HEK 293SF-332 3F6 GMP compliant cells as described previously (25). Harvests containing eVLPs were 333 purified either by ultracentrifugation using sucrose cushion as described previously (25)  Alum Adjuvant (Adju-Phos®, Brenntag) or saline control at monthly intervals or as indicated for each experiment. Blood samples were collected every 2 weeks starting 2 weeks before 382 the first injection. Sera were either pooled or analyzed individually. 383 Challenge with 10,000 PFU of ZIKV Philippine 2013 was done 4 weeks after final 384 injection in C57BL/6-IFNR -/mice by the IP route. After infection, the mice were monitored 385 daily for body weight and clinical behavior as indicated in figure legends and blood was 386 collected at the end of the study. 387 388

Monitoring of immune response 389
Direct ELISA described earlier (25) was used to measure Ab binding titers to 390 recombinant E and NS1 protein from ZIKV (both from Meridian LifeScience). Plaque 391 Reduction Neutralization Test (PRNT) was used to measure ZIKV-specific neutralizing 392 activity in mouse sera. Vero cells were seeded at 3x10 5 cells/mL in 24-well plates 24h prior 393 PRNT. On the day of assay, virus and serially diluted serum samples were mixed and 394 incubated for 1h at 37±1°C. Supernatant from cell-seeded 24-well plates was decanted then 395 100 ml of virus/serum mixture was transferred from the dilution plate to the cells. After 1h 396 adsorption, agarose-containing overlay media was added and plates were incubated at 37°C, 397 5%CO2 for 3 days. The cells were fixed and stained using crystal violet solution and plaques 398 were counted visually. The neutralizing antibody titer was expressed as the highest test 399 serum dilution for which the virus infectivity is reduced by 50% or by 90%, as indicated in 400 legends. NS1-specific IFN and IL-5 production by ZIKV-specific splenic T cells were 401 enumerated by Enzyme Linked ImmunoSpot (ELISPOT) assay following manufacturer's 402 instructions (Mouse IFN-/IL-5 ELISpot Basic kits; Mabtech). Mouse splenocytes were left 403 unstimulated or stimulated with an NS1 peptide pool (PepMix ZIKV NS1 Ultra; JPT Peptide 404 Technologies, Germany) designed to cover the high sequence diversity of ZIKV. Spots were 405 counted at ZellNet Consulting (NJ, USA) using KS ELISPOT Reader System (Carl Zeiss,406 Thornwood, NY, USA) with KZ ELISPOT Software Version 4.9.16. The background was