Japanese encephalitis virus live attenuated vaccine strains display altered immunogenicity, virulence and genetic diversity

Japanese encephalitis virus (JEV) is the etiological agent of Japanese encephalitis (JE). The most commonly used vaccine used to prevent JE is the live-attenuated strain SA14-14-2, which was generated by serial passage of the wild-type (WT) JEV strain SA14. Two other vaccine candidates, SA14-5-3 and SA14-2-8 were derived from SA14. Both were shown to be attenuated but lacked sufficient immunogenicity to be considered effective vaccines. To better contrast the SA14-14-2 vaccine with its less-immunogenic counterparts, genetic diversity, ribavirin sensitivity, mouse virulence and mouse immunogenicity of the three vaccines were investigated. Next generation sequencing demonstrated that SA14-14-2 was significantly more diverse than both SA14-5-3 and SA14-2-8, and was slightly less diverse than WT SA14. Notably, WT SA14 had unpredictable levels of diversity across its genome whereas SA14-14-2 is highly diverse, but genetic diversity is not random, rather the virus only tolerates variability at certain residues. Using Ribavirin sensitivity in vitro, it was found that SA14-14-2 has a lower fidelity replication complex compared to SA14-5-3 and SA14-2-8. Mouse virulence studies showed that SA14-2-8 was the most virulent of the three vaccine strains while SA14-14-2 had the most favorable combination of safety (virulence) and immunogenicity for all vaccines tested. SA14-14-2 contains genetic diversity and sensitivity to the antiviral Ribavirin similar to WT parent SA14, and this genetic diversity likely explains the (1) differences in genomic sequences reported for SA14-14-2 and (2) the encoding of major attenuation determinants by the viral E protein.


INTRODUCTION
Japanese encephalitis virus (JEV) is the prototype member of the JE serogroup within the genus Flavivirus, which contains other mosquito-borne encephalitic viruses such as West Nile and St. Louis encephalitis. JEV is the leading cause of epidemic, viral encephalitis in mainland Asia and the Western Pacific causing an estimated 57,000 to 175,000 cases annually with a 20-30% case fatality rate. Thirty to fifty percent of those who survive the acute infection report serious neurological sequelae 1,2 . There are no licensed antivirals to treat the disease making vaccination the primary control strategy. A number of JEV vaccines have been licensed including live attenuated (LAV), chimeric LAVs and inactivated vaccines. The most commonly used in Asia is the LAV strain SA14-14-2, which is highly efficacious showing 90% seroconversion after a single immunization in multiple clinical trials [3][4][5] .
Strain SA14-14-2 was derived from the wild-type (WT) strain SA14 by plaque purification after serial passage in primary hamster kidney (PHK) cells and suckling mice 6 . The genomic sequences of SA14-14-2 and SA14 have been published a number of times but the published sequences differ between groups, which makes determination of attenuating mutations difficult [7][8][9][10][11][12] . While one study reports that there are no genomic changes after passage of SA14-14-2 in cell culture, mice or mosquitoes 13 others report consensus changes after very few passages in similar substrates. These changes include three amino acid substitutions after passage in C6/36 cells, four amino acid substitutions after 22 passages in PHK cells, one amino acid substitution after passage in baby hamster kidney (BHK) cells, one amino acid substitution after passage in Vero cells, and four amino acid substitutions after passage in mouse brain 10,14,15 . There have been limited studies to resolve the differences in genomic sequences 12,16 or the conditions that favor changes to the vaccine genotype 17 .
JEV strains fall into a single serogroup that contains five genotypes. All current JEV vaccines are derived from genotype III. Licensed JEV vaccines appear to provide protective immunity against infection by the different genotypes, although the situation for genotype V is unclear [25][26][27] . To date there are only three isolates of genotype V and there are limited crossneutralization data in vitro. Mouse studies have shown that sera from mice immunized with SA14-5-3 and SA14-2-8 exhibit limited cross neutralization against multiple genotypes whereas SA14-14-2 does provide effective cross neutralization 24,[27][28][29] .
The flavivirus genome consists of three structural (capsid [C], premembrane [prM] and envelope [E]) and seven non-structural (NS) genes (NS1, NS2A, NS2B, NS3, NS4A, NS4B and NS5). The E protein is the major antigenic driver of the humoral immune response to the virus, thereby dictating the success of vaccine cross-protection. Studies have implicated substitutions in the E protein as the major determinants of attenuation of SA14-14-2 9,13,15,17,[30][31][32][33][34][35][36][37][38] with the NS proteins, which constitute the replication complex, playing no or a very limited role 11,17,39,40 . In fact, a chimeric SA14 virus containing SA14-14-2 prM/E genes was completely attenuated in a 3-week-old Kumming mouse model, demonstrating the importance of E protein in attenuation 41 . This mechanism of attenuation differs from that of another flavivirus LAV, yellow fever (YF) 17D where both structural and NS proteins contribute to the attenuated phenotype. It has been shown that the genome of the 17D vaccine virus is significantly less diverse than the WT parent strain from which it was derived 42,43 . The homogenous genotype of 17D has been attributed to the increased fidelity of the replication complex, which was shown using resistance to the nucleoside analog Ribavirin and is thought to contribute to the stability and safety of the vaccine 44 . Next generation sequencing (NGS) technology has yet to be applied to JEV in these contexts and so the diversity of the WT SA14 and vaccine strains is unknown. The variability of the SA14-14-2 genomic sequences described above suggests it displays more diversity than YFV 17D, potentially distinguishing their mechanisms of attenuation despite similar empirical derivation strategies and clinical efficacies.
WT JEV infection has been modeled using adult and suckling mice, and multiple outbred and inbred strains 10,11,13,17,31,34,[45][46][47][48][49] . In general, other than newborn mice, WT JEV strains only cause a lethal infection when given intracranially or by a peripheral route after the blood brain barrier has been perturbed whereas it has been shown that SA14-14-2 is avirulent in suckling mice by both intracranial and intraperitoneal routes (Yang, 2014). JEV LAVs are 100% attenuated in most of these models and in many cases are cleared so efficiently that no neutralizing antibody response is generated. For SA14-14-2, a dose of one million pfu is non-lethal in immunocompetent mice even if the virus is administered directly into the brain 36 . SA14-14-2 has only shown virulence in interferonαβγ receptor knockout (AG129) mice 47 .
In order to better understand the attenuated phenotype of SA14-14-2, we have undertaken a series of studies that compare genetic diversity, mouse virulence phenotype and immunogenicity of the vaccine strains derived from WT strain SA14 (SA14-14-2, SA14-5-3 and SA14-2-8). We demonstrate that SA14-14-2 has both high genetic diversity and sensitivity to the antiviral Ribavirin similar to WT parent SA14. This apparent genetic diversity likely explains the differences in genomic sequences reported for SA14-14-2 plus why the E protein encodes the major determinants of attenuation. In addition, despite displaying a mechanism of attenuation that distinguishes it from YF 17D LAV, SA14-14-2 maintains a similar level of attenuation in susceptible mouse models.
The SA14-14-2 stock used in our studies differed from all previously published isolates by four nucleotides that encoded one synonymous change at nucleotide 9122 and three nonsynonymous consensus changes at nucleotides 1117 (E-N47S), 1512 (E-K179E) and 7656 (NS4B-D131N) 10,51,52 (Table 1). As residues E-47N, E-179K and NS4B-131D were present in all previously published SA14-14-2 strains and in progenitor Clone12-1-7 (Genbank submission: AF416457), it can be presumed that these residues constitute part of a conserved vaccine genotype. The changes detected in the SA14-14-2 strain utilized in these studies (E-47S, E-179E and NS4B-131N) would therefore be considered mutations from this genotype, though not necessarily reversions to WT. Compared to the only other previously published commercial vaccine, the vaccine ampule of SA14-14-2 used in this study differs by three residues (E-N47S, E-M279K, and NS4B-D131N, Table 1) 10 .
SA14-14-2 exhibits significantly lower levels of entropy compared to WT SA14 (p = 0.014) (Fig. 4a) and significantly higher levels of entropy than SA14-5-3 (p < 0.003) and SA14-2-8 (p < 0.0001). Areas of the genome displaying high diversity in WT SA14 did not overlap with those showing high diversity in SA14-14-2 suggesting diversity of SA14-14-2 was due to selection during passage in PHK cells and mice, rather than as a consequence of derivation from SA14 (Fig. 4b, c). For SA14-14-2, the E protein gene displayed the highest levels of diversity, followed by NS5 (Fig. 4b).
Very low levels of diversity were found across the entire genome of SA14-5-3 ( Fig. 4d and Table S8). Only two peaks of entropy were recorded, both in NS3 (NS3-235 and NS3-446). These positions do not differentiate SA14-5-3 from SA14. Two high frequency, non-synonymous SNVs were also detected at these positions (NS3-V235A (3.5%) and NS3-N446K (40.5%)) but these SNVs did not lead to a consensus change upon passage. SA14-2-8 displayed very low levels of entropy (Fig. 4d) and few, low frequency SNVs (Table S9). This was expected as the virus was derived by (1) UV irradiation and was then (2) plaque purified during its attenuation from SA14. There are peaks of entropy at E-160 and NS2B-7 that represent reversions to WT SA14. These few areas of heterogenicity are reproduced after passage and correspond with SNVs of frequency 10% and 1.5%, respectively. Overall, the average entropy and SNV profile was not changed during passage. Both initial and final passage of SA14-2-8 showed very low levels of diversity across the entire genome.
Changes to the E protein were detected in SA14-14-2 but not SA14-2-8 when isolated from brains of AG129 mice Virus recovered at euthanasia from the brains of mice infected with SA14-14-2 was sequenced using Illumina methods. In all brains consensus changes occurred at E-S47D and E-E179K. Both were SNVs detected in the sequencing studies described above suggesting that these residues are important to how SA14-14-2 enters or replicates in the brain of mice. E-47 differentiates  SA14-14-2 and SA14-5-3 from SA14. The E-47 amino acid substitution detected in mouse brains (aspartic acid) is a reversion to SA14-5-3 not to WT SA14 (asparagine). No consensus residue changes were recorded outside the E protein. It is important to note that in comparison of SA14 to SA14-14-2, the E-E179K substitution is dictated by a single nucleotide reside change, G1117A, in the second base of the codon (AGC). However, in mouse brain G1117A is accompanied by an additional A1116G mutation in the first base of the codon such that the double mutation encodes the glutamine mutation (GAC). Due to incomplete viral sequence of SA14-2-8 virus in mouse brain by NGS, Sanger methods were used to show that no nucleotide changes took place in the E gene in the brains of AG129 mice.
Again, no consensus changes to the genome of SA14-2-8 were detected in the brains of A129 mice. As no mice infected with SA14-14-2 succumbed to infection, the brains were not assayed for viral titer as it was assumed that any infection would have been cleared by the end of the study (28 dpi).

DISCUSSION
In this manuscript, we have shown that, as expected, SA14-14-2 has the most favorable safety/immunogenicity profile of the SA14derived vaccine strains in vivo. It is also clear that the mechanism of attenuation of SA14-14-2 is different to that of a previously investigated flavivirus LAV (YFV 17D), in that 17D is characterized by a highly homogenous genotype whereas that of SA14-14-2 is heterogenous. Interestingly, SA14-14-2 is the most diverse of the three vaccine strains derived from SA14. Nevertheless, both YF 17D and JE SA14-14-2 vaccines share a similar phenotype in mouse models, good immunogenicity, cross-protective response and impressive safety profile making them highly effective LAVs 55 .
SA14-14-2 vaccine is produced in PHK cells using a seed-lot system much like that of 17D though the YFV vaccine is produced in chicken eggs 56 . In this substrate, SA14-14-2 has been shown that primary, master, working and commercial seeds (up to eight passages) and an experimental 17 passages of the virus maintain a stable genotype 9 . When these genomic sequences are compared to other SA14-14-2 GenBank submissions, it is clear that consensus sequence changes occur with extended passage in PHK cells or passage outside of this cell substrate (Table 1) 7,57 . This was shown in these studies where passage of SA14-14-2 three times in Vero cells and one in C6/36 cells to generate the working stock resulted in two additional nucleotide consensus changes, one a nonsynonymous change at nucleotide 1512 (E-K179E) and the other a synonymous change at nucleotide 9122 (NS5-482G). As E-179E is the consensus residue in SA14-2-8 it would suggest that either a lysine or glutamine can be present at this residue without affecting attenuation in humans.
The differences between the published genomic sequences of SA14-14-2 7-12 are explained by the apparently high levels of genetic diversity of this vaccine strain. This may be considered a concern for a LAV but the phenotypic data reported in this paper are very similar to that of the commercial vaccine indicating that the attenuated phenotype is maintained despite the apparent genetic variation. In fact, the mean diversity level of SA14-14-2 is only slightly, though statistically significantly, lower than WT parent strain SA14. As SA14-14-2 is highly susceptible to the antiviral Ribavirin, the cause of genetic plasticity could be localized to the NS proteins. Resistance to Ribavirin, a GTP analog, has been shown to be a marker of flavivirus replication complex fidelity 44 . Susceptibility to the drug suggests that the replication complex is not able to distinguish the analog from guanosine, which results in an increase in mutation frequency. The susceptibility of SA14-14-2 to Ribavirin differentiates the virus from yellow fever 17D, which was shown to be relatively resistant to Ribavirin 56 .
It is clear that the genetic diversity of SA14-14-2 does not make the virus susceptible to reversion to virulence. We hypothesize that this is because the genetic diversity of SA14-14-2 is not random upon passage, instead being localized to discrete regions yielding genotypes that retain an attenuated phenotype. With this knowledge it can be assumed that previously submitted SA14-14-2 sequences are correct despite consensus changes, representing a selection for a different, attenuated consensus [8][9][10][11][12]50 . As the vaccine has an impressive clinical safety record, it can be assumed that this multigenic genotype is due to multiple attenuating residues that maintain the attenuated phenotype. This type of mechanism has been shown in other flavivirus vaccine candidates and is the rationale for engineering attenuating mutations in multiple genes of rationally designed flavivirus vaccines 55,56 . In one such study, the multigenic nature of SA14-14-2 attenuation was shown using a chimeric vaccine, Chimerivax-JE, which utilizes SA14-14-2 prME genes in a YFV 17D backbone. Single reversions to WT did not have great effect on neurovirulence whereas groups of changes (especially clusters including E-F107L or E-K138E) did significantly increase virulence of the vaccine virus in mice. To this end, we have shown that the critical attenuating mutations at E-107 and E-138 are extremely stable in SA14-14-2 with no SNVs detected at these positions at any passage making it unlikely that these reversions would occur. This robust genotype may lend itself to the protective capacity of the vaccine, which has been shown to be affective against multiple JEV genotypes 8 .  The mechanism of SA14-14-2 diversity can be mapped using SA14-5-3, a progenitor of SA14-14-2, which is less genetically diverse and was over attenuated clinically. Only four amino acid residues in the replication complex (NS1-H351D, NS4B-H131D, NS5-H639Q and NS5-A671V) differentiate SA14-5-3 from SA14-14-2. Of these four residues, NS5-671 is potentially important as it extends into the pocket of the RdRp active site and is in close proximity to the conserved catalytic site in motif C of the RdRp (NS5-668). This site participates in binding GTP during preinitiation of replication and NS5-671V is conserved in all other strains of JEV as well as WNV, YFV and DENV-2 suggesting that any alteration could affect the efficiency of viral replication and potentially susceptibility to ribavirin 58,59 . A more distantly related vaccine virus, SA14-2-8 was also overattenuated clinically but was resistant to Ribavirin, showing little loss of infectivity titer even at the highest concentration of drug tested. Interestingly, SA14-2-8 maintains an NS5 gene that is identical to WT SA14, which supports the hypothesis that the flavivirus replication complex as a whole, not exclusively the RdRp, contributes to viral diversity 60 . Furthermore, SA14-2-8 only differs from SA14 at four residues (NS1-P126S, NS2B-F7V, NS3-A105G and NS4A-I255V) in the replication complex, suggesting that these residues play a key role in restricting the viral diversity of SA14-2-8.
Combining the results of the genotypic analysis and phenotypic profiles of the three vaccines, SA14-14-2 was the most genetically diverse of the vaccine strains and showed the most balanced safety/immunogenicity profile. In A129 mice, SA14-14-2 was completely attenuated, suggesting an important role for lack of type-2 interferon antagonism in the safety of the vaccine. SA14-5-3 and SA14-2-8 had very low levels of diversity and very few SNVs detected but radically distinct phenotypes in mice. SA14-5-3 was completely attenuated in the highly susceptible AG129 mouse model whereas SA14-2-8 was significantly more virulent than SA14-14-2 in both AG129 and A129 mouse models. As no reversions were detected in the brains of either AG129 or A129 mice, we speculate that this virulence phenotype is due to SA14-2-8 being more genetically similar to the parent strain WT SA14 than SA14-5-3. The two viruses differ by only 35 nucleotides and 13 amino acids, making SA14-2-8 more similar to its WT parent than to either of the two other vaccine strains tested. In particular, the maintenance of the WT residue at E-107 likely impacts the attenuation of SA14-2-8 in mice as it has been shown to be a major determinant of neurovirulence in SA14-14-2 (Fig. 7) 34 .
Although our studies have not identified a correlation to virulence, the genetic diversity differences between SA14-14-2 and SA14-2-8 could be important to vaccine immunogenicity as A129 mice that survived SA14-2-8 infection had significantly lower antibody titers compared to those infected with SA14-14-2. SA14-5-3 and SA14-14-2 are genetically very similar, with only three changes distinguishing the E proteins of the two vaccines. We postulate that the high frequency of SNVs in SA14-14-2 E and NS genes potentially affect the phenotype of the virus with respect to both immunogenicity and replicative efficiency. We also speculate that high frequency SNVs in the genotypic population of SA14-14-2 enabled the virus to cause disease in the highly susceptible AG129 model and caused the virus to be highly immunogenic in the A129 model. As SA14-14-2 contains a genetically diverse E gene (i.e., multiple SNVs at a high percentage of the RNA Fig. 7 Summary of attenuating E protein mutations in JEV vaccine strains. The three attenuated vaccine strains were empirically derived which led to distinct combinations of genotypes and phenotypes (A). It is apparent that genetic diversity doesn't correlate with attenuation in mice. It is proposed that changes to the E protein of SA14-2-8 (shown in green) and SA14-14-2 (shown in blue) are responsible for differences in mouse virulence (B). population) it is possible that the immune system sees a range of virions once administered, leading to a diversified immune response. In support of this hypothesis, WT strains of JEV show extensive genetic variation yet the SA14-14-2 vaccine is efficacious against all known strains of WT JEV 59,61,62 . Additionally, studies show that monoclonal antibodies induced by genotypes I and III viruses bind and neutralize viruses in other genotypes [63][64][65] . Our hypothesis is supported by the observation that these SNVs were not detected in the SA14-5-3 genome where the virus was 100% attenuated in the most susceptible model. In fact, the vaccine ampule passaged to create the working stock for these studies differed from the published vaccine sequence above (MH258852) at four nucleotides, which resulted in three amino acid substitutions (E-S47N, E-K279M, and NS4B-N131D) 10 . Interestingly, SNVs encoding the majority variant of the previously published commercial vaccine at each of these four positions were detected in the SA14-14-2 strain used here. This suggests that a subset of the viral population in the vaccine ampule used here was genetically identical to that of the previously published ampule 10 .
It should be emphasized that the presented findings do not extensively address the stability of the current vaccine genotype over time or passage, and are instead focused on genetic properties and phenotype of progenitor strains; the findings strongly suggest utility of using deep sequencing for monitoring of empiric vaccine population metastability to ensure both safety and protection characteristics.
The virulence of SA14-14-2 in both A129 and AG129 mouse models greatly resembled that of YFV 17D, which has an MST of 10.1 days in three to four week old AG129 mice by the subcutaneous route but is completely avirulent in age matched A129 inoculated by the same route 55 . Thus, our results are consistent in suggesting that NS genes and viral diversity have no role in attenuation of JE SA14-14-2 but possibly do play a role in immunogenicity and the E protein encodes determinants of attenuation, which is supported by amino acid substitutions involved in mouse virulence being located in the E protein (Fig. 7). Despite differing viral diversity, none of the vaccine strains display evidence of defective interfering RNA species by NGS, which was recently detected in the live attenuated influenza vaccine 66 . Thus, though 17D and SA14-14-2 maintain distinct genotypes, their phenotypes are comparable. The techniques employed during the generation of SA14-14-2 are likely responsible for its unique genotype and solid attenuated phenotype. SA14-5-3 and SA14-2-8 had been shown to be over attenuated in clinical trials leading vaccine developers to passage SA14-5-3 in suckling mice to increase the immunogenicity of the vaccine. This passaging likely selected a population for SA14-14-2 that was not clonal, unlike 17D which was passaged hundreds of times in chicken tissue. The diverse yet attenuated viral population of SA14-14-2 provides cocktail of antigens that induce robust and protective immune response and represents a unique attenuation mechanism for liveattenuated vaccines.

Passaging of viral stocks
A freeze-dried vaccine ampule of SA14-14-2 was obtained from NICPBP (Beijing) in 1987/1988. We do not have the lot number. Thus, the vaccine sample is derived from manufacture in the original production facility, which is different to the current vaccine that is manufactured in the new facility. The vaccine was reconstituted in HPLC grade water and passaged once in C6/36 and twice in Vero cells. SA14-2-8 12 was passaged once in C6/ 36 cells and once in Vero cells. SA14-5-3 12 was passaged three times in Vero cells. Wild-type (WT) strain SA14 was obtained from the World Reference Center for Emerging Viruses and Arboviruses (WRCEVA, Galveston, Texas). Passages in Vero and C6/36 cells occurred in minimal essential media (MEM) supplemented with L-glutamate, penicillin-streptomycin, non-essential amino acids, and 2% fetal bovine serum by volume. Viral infectivity assays were undertaken in Vero cells by focus forming assay using anti-JEV mouse immune ascites fluid kindly provided by the World Reference Center for Emerging Virus and Arboviruses (WRCEVA, UTMB), biotinylated goat anti-mouse IgG and NuetrAvidin-HRP (ThermoFischer) 71 .

Next generation sequencing (NGS) of JEV vaccine strains
RNA was extracted from the seed stock and final passage of SA14 and SA14-14-2, plus the final two passages each of SA14-2-8 and SA14-5-3 using the Qiagen QIAmp viral RNA isolation kit (Qiagen). cDNA libraries were prepared using random hexamers (TruSeq RNA V2 kit) and sequenced on either the Illumina HiSeq1500 platform or the NextSeq550 platform by the UTMB NGS Core. Two viruses (SA14-14-2 and SA14-2-8) were sequenced on both platforms to evaluate consistency. No variability was identified. A de novo consensus sequence was generated for each sample using ABySS V1.3.7 assembly of paired end reads with k values from 20 to 40. The consensus sequences of each virus were then compared to previously published consensus sequences of the same strain using BioEdit (V 7.2.5).
The reads obtained from Illumina sequencing were trimmed to a minimum length of 35 bases and quality controlled (Q-score ≥30) using the open source software Trimmomatic (V 0.39). Reads were aligned using Bowtie2 (Version 2.2.6) very-sensitive-local presets to the de novo consensus sequences generated. The resulting alignment was sorted by coordinate using PicardTools (V 2.20.5) SortSam and PCR duplicates were removed using PicardTools MarkDuplicates. Shannon entropy was calculated using previously published means 44,60 . Single nucleotide variants (SNVs) were called using the open source software LoFreq V2 with a variant cutoff of 1% 67 . Shannon's entropy and SNVs were calculated for all samples excluding the final passage of SA14-5-3, which had low coverage.

Viral infectivity reduction of antiviral ribavirin
Viral infectivity reduction curves in the presence of ribavirin were undertaken as previously reported 44 . Briefly, Vero cells were inoculated with SA14, SA14-14-2, SA14-5-3 or SA14-2-8 at a multiplicity of infection of 0.05. Ribavirin was added at 2X concentration (1:2 dilution curve starting at 3 mM), allowed to incubate for 30 minutes at room temperature and then MEM supplemented with 2% FBS added to reduce the concentration to 1X. After 48 hours, cell culture supernatants were collected, and virus titrated via focus forming assay. Each assay was completed in triplicate.

Mouse virulence studies
All animal procedures were approved by the University of Texas Medical Branch (UTMB) Institutional Animal Care and Use Committee and studies were carried out in compliance with the recommendations of the Guide for the Care and Use of Laboratory Animals (National Research Council).
Following inoculation mice were weighed daily and monitored for morbidity for 28 days. Animals exhibiting clinical signs of neurological disease or with weight loss >20% of initial body weight were humanely euthanized and counted as dead on the following day of the study for analysis. On day three post inoculation, blood was collected by retroorbital bleed to quantify viremia using iScript cDNA synthesis kit (BioRad) and PF1S and PF2R-bis primers 68 . In addition, for some animals the brains were harvested at the time of euthanasia and homogenized using 1.5 mm zirconium beads. After clarification by centrifugation, viral load was determined using focus-forming assays and expressed as FFU/gram. Viral RNA was extracted using Qiagen Viral RNA isolation kits and sequenced using NGS. The E protein gene of viruses that returned incomplete sequences by NGS were Sanger sequenced using E protein primers described previously 69 Only changes to the E protein are reported.
At the end of the study, surviving animals were sacrificed, serum was collected and heat-inactivated by incubation at 56°C for 30 minutes. The samples were stored frozen at −20°C until the neutralizing antibody titers were determined by focus reduction neutralization tests (FRNT 50 ). Briefly, 2-fold serial dilutions were incubated with 50 FFU of infecting virus that was subsequently used to infect Vero cells. After four days, cells were fixed E.H. Davis et al.