Second-Generation Live-Attenuated Candid#1 Vaccine Virus Resists Reversion and Protects against Lethal Junín Virus Infection in Guinea Pigs

ABSTRACT

Live-attenuated virus vaccines are highly effective in preventing viral disease but carry intrinsic risks of residual virulence and reversion to pathogenicity. The classically derived Candid#1 virus protects seasonal field workers in Argentina against zoonotic infection by Junín virus (JUNV) but is not approved in the United States, in part due to the potential for reversion at the attenuating locus, a phenylalanine-to-isoleucine substitution at position 427 in the GP2 subunit of the GPC envelope glycoprotein. Previously, we demonstrated facile reversion of recombinant Candid#1 (rCan) in cell culture and identified an epistatic interaction between the attenuating I427 and a secondary K33S mutation in the stable signal peptide (SSP) subunit of GPC that imposes an evolutionary barrier to reversion. The magnitude of this genetic barrier is manifest in our repeated failures to rescue the hypothetical revertant virus. In this study, we show that K33S rCan is safe and attenuated in guinea pigs and capable of eliciting potent virus-neutralizing antibodies. Immunized animals are fully protected against lethal challenge with virulent JUNV. In addition, we employed a more permissive model of infection in neonatal mice to investigate genetic reversion. RNA sequence analysis of the recovered virus identified revertant viruses in pups inoculated with the parental rCan virus and none in mice receiving K33S rCan (P < 0.0001). Taken together, our findings support the further development of K33S rCan as a safe second-generation JUNV vaccine.

IMPORTANCE Our most successful vaccines comprise weakened strains of virus that initiate a limited and benign infection in immunized persons. The live-attenuated Candid#1 strain of Junín virus (JUNV) was developed to protect field workers in Argentina from rodent-borne hemorrhagic fever but is not licensed in the United States, in part due to the likelihood of genetic reversion to virulence. A single-amino-acid change in the GPC envelope glycoprotein of the virus is responsible for attenuation, and a single nucleotide change may regenerate the pathogenic virus. Here, we take advantage of a unique genetic interaction between GPC subunits to design a mutant Candid#1 virus that establishes an evolutionary barrier to reversion. The mutant virus (K33S rCan) is fully attenuated and protects immunized guinea pigs against lethal JUNV infection. We find no instances of reversion in mice inoculated with K33S rCan. This work supports the further development of K33S rCan as a second-generation JUNV vaccine.

INTRODUCTION

Arenaviruses (genus Mammarenavirus) are endemic in rodent populations worldwide, and several species can be transmitted to humans to cause severe, life-threatening hemorrhagic fevers. In the absence of licensed vaccines and effective antiviral therapeutics, these viruses pose grave threats to public health and national security (1). Nearly half a million people are infected annually by Lassa virus (LASV) in West Africa, causing more than 5,000 deaths (2). Until recently, several thousand agricultural workers suffered from Junín virus (JUNV) infection each year in the Pampas regions of Argentina (3). Fortunately, the incidence of Argentine hemorrhagic fever (AHF) has now been significantly reduced thanks in large part to the live-attenuated JUNV vaccine Candid#1 (4).

The Candid#1 vaccine was developed in the 1980s by a consortium of U.S. and Argentine researchers using a conventional attenuation strategy in which a pathogenic JUNV isolate (XJ) was serially passaged through guinea pigs and mice and in cell culture (4). The resultant Candid#1 strain was attenuated in mice, guinea pigs, and rhesus macaques and engendered protective immunity against lethal JUNV infection (5). Under the auspices of the U.S. Food and Drug Administration, the vaccine was proven to be safe and efficacious in human volunteers (6). The Candid#1 vaccine became available in Argentina in 1992 for seasonal use in at-risk agricultural workers, and as of 2006, more than a quarter of a million doses have been administered (3). However, this vaccine is unlikely to be licensed in the United States due to concerns regarding the stability of the attenuation phenotype of Candid#1.

During the course of passage leading to Candid#1, the virus acquired a number of mutations in the envelope glycoprotein GPC (Fig. 1) and throughout the bipartite single-stranded ambisense RNA genome (7, 8). Recent reverse-genetics studies have localized the key determinant of Candid#1 attenuation to the phenylalanine-to-isoleucine substitution at position 427 (F427I) in the transmembrane domain (TMD) of the GP2 fusion subunit of GPC (8, 9). F427 is conserved in all pathogenic strains of JUNV, and virulence is abrogated upon the introduction of the F427I mutation. The mechanism whereby the change to I427 mediates attenuation is unknown. The four additional amino acid changes identified in Candid#1 GPC are not sufficient for attenuation but may contribute to virus fitness in cell culture and/or provide support to the attenuation phenotype (9).

FIG 1.

Derivation of Candid#1. (Top) Clinical isolate XJ was serially passaged in mouse brain (MB), guinea pigs (GP), and fetal rhesus lung cells (FRhL) to generate Candid#1. The number of passages separating the intermediate isolates is shown, and their virulence (v) or attenuation (a) in the respective animal species is noted. (Bottom) Amino acid substitutions in GPC detected in Candid#1 and selected intermediate isolates. Position 427 in GP2 is highlighted. (Data were taken from reference 8.)

Although the Candid#1 vaccine is reported to have a favorable safety profile (10), follow-up among immunized migrant workers is incomplete, and reliance on a single nucleotide change for attenuation makes reversion all but inevitable. Indeed, we have demonstrated a strong evolutionary drive toward reversion at this position in cell culture (11). Revertants that arise in the course of vaccination may pose a threat to immunized individuals and their naive or immunocompromised contacts (12). Furthermore, the risk of reversion raises weighty concerns among regulators and vaccine manufacturers. This liability of the Candid#1 vaccine motivates our studies to design a recombinant Candid#1 (rCan) variant with an expanded margin of safety.

GPC is responsible for mediating virus entry into the host cell and is the target for virus-neutralizing antibodies (VN Abs) (13–15) and small-molecule entry inhibitors (16, 17). Upon GPC binding to its cell-surface receptor, transferrin receptor 1 for the pathogenic New World (NW) viruses, including JUNV (18), and α-dystroglycan for Old World (OW) arenaviruses (19), the virion is endocytosed, and GPC-mediated fusion of the viral and cellular membranes is triggered by low pH in the maturing endosome (20). GPC is synthesized as a precursor that trimerizes and is cleaved by the Golgi S1P/SKI-1 protease (21–23) to generate the mature receptor-binding (GP1) and transmembrane fusion (GP2) subunits. Unlike other class I viral fusion proteins, GPC retains a 58-residue signal peptide as a third noncovalently associated subunit in the mature complex (Fig. 2) (24–26). We have shown that the stable signal peptide (SSP) contains two hydrophobic regions (TM1 and TM2) that span the membrane to form a hairpin structure (27). SSP is myristoylated at its cytoplasmic N terminus (25), and this modification is important for efficient cell-cell fusion activity (25, 28, 29). Moreover, we have discovered that TM1 and the central ectodomain loop of SSP interact with the TMD and membrane-proximal external region (MPER) of GP2 to form a functional unit that senses endosomal pH and triggers the large-scale conformational change in GPC leading to membrane fusion (30–32). We have shown that amino acid substitutions that reduce positive polarity at K33 in the ectodomain loop of SSP correspondingly decrease the pH required for cell-cell fusion by ectopically expressed wild-type (F427) GPC and that alanine and serine replacements abolish fusion at any physiologically relevant pH (31). Importantly, fusion activity can be rescued by complementing mutations in the TMD and MPER of GP2, including by the attenuating F427I mutation in Candid#1 (11, 31). This epistatic interaction extends to the intact virus, and K33S rCan is readily rescued by reverse genetics and displays growth properties in cell culture similar to those of the parental virus (11). As reversion to the pathognomonic F427 in K33S rCan is expected to yield a fusion-defective GPC, we speculated that this epistasis can impose a strong evolutionary barrier against reversion. Indeed, K33S rCan retains both K33S and F427I upon serial passage in cell culture under conditions that result in the fixation of the I427F mutation in parental rCan (11).

FIG 2.

Subunit organization of arenavirus GPC. (Left) Schematic drawing of the tripartite GPC protomer. Position 427 in GP2 and position 33 in SSP are labeled and indicated by red and black circles, respectively. Features of GPC include SSP myristoylation (myr), the binuclear SSP-GP2 zinc finger (54, 55) (gray circles), and the N- and C-terminal heptad repeat regions in GP2 (thick lines), diagnostic of class I viral fusion proteins. The drawing is not to scale, and the relative position of SSP is unknown. (Right) Cryo-electron tomography (CryoET) reconstruction of the GPC trimer at a 14-Å resolution (from reference 56). SSP and the transmembrane domain of GP2 are not visible.

Here, we characterize K33S rCan in small-animal models of JUNV infection to evaluate its safety, genetic stability, immunogenicity, and protective efficacy. We show that K33S rCan remains attenuated in guinea pigs and elicits strong protective immunity against infection by the pathogenic Romero strain of JUNV. A single immunization induces a potent VN Ab response and prevents viremia, morbidity, and mortality upon lethal challenge. Because Candid#1 cannot be reisolated from immunized guinea pigs (33), we investigated reversion to the pathognomonic I427F genotype by using an established neonatal mouse model in which intracranial (i.c.) inoculation of Candid#1 is lethal. RNA sequence (RNA-Seq) analysis of virus isolated from brain tissue revealed significant I427F reversion in mice receiving parental rCan but no instances of reversion in mice inoculated with K33S rCan. Collectively, our studies demonstrate the propensity for reversion by rCan in live animals and support the genetic stability, safety, and protective efficacy of a second-generation K33S rCan vaccine.

RESULTS

The Hartley guinea pig provides an established model for lethal JUNV infection by pathogenic isolates of JUNV and for attenuation of the Candid#1 vaccine strain (4, 9, 34). This model serves as a release criterion for the newly manufactured Argentine vaccine (35) and is frequently used experimentally to demonstrate vaccine protection against subsequent challenge with virulent JUNV. Accordingly, we employed this model to determine the safety, immunogenicity, and protective efficacy of the K33S rCan vaccine (Fig. 3A). In brief, guinea pigs were vaccinated intraperitoneally (i.p.) with 10³ or 10⁴ PFU of either K33S rCan or the parental rCan (day −23) and monitored for 21 days to detect any adverse clinical signs and to assess immunogenicity. Animals were subsequently challenged by i.p. inoculation with a known lethal dose (100 PFU) of recombinant Romero JUNV (rRom) (day 0) and monitored for an additional 36 days to assess clinical signs of disease, viremia, and mortality.

FIG 3.

Safety and immunogenicity of K33S rCan in guinea pigs. (A) Overview of K33S rCan safety, immunogenicity, and efficacy study design. (B and C) Guinea pigs in each group (n = 5) were vaccinated by i.p. injection containing 10³ or 10⁴ PFU of K33S rCan or rCan and monitored for body temperature (B) and weight change (C) for 23 days. Animals vaccinated with the PBS placebo and unvaccinated normal controls were included for comparison. The body temperature data represent the group means and standard errors of the means (SEM). The percent changes in weights represent the group means and SEM relative to initial animal starting weights obtained 4 days before vaccination. (D and E) Three weeks after immunization, blood samples were collected, and sera were analyzed for anti-NP (D) and VN Ab (E) titers. Panels show results from individual animals as well as group means and SEM. One guinea pig in the low-dose K33S rCan group was excluded from the analysis due to vaccination failure. Results below the limits of detection (1:20 and 1:10, respectively) were assigned a value of 1. (F) The correlation of anti-NP versus VN Ab titers was assessed by Pearson’s r test (r = 0.74).

Attenuation and immunogenicity of rCan and K33S rCan viruses.

As anticipated, infection with rCan or K33S rCan was clinically benign in guinea pigs at both low and high doses, with no apparent behavioral changes or significant changes in body temperature or weight (Fig. 3B and C). Immunogenicity was evaluated 21 days after immunization using a sandwich enzyme-linked immunosorbent assay (ELISA) to quantitate antibodies directed against the virus nucleoprotein (NP). We found that all but one immunized animal developed high levels of anti-NP antibodies (Fig. 3D). The one guinea pig that failed to induce detectable levels of antibody likely represents a failed immunization, and this animal was excluded from the analysis. Of note, antibody titers were modestly lower in the K33S rCan-inoculated animals at both dose levels, suggesting a lesser degree of virus replication in the K33S mutant. While the two viruses appeared to grow comparably in cell culture, K33S rCan foci are smaller than those of the parental rCan (11), and this property may contribute to somewhat reduced replication in vivo.

The VN Ab response was quantitated using a conventional 50% plaque reduction neutralization titer (PRNT₅₀) assay (36). Candid#1 virus was incubated with serial dilutions of sera from the immunized animals, and the remaining infectious virions were enumerated by infection of African green monkey Vero cells. Using this assay, we detected potent VN Abs in all immunized animals (Fig. 3E). Both rCan and K33S rCan vaccines were comparably immunogenic at each dose level, and titers increased 5- to 10-fold at the higher dose. At all dose levels, VN Ab titers were higher than those reported in immunized persons and convalescent-phase plasma (37). Overall, VN Ab titers correlated with levels of antibodies directed to NP (Fig. 3F), suggesting comparable exposures of VN epitopes in both vaccine viruses. We conclude that the K33S rCan vaccine is similar to the parental rCan vaccine in safety and immunogenicity.

Protective efficacy against lethal JUNV challenge.

VN Ab levels have been shown to correlate with vaccine protection in immunized individuals, and convalescent-phase plasma containing VN Abs is used therapeutically to treat AHF in Argentina (3, 38). In order to evaluate and compare the protective efficacies of the rCan and K33S rCan vaccines, guinea pigs were transferred to animal biosafety level 3+ (ABSL3+) containment after 23 days and challenged by i.p. inoculation with a lethal dose (100 PFU) of rRom (Fig. 3A, day zero). All animals that had received phosphate-buffered saline (PBS) as a placebo vaccine showed elevated body temperature and weight loss starting at 6 days postinfection (p.i.) (Fig. 4A and B), and all succumbed to infection by day 15 (Fig. 4C). The above-mentioned animal in which vaccination had apparently failed also succumbed to rRom infection (not shown). In contrast, guinea pigs that received the rCan or K33S rCan vaccine at either dose were fully protected. Immunized guinea pigs did not display elevated body temperature in response to rRom challenge and continued to gain weight comparably to sham-challenged control animals. Viremia was undetectable on day 12 p.i. in immunized guinea pigs, whereas animals that received placebo carried >10⁵ infectious particles/ml of serum (Fig. 4D). Together, these findings demonstrate that K33S rCan offers complete protection against disseminated JUNV infection and disease.

FIG 4.

Efficacy of K33S rCan vaccination against lethal challenge with JUNV Romero. (A to C) Vaccinated guinea pigs (n = 5/group) were infected i.p. with 100 PFU of rRom and monitored for body temperature (A), weight change (B), and survival outcome (C) for 36 days. Unvaccinated, sham-infected controls were included for comparison. The body temperature data represent the group means and SEM. The percent changes in weights represent the group means and SEM for surviving animals relative to their starting weights on the day of rRom challenge. ** indicates a P value of <0.01 compared to animals vaccinated with the PBS placebo, as determined using the Mantel-Cox log rank test. (D) Twelve days after challenge, blood samples were collected, and sera were assayed for infectious rRom titers. The panel shows individual animal titers as well as group means and SEM. The x axis is drawn at the limit of detection (1.49 CCID₅₀/ml), and virus was undetected in all immunized animals. *** indicates a P value of <0.001 compared to animals vaccinated with the PBS placebo.

Reversion of the attenuating F427I mutation in rCan.

While comparable to rCan in safety, immunogenicity, and protective efficacy, K33S rCan was designed to minimize reversion at the attenuating locus in Candid#1. We previously demonstrated that K33S rCan does not revert in cell culture under conditions that elicit facile reversion in rCan (11). We reasoned that resistance to reversion in K33S rCan is due to the synthetic lethal interaction between K33S and the revertant F427. Notably, several attempts to generate the hypothetical K33S I427F rCan revertant were unsuccessful despite concurrent success in rescuing the parental or K33S rCan (not shown). We conclude that back-mutation to the pathognomonic F427 in K33S rCan results in a nonviable virus, thereby precluding reversion.

Given the documented absence of viremia in guinea pigs immunized with Candid#1 (33), we were unable to characterize reversion in this model. In contrast, neonatal (2-day-old) mice are susceptible to lethal infection by Candid#1 delivered i.c. (8, 39–41). Extensive virus replication in this model thus provides an opportunity to study virus evolution and reversion in a live animal. To this end, we confirmed that infection by rCan or K33S rCan was likewise fatal (Fig. 5A). Pups receiving 10 PFU of either recombinant virus succumbed to infection within 16 days, whereas cohort animals inoculated with PBS survived.

FIG 5.

Susceptibility of 2-day-old mice to K33S rCan. Cohorts of CD-1 mouse pups from 2 litters were inoculated by i.c. injection with 10 PFU of K33S rCan (n = 9) or rCan (n = 10) or the PBS vehicle (sham infected) (n = 4). Survival outcomes (A) and virus titers in the brains of moribund mice (B) are shown. Individual pup brain homogenate titers are shown, as are group means and SEM. Points circled in black indicate pups from which virus was expanded for sequence analysis. Differences in K33S rCan and rCan group survival outcomes and brain viral titers were not statistically significant.

Genetic changes were characterized in virus isolated from the brains of moribund mice at 13 to 15 days p.i. (Fig. 5B). Three pups from both the rCan and K33S rCan groups were chosen based on higher virus loads in the brain homogenate, and 10⁴ PFU of infectious virions from each pup were expanded in Vero cells to obtain sufficient viral RNA for RNA-Seq analysis. The goal was to amplify an equally complex population of the virus from each animal in order to accurately compare the frequencies of reversion. We note that expansion per se does not result in genetic changes (11, 29), as Candid#1 is traditionally grown in Vero cells. Supernatants from the expanded Vero cell cultures were pooled by group, and RNA was isolated from each pool for RNA-Seq analysis. The absolute amount of viral RNA in each sample could not be quantified as the vast majority of RNA isolated from the cell culture supernatant is cell derived.

RNA-Seq was performed by Genewiz. In brief, rRNA was depleted from the samples, and barcoded libraries were multiplexed onto a single flow cell for 2-by-150 paired-end sequencing using an Illumina HiSeq4000 instrument. Trimmed reads were initially filtered against the African green monkey (Chlorocebus sabeus) genome to remove excess cell-derived sequences and subsequently aligned to the two segments of the reverse-genetics rCan genome (11, 34). GPC and NP are encoded in an ambisense fashion by the S (small) RNA segment, and likewise, the Z matrix protein and L RNA-dependent RNA polymerase (RdRP) are encoded by the L (large) segment. Sequences were inspected using the Broad Institute Integrated Genomics Viewer (IGV).

Sequence analysis revealed 115 instances of I427F reversion in the brains of mice inoculated with rCan (in 19,784 total reads) and no instances in mice inoculated with K33S rCan (in 2,846 total reads) (Table 1). The frequency of reversion in rCan (nucleotide A1367T) was significantly higher than that in K33S rCan (P < 0.0001 by one-sided Fisher’s exact test) and higher than that predicted based on the 10⁻⁴ error rate of the virus RdRP (42) (multinomial test with Monte Carlo P value of <0.0001). Despite the relatively low frequency of revertant genomes in rCan-inoculated mice (0.58%), these findings clearly demonstrate that positive selection for back-mutation to the pathognomonic F427 is active in live animals as well as in cell culture. Importantly, the failure to detect reversion in the K33S rCan population is consistent with the significant barrier to reversion imposed by synthetic lethality.

TABLE 1.

Genetic variants in mouse brains^a

^a#, variants present at ≥0.5% of total reads. *, changes incorporated through the K33S mutagenic oligonucleotide are included in the reference genome (see reference 11) and are thus not listed; no variations at these positions were noted. Variants encoding GPC amino acid 427 are highlighted in red; boldface entries indicate the virus gene bearing the variant. Abbreviations: Ref, reference nucleotide; Var, variant nucleotide; AA, amino acid; IGR, intergenic region.

Additional variants identified at a frequency of ≥0.5% are listed in Table 1. A nucleotide C342T variant in the S RNA is present at a frequency of 41% in the K33S rCan population but absent from rCan. The resulting P85L mutation is in a highly variable region of GP1 that is unresolved in crystal structures (14, 43). Another change, in the L RdRP (nucleotide T2082C in the L RNA; amino acid K1668T), is present in 27% of rCan sequences and absent in K33S rCan. Neither of these changes has been noted in our previous studies of rCan viruses passaged in cell culture (11, 29), and their presence in only one or the other of the present viruses suggests that they are not associated with growth in mice. The consequences of these variations are unclear.

DISCUSSION

Live-attenuated virus vaccines have been among the most successful preventative interventions in medical history (e.g., smallpox, oral poliovirus, measles-mumps-rubella [MMR], varicella, and yellow fever). These vaccines provide strong and durable immunity but carry intrinsic risks of residual virulence and reversion to a pathogenic form (44). This liability is perhaps best exemplified by the emergence of pathogenic variants upon replication of the oral polio vaccine in the host, leading to persistent outbreaks of vaccine-derived poliovirus throughout the world (45). These concerns are heightened for the Candid#1 vaccine due to its reliance on a single nucleotide change for attenuation (8, 9). Although there have been no reports of Candid#1 reversion or vaccine-derived disease among at-risk agricultural workers in Argentina, we note that circulating vaccine-derived poliovirus was first reported many years and 100 million doses after introduction (46) and that it was an additional 20 years before the molecular basis for phenotypic reversion was defined (47, 48).

Our previous studies of Candid#1 evolution in cell culture confirmed a strong drive toward reversion at position 427. Serial passage of parental rCan in the presence of a mild lysosomotropic agent or in response to introduced mutations resulted in the complete replacement of the attenuating I427 by the pathognomonic F427 (11). In the present study, we report that positive selection for the revertant virus also occurs in the established neonatal mouse model of lethal rCan infection (8, 39–41). Within 13 to 15 days of infection with parental rCan, 0.58% of the virus population contained the revertant I427F allele. The lower frequency of this allele in mice than that obtained in cell culture may reflect the short period of growth in the animal and the continued presence of the original virus population in the brain. Nonetheless, our finding provides proof of principle that rCan can revert to the pathognomonic genotype in live animals.

We had discovered that reversion in cell culture could be blocked by taking advantage of the epistatic interaction between K33S in SSP and I427 in GP2 (11). As formal proof of the synthetic lethal interaction between K33S and the revertant F427, we report that repeated attempts to rescue such a revertant by reverse genetics have been unsuccessful, consistent with the lack of viability of the hypothetical K33S I427F virus. In keeping with our previous findings in cell culture, we detected no instances of K33S rCan reversion in the brains of 2-day-old mice following i.c. inoculation. These findings validate and extend our previous conclusion that epistasis between K33S and I427 creates a formidable barrier to reversion in K33S rCan.

Complementation of the K33S defect by I427 rescues membrane fusion activity in K33S rCan, but its effects on virulence were unknown. By using the gold-standard model for Candid#1 attenuation, we demonstrate that K33S rCan remains fully attenuated in guinea pigs. Infection with 10³ or 10⁴ PFU of either rCan or K33S rCan is benign, with immunized guinea pigs showing no clinical signs of disease. In contrast, inoculation of 100 PFU of the pathogenic rRom is uniformly lethal. Furthermore, both attenuated rCan viruses are strongly immunogenic in this model. VN Ab titers elicited by rCan and K33S rCan are comparable and higher than those reported in immunized persons and in convalescent-phase plasma used therapeutically (3, 37, 38). To directly assess protective efficacy, immunized guinea pigs were challenged with a lethal dose of virulent rRom. All animals receiving either dose of the rCan or K33S rCan vaccine were fully protected. rRom viremia was undetectable in vaccinated guinea pigs, whereas animals receiving a placebo vaccine carried >10⁵ infectious particles/ml of serum. Taken together, these studies indicate that K33S rCan is comparable to parental rCan in attenuation and in its ability to elicit VN Abs and protective immunity in guinea pigs.

The safety and protective efficacy of K33S rCan in guinea pigs and its genetic resistance to reversion in mice support the development of K33S rCan as a second-generation JUNV vaccine. It is furthermore likely that key features of the critical interaction between SSP and GP2 subunits are conserved among mammarenaviruses. Thus, future investigations may identify analogous attenuating determinants and epistatic interactions in other hemorrhagic fever arenaviruses, including LASV.

MATERIALS AND METHODS

Molecular reagents, cells, and antibodies.

Candid#1 reverse-genetics plasmids mPol-I-Sag, mPol-I-Lag, pC-L, and pC-NP were kindly provided by Juan Carlos de la Torre (Scripps Research Institute, La Jolla, CA) (34). The pathogenic recombinant Romero virus (rRom) was kindly supplied by Slobodan Paessler (University of Texas Medical Branch, Galveston, TX) (34). BHK-21 and Vero-76 cells were obtained from the ATCC. Monoclonal antibody (mAb) AG12 directed to JUNV nucleoprotein (NP) (13) was obtained through the NIAID Biodefense and Emerging Infections Research Program (BEI Resources).

Reverse-genetics methodology.

Recombinant viruses were generated as previously described (29, 34). In brief, BHK-21 cells were cotransfected with the two antigenome plasmids mPol-I-Sag and mPol-I-Lag and plasmids expressing the viral polymerase and nucleoprotein (pC-L and pC-NP, respectively) to initiate replication of the transcribed genomic RNAs. The rescued viruses were subsequently expanded in Vero cells, and virus stocks were prepared from freeze-thawed cells and the culture supernatant.

The derivation and cell culture characterization of K33S rCan were previously described (11). In the present study, rCan and K33S rCan were amplified upon rescue by 2 to 4 passages in Vero cells to generate stocks of 3.6 × 10⁶ PFU/ml and 1.5 × 10⁶ PFU/ml, respectively. Sequences spanning the relevant regions of GPC (K33S and F427I) were initially confirmed by reverse transcription and Sanger sequencing. All studies using rCan viruses were performed using BSL2 and ABSL2 facilities and operations, as approved by the respective Institutional Biosafety Committees of the University of Montana and Utah State University (USU). Work with pathogenic rRom was conducted by Candid#1-vaccinated personnel using BSL3+ and ABSL3+ facilities and operations at USU (49).

Ethics statement.

All animal procedures performed at USU complied with USDA guidelines and were approved by the USU Institutional Animal Care and Use Committee. USU is an AAALAC-accredited institution.

Animals.

Male Hartley guinea pigs (300 to 350 g) and timed-pregnant CD-1 mice were obtained from Charles River Laboratories. IPTT-300 electronic transponders were subcutaneously implanted in guinea pigs for identification and temperature measurements using a handheld DAS 6002 scanner (Biomedic Data Systems).

Safety, immunogenicity, and efficacy of experimental rCan and K33S rCan vaccine viruses in guinea pigs.

Four days prior to immunization, guinea pigs were weighed and grouped (n = 5 for vaccine groups, and n = 3 for the sham infection group) to minimize weight variation across the cohorts. The rCan and K33S rCan viruses were administered by i.p. injection at doses of either 10³ or 10⁴ PFU. The PBS vehicle was administered to the vaccine placebo group. Three weeks after immunization, blood was obtained from each animal by cranial vena cava sampling under deep isoflurane anesthesia, and sera were analyzed for VN Ab titers. Two days later, the guinea pigs were transferred to ABSL3+ containment and challenged i.p. with 100 PFU of pathogenic rRom. The animals were bled a second time 12 days after rRom challenge to assess viremia and observed for a total of 36 days p.i. for morbidity and mortality. Body weights and temperatures were recorded at specified times during the course of the study, and moribund animals, defined by body temperatures below 3°C of their starting temperature or having lost 20% or more of their starting weight, were humanely euthanized.

Anti-NP antibody ELISA.

Antibodies directed to NP were detected using a modification of a previously described sandwich ELISA (50). Briefly, mAb AG12 was used to capture NP in solubilized rCan virus stocks onto the plate. Serial dilutions of guinea pig sera were incubated to detect antibodies directed to NP, which were visualized using a goat anti-guinea pig IgG(H+L) antibody-horseradish peroxidase (HRP) conjugate (Novex) and the Sure Blue TMB (3,3',5,5'-tetramethylbenzidine) substrate (KPL). Optical densities were measured at 450 nm using a VMax microplate reader (Molecular Devices). Titers were defined as the reciprocal of dissociation constant (K_d) values generated using nonlinear curve fitting and the one-site total binding model implemented in Prism (GraphPad Software).

Plaque reduction neutralization test.

The PRNT₅₀ (dilution of serum required to reduce plaques by 50%) was determined as previously described (36). Briefly, 50 PFU of Candid#1 virus was incubated at 37°C for 1 h with equal volumes of serial dilutions (1:10 to 1:5,120) of guinea pig sera. Prior to incubation, sera were heat inactivated in a 55°C water bath for 30 min. The mixtures were then added to Vero cell monolayers and incubated with occasional rocking at 37°C for 1.5 h. After the removal of the virus inoculum, an overlay of 1.3% methylcellulose was added, and the plates were incubated for 7 days for plaque development. The plaques were resolved by removing the overlay and staining the cells with crystal violet. The PRNT₅₀ values were determined by regression analysis of the number of plaques at each dilution relative to the number of plaques observed with normal serum from sham-infected guinea pigs.

Serum virus titers.

Virus titers were determined using a cell culture assay as previously described (51). Briefly, serum samples were serially diluted and added to triplicate wells of Vero cell monolayers in 96-well microplates. Foci of viral cytopathic effect were enumerated at 11 days p.i., and 50% cell culture infectious dose (CCID₅₀) endpoints were calculated as described previously (52). The assay limit of detection was 1.49 CCID₅₀/ml.

Intracranial inoculation of 2-day-old mice.

Two days after birth, pups from mixed-sex litters from two timed-pregnant CD-1 mice were inoculated i.c. with 10 PFU rCan (n = 10 [5 from each litter]) or K33S rCan (n = 9 [5 from the first litter and 4 from the second litter]) and returned to their dams to continue suckling. Four pups (2 from each litter) were sham inoculated with 10 μl of the PBS vehicle. Moribund mice were euthanized, and the brains were collected for virus quantification and amplification for RNA-Seq analysis. Brains were homogenized in a fixed volume of minimum essential medium, and virus titers were determined by a plaque assay. Viruses from three selected animals per group were amplified in Vero cells, and culture supernatants were pooled by group for RNA isolation using the Quick-RNA viral kit (Zymo Research). The quantity and quality of the RNA were assessed by using a NanoDrop instrument (Thermo Fisher Scientific), and samples were stored at −80°C for RNA-Seq analysis.

RNA-Seq of virion RNAs.

Virion RNA samples were provided to Genewiz (South Plainfield, NJ) for library construction and deep sequencing. In brief, RNA was depleted of cellular rRNA using the FastSelect rRNA HMR (human/mouse/rat) kit (Qiagen) and fragmented by incubation for 15 min at 94°C. Barcoded sequencing libraries were constructed using the NEBNext Ultra II RNA library preparation kit for Illumina (New England BioLabs) based on the manufacturer’s recommendations. Libraries were validated using a Tapestation 4200 instrument (Agilent), quantified using a Qubit 2.0 fluorometer (Thermo Fisher Scientific), and then multiplexed onto a single flow cell for sequencing using a 2-by-150 paired-end configuration in an Illumina HiSeq4000 instrument. Raw data were converted to FASTQ files and demultiplexed using Illumina bcl12fastq 2.17 software.

Bioinformatics analysis was also performed by Genewiz. In brief, sequence reads were trimmed and screened for quality using Qiagen CLC Genomics Server 9.0. Cellular RNA sequences, which comprise ∼99% of reads, were removed by annealing to the reference genome for African green monkey Vero cells (assembly Chlorocebus_sabeus_1.1 [GenBank accession number GCA_000409795.2]), and viral sequences were then aligned to the sequence of the rCan S and L reverse-genetics plasmids (modified from sequences under GenBank accession numbers AY746353.1 and AY746354.1, respectively). Variants present at a frequency of ≥0.5% were extracted using the CLC Genomics Server program, and the aligned sequences were visualized using the Broad Institute Integrated Genomics Viewer (IGV_2.6.3) (53).

Statistical analysis.

The Mantel-Cox log rank test was used for analysis of Kaplan-Meier survival curves. One-way analysis of variance (ANOVA) with Dunnett’s posttest to correct for multiple comparisons was performed to evaluate differences in PRNT₅₀ values and virus titers. These analyses were performed using Prism 9 software (GraphPad Software). Differences in I427F reversion rates in brain tissue were judged using one-sided Fisher’s exact test and a multinomial test with a Monte Carlo P value, performed using R statistical software (www.r-project.org).

Data availability.

The next-generation sequence reads generated in this project are available through the NIH Sequence Read Archive (SRA) under BioProject accession number PRJNA725261.