Molecular mechanisms underlying attenuation of live attenuated Japanese encephalitis virus vaccine SA14-14-2

The live attenuated Japanese encephalitis virus vaccine SA14-14-2 demonstrated ≥ 95 % efficacy and is today the vaccine of choice against JEV globally. Relative to its parent strain SA14, SA14-14-2 carries 46 nucleotide and 24 amino acid alterations, with 8 of the latter located within the envelope glycoprotein. The vaccine strain also fails to synthesize the nonstructural protein NS1’ owing to a silent mutation that abrogates a-1-frameshifting event close to the 5’ end of the NS2A coding sequence. Previous studies employing reverse genetics and mouse models implicated both absence of NS1’ and mutated E, in attenuation of SA14-14-2. We demonstrate progressive reduction in ER stress sensor PERK levels and increased expression of CEBP-homologous protein (CHOP), accompanied by dephosphorylation of eIF2α, inhibition of autophagy maturation and necroptosis following infection of cultured cells with wild-type JEV strain P20778. Autonomous expression of NS1’ caused constitutive up-regulation of CHOP and loss of PERK. Conversely, infection with SA14-14-2 led to significantly increased IRE-1α activation, ER chaperone levels and autophagy. We report labile conformational epitopes accompanied by drastically reduced folding kinetics of intracellular SA14-14-2 envelope protein engendered by sluggish oxidation of cysteine sulfhydryl groups to form disulfide bonds within the endoplasmic reticulum along with altered envelope epitopes in extracellular SA14-14-2 viral particles. We also demonstrate near total conversion of prM to pr and M in SA14-14-2 virus particles. These alterations were accompanied by enhanced activation of mouse and human antigen presenting cells by SA14-14-2 along with superior CD8+ recall T cell responses to viral structural proteins in volunteers vaccinated with SA14-14-2. Author Summary The random process of cell culture passage adopted in generation of most live attenuated virus vaccines leads to fixation of multiple nucleotide changes in their genomes and renders it difficult if not impossible to pinpoint those mutations primarily responsible for their attenuated phenotype. Identifying the precise attenuating mutations and their modi operandi should aid in developing rationally attenuated vaccines for other viruses. We discovered that wild type (WT) JEV uses the nonstructural protein NS1’ to take over the host protein synthesis machinery to produce viral proteins. Loss of NS1’ in SA14-14-2 deprives the vaccine strain of this ability. Viruses uniformly target host death pathways to avoid generating potent antiviral immune responses. WT JEV prevents autophagy maturation. Conversely the SA14-14-2 vaccine activates autophagy due to unresolved ER stress caused by inability of its envelope glycoprotein to fold promptly post synthesis. Combined with enhanced proteolytic cleavage of the viral prM protein in SA14-14-2, this resulted in altered envelope epitopes on extracellular SA14-14-2 virus particles. These changes culminated in enhanced activation of innate and adaptive immune responses by SA14-14-2.


INTRODUCTION
The genus Flavivirus, in the family Flaviviridae, comprising numerous vector-borne human 55 pathogens has spawned two highly efficacious live attenuated vaccines in Yellow Fever virus 56 (YFV)-17D and Japanese encephalitis virus (JEV) -SA14-14-2, both of which have contributed 57 to significant reduction of disease incidence from their respective viral pathogens. Flavivirus 58 genomes are single strand RNA of positive polarity and encode a single large polyprotein 59 which is processed by host and virally encoded proteases to give rise to 3 structural (capsid, C; 60 envelope, E; and premembrane, prM) and 7 non-structural (NS; 1, 2A, 2B, 3, 4A, 4B and 5) 61 proteins. JEV is represented by a single serotype and 5 genotypes [1] and was first isolated in 62 Japan in 1934, giving us the prototype Nakayama strain [2]. Based on meta-analysis of 63 published literature and national incidence estimates of 124 countries the annual global 64 incidence of JEV encephalitis was computed around 67,900 [3]. 65 Of the 24 amino acid changes identified in SA14-14-2 relative to its parent Chinese strain 66 SA14, the largest number, namely 8 were located within the envelope glycoprotein E [4,5]. 67 Similarly, 8 of the 22 amino acid changes in YF17D relative to its parent Asibi strain were 68 found in the E protein [6]. Studies in the mouse model implicated multiple mutations within 69 the E gene of SA14-14-2 in the attenuated phenotype of this vaccine strain [7][8][9][10]. Additionally, 70 a silent G to A mutation near the beginning of the NS2A gene which led to abrogation of -1 71 ribosome frameshifted NS1' synthesis was reported to determine loss of neurovirulence of 72 SA14-14-2 [11]. Both viral E and NS1' are secreted glycoproteins synthesized on endoplasmic 73 reticulum (ER)-bound ribosomes, that are dependent on the ER lumen for glycosylation and 74 folding. The molecular mechanisms underlying the attenuation caused by these mutations have 75 not been elucidated. 76 As obligate intracellular parasites, viruses rely on usurping the host translational  The most conserved ER stress sensor IRE-1α, an ER transmembrane kinase and 100 endoribonuclease which is activated when unfolded proteins accumulate within the ER, serves 101 to match protein folding burden within the ER with capacity. IRE-1α generates the 102 transcription factor XBP-1 [25] through nonconventional cytoplasmic splicing which in turn 6 103 helps to maintain ER homeostasis and prevent activation of cell death pathways caused by 104 sustained ER stress. The IRE-1α-XBP-1 axis was reported to be vital and essential for 105 development and survival of dendritic cells [26], certain subsets of which were shown to 106 constitutively activate the IRE-1α-XBP-1 pathway [26,27]. More importantly, both 107 transcription and splicing of XBP-1 mRNA were reported to be increased in response to viral 108 or bacterial infection within antigen specific CD8 + T cells, whose differentiation into killer cell

127
Differential modulation of PERK-eIF2α-CHOP pathway by wild type and vaccine strains 128 of JEV. 129 As a first step to query the underlying basis for attenuation of SA14-14-2 we compared 130 activation of the ER stress sensor PERK in JEV-infected cells. We observed progressive  (Fig 1C and D To query a role for NS1' (absent in SA14-14-2) in modulating the PERK pathway, we created 150 cells stably expressing NS1 or NS1' of P20778 by lentivirus transduction. PS cells stably 151 expressing NS1' revealed constitutive high expression of CHOP (Fig 2, lanes 9 to 12). Cells 152 expressing P20778 NS1 protein were distinctly devoid of CHOP, even on long exposure (Fig   153   2, lanes 5 to 8). Surprisingly, NS1'-induced CHOP expression did not lead to 154 dephosphorylation of eIF2α, presumably due to vital compensatory pathways in cells 155 conditioned to constitutively express CHOP over multiple passages. We also observed 156 progressive loss of PERK levels in NS1'-expressing cells over a span of 36 hours (Fig 2, lanes 157 9 to 12). These results revealed a role for the C-terminal frameshifted 52 amino acid segment 158 of NS1' in modulating levels of CHOP and PERK following viral infection. Loss of NS1' in 159 SA14-14-2-infected cells would understandably cripple the vaccine strain due to its inability 160 to appropriate the translation machinery of the host.

161
WT JEV blocks autophagy maturation. 162 The sustained high phosphorylation levels of eIF2α in SA14-14-2-infected cells led us to  The observed persistent eIF2α phosphorylation combined with IRE-1α activation and increased 198 autophagy flux pointed to unresolved protein folding stress within SA14-14-2-infected cells. 199 We therefore investigated levels of ER chaperones in virus infected cells. We observed 200 dramatic upregulation of BiP and calnexin/calreticulin only in SA14-14-2-infected PS and 201 N2A cells (Fig 4A and B). Keeping in mind the eight mutated residues of SA14-14-2 envelope, 202 we proceeded to probe the stability and folding of envelope protein in cells infected with JEV. 203 We utilized a rabbit polyclonal antiserum specific to JEV E protein to determine the SA14-14-2 is also spared from degradation observed for E, suggesting that the SA14-14-2 E 254 was specifically targeted presumably by ER-associated degradation (ERAD) machinery of 255 infected cells owing to its tardy folding kinetics. We observed no such differential instability 256 of P20778 E relative to NS3 and NS5 (S5 FigA).

257
When we investigated the susceptibility of nascent E protein to reduction by DTT, we In order to ask if the above differences in the cell biology of infection with P20778 and SA14-288 14-2 would be reflected in the immune response to the virus, we compared recall T cell 289 responses in individuals exposed to circulating wild type JEV in endemic regions with those 290 vaccinated with SA14-14-2. Capsid and E protein-specific T cells were both characterized by 291 the dominance of CD4 + over CD8 + subsets in those naturally infected with circulating WT 14 297 Importantly, significantly greater percentages of capsid-specific CD8 + T cells secreting IFN-γ, 298 TNF-α or MIP-1β including polyfunctional ones were observed in vaccinated individuals 299 compared to infected HV and recovered JEV patients (Fig 7A, top panel). E-specific CD8 + T 300 cells secreting IFN-γ, TNF-α and MIP-1β including polyfunctional T cells showed 301 enhancement in SA14-14-2 vaccinated individuals relative to recovered JEV patients (Fig 7A,   302 middle panel), in keeping with the earlier reported absence of JEV-specific CD8 + T cells in 303 recovered JEV patients [41,42]. In contrast, we did not observe such enhancement of NS3-304 specific T cells in vaccinees compared to naturally infected individuals (Fig 7A, lower panel).

305
NS3 is known to be the strongest stimulator of human CD8 + T cells in JEV-endemic cohorts 306 [42,43]. IL-2 responses to JEV proteins were relatively weak (data not shown). These results 307 suggested superior presentation of viral structural proteins to CD8 + T cells following SA14-308 14-2 infection.

309
When we queried the ability of P20778 and SA14-14-2 to stimulate activation of 310 antigen presenting cells, we observed enhanced death of primary mouse dendritic cells and 311 human monocyte cell line THP-1 detected by Annexin-V and propidium iodide staining 312 following 24 h infection by SA14-14-2 compared to P20778 (Fig 7B). In parallel, we also 313 detected impressive increases in levels of inflammatory cytokines IL-12p40, IL-6 and TNF-α 314 secreted from BMDC infected with SA14-14-2 compared to P20778 (Fig 7C). While we have 315 not investigated the mode of death in infected cells, the above data suggest that presentation of 316 viral antigens following infection with SA14-14-2 would be far more efficient relative to 14-2 appears to be the immediate cause of autophagy enhancement mediated by persistent 399 eIF2α phosphorylation. The abundant levels of viral NS proteins 1, 3 and 5 in SA14-14-2 400 infected cells (Fig 3B, S1 FigA) would suggest that despite IRE-1α activation, viral mRNA 401 degradation by regulated IRE-1α dependent decay (RIDD) was not triggered. Among these NS 402 proteins, it is particularly notable that NS1 of SA14-14-2 escapes degradation despite being 403 ER localized (Fig 3B), further confirming that the targeted susceptibility of SA14-14-2 E to 404 degradation emanates from its delayed folding kinetics within the ER. 405 We suspect that the unstable conformation of SA14-14-2 E protein permits superior to clarify whether the nearly complete prM cleavage renders these particles smooth and 422 homogeneous in appearance.

423
The observed significantly higher levels of inflammatory cytokines from primary 424 BMDC infected with SA14-14-2 as well as significantly greater death of infected BMDC and 425 THP-1 human monocytes compared to P20778, most likely resulted from the enhanced 426 autophagy triggered by SA14-14-2. This ability of SA14-14-2 to better activate BMDC was 427 also evident when comparing two earlier reports [59,60]. In keeping with this observation, we  (Table S1) to obtain the full length NS1' gene. OSV 387 and OSV388 were 543 designed to disrupt the slippery heptanucleotide and pseudoknot structure at the beginning of 24 544 the NS2a gene of JEV that promote ribosomal frameshifting, without altering the amino acid 545 sequence of the NS1' C-terminus. The full length NS1' gene was amplified using forward 546 primer OSV 389 and reverse primer OSV393 containing the haemagglutinin (HA) tag sequence 547 followed by a termination codon and a NotI site. Cohesive ends were generated by restriction Peptides were coated at a concentration of 1 µg/ml, 100 µl/ well. 100 µl containing 10 4 pfu of 619 either P20778 or SA14-14-2 virus in 100 mM carbonate buffer pH-9.6 was coated on high  Kruskal-Wallis test with Dunn's test or parametric ANOVA with Bonferroni correction for 683 multiple comparisons, respectively. Band intensities quantitated using ImageJ were compared 684 using unpaired Student's t test with alpha set at 0.05.

685
Acknowledgements: 686 We thank Sai Pallavi Pradeep and Madhusudan Thirumallesh for help with compiling figures.