All work with infectious viruses and VRPs was performed in a biosafety level 4 facility at the Centers for Disease Control and Prevention (Atlanta, GA).

A cassette for expressing a codon-optimized open reading frame for LASV Josiah glycoprotein precursor (GPC; corresponding to HQ688672) was inserted into the Vero E6 cell line, using the Flp-In system (Invitrogen).

LASV Josiah was rescued as described elsewhere [11]. The rescue constructs were modified to enable rescue of nonspreading VRPs by replacing the GPC open reading frame with the ZsGreen (ZsG) open reading frame (Clontech). For some VRPs, mutations D389A and G392A were made to inactivate the nucleoprotein (NP) exonuclease domain [12, 13], or the LASV Z open reading frame was replaced with that of Mopeia virus (NC_006574). VRPs were rescued on BSR T7/5 cells (a generous gift from Karl-Klaus Conzelmann), using rescue constructs for the L and S segments and a helper construct for LASV GPC.

Antibodies used included an α-LASV NP (703079-Lassa Josiah HMAF; in-house reagent) for immunoblotting; an α-LASV NP (mouse monoclonal 52-129-18) for immunofluorescence analysis; an α-LASV GP1 (monoclonal antibody 52-74-7) for immunoblotting and immunofluorescence analysis; an α-LASV GP2 (monoclonal antibody 52-85-6) [14] for immunoblotting; an anti–influenza virus hemagglutinin tag, clone HA-7 (Sigma-Aldrich), for immunofluorescence analysis; secondary α-mouse immunoglobulin G (IgG) antibodies coupled to Alexa488, Alexa594, or Alexa633 fluorophores (Molecular Probes), for immunofluorescence analysis; and a horseradish peroxidase–conjugated β-actin antibody (GenSript A00730) for immunoblotting.

Commercial and custom-made assays against the LASV Josiah/VRP NP sequence, as well as guinea pig GAPDH messenger RNA (mRNA), interferon β (IFN-β) mRNA, 18S RNA, and CCL5/RANTES mRNA, were used as described in the Supplementary Materials. For in vitro experiments, the relative abundance of each analyte was calculated using a correction for PCR efficiency [15] and GAPDH as the internal control.

Experiments with strain 13/N guinea pigs from our in-house breeding colony were conducted under biosafety level 4 containment as described in the Supplementary Materials. A group size of 5 animals each was chosen so that a statistically significant survival difference (defined as a difference with a P value of < .05, by the Fisher exact test) between vaccinated and mock-vaccinated animals would be detected if complete protection was achieved after vaccination and if mortality in the mock-vaccinated group was 80%–100%.

Plasma was inactivated using 5 × 10⁴ Gy of gamma irradiation from a ⁶⁰Co source. For IgG ELISA, lysates from either Vero E6 cells or LASV Josiah–infected Vero E6 cells were used as the coating antigen. To detect immunoglobulin M (IgM) antibodies, a capture ELISA was used. ELISA measurements were repeated twice, and the average titer is reported for samples that yielded positive results on both repeats.

The statistical significance of differences in guinea pig survival was evaluated using the 2-sided Fisher exact test (GraphPad Prism 7.03 software).

To produce VRPs that resemble infectious virus but do not spread beyond the initially infected cells, we modified our LASV Josiah strain reverse-genetics system [11] by replacing the gene for the GPC, coded by the S segment, with the gene for the fluorescent protein ZsG (Figure 1A). VRPs could be rescued when the resulting plasmid pLasSΔGPC+ZsG and the native pLasL plasmid were transfected into BSR T7/5 cells together with an expression construct for LASV GPC. When passaged in parallel in Vero E6 cells, VRPs and authentic virus showed similar NP immunofluorescence signals at day 1 of infection, suggesting comparable replication efficiencies (Figure 1B). Spread of authentic virus in the monolayer was evident on day 2 after infection, with virtually all cells infected by day 7, and the supernatant readily infected fresh cell monolayers. In contrast, VRPs did not spread in the primary culture or pass into secondary cells, confirming the expected single-cycle phenotype.

While Lassa VRPs could be rescued in BSR T7/5 cells with the help of a GPC expression construct, low titers were obtained. Therefore, a cell line stably expressing GPC was cloned to enable large-scale VRP production. Vero E6 cells were chosen as the basis of this cell line because they exhibit contact inhibition and, most importantly, lack the gene cluster for type I IFN genes [16]. These characteristics make Vero E6 cells the standard choice for growing virus and VRP stocks free of the potent inducers of antiviral defenses.

The cell line Vero-LASV-GPCco (cloned as described in Methods) showed marked GP1 immunofluorescence staining (Figure 2A). To test whether Vero-LASV-GPCco cells can support VRP production, spread of the ZsG signal in the monolayer was observed by imaging the same areas repeatedly after VRP inoculation. VRPs inoculated at a low multiplicity spread to most cells in the Vero-LASV-GPCco culture within 3 days (Figure 2B). Interestingly, initial VRP infectivity was reduced in Vero-LASV-GPCco cells as compared to Vero E6 cells. Although we did not study the reason for this, the lower infectivity upon GPC expression may have been due to reduced expression of functional α-dystroglycan, a receptor for LASV, on the cell surface [17]. Nonetheless, 1 passage in Vero-LASV-GPCco cells yielded VRP titers approximately 3 logs above those seen after the initial rescue, showing that the stable cell line can be used for efficient VRP propagation (Figure 3A).

To characterize GP expression in more detail, parental Vero E6 cells and Vero-LASV-GPCco cells were inoculated with LASV, VRP, or mock inoculum. The immunoblot pattern of GP1, GP2, and higher-molecular-weight GPC expression in mock-infected Vero-LASV-GPCco cells closely resembled that of virus-infected Vero E6 cells (Figure 3B), indicating that GPC expressed by the stable cell line was processed the same way as during virus infection. As expected, VRP-inoculated Vero E6 cells contained NP but no detectable viral glycoproteins. In contrast, VRP-inoculated Vero-LASV-GPCco cells expressed NP and GP1/2 similarly to virus-infected Vero E6 cells, albeit with somewhat reduced GP1/GP2 levels. Taken together, the results show that the gene encoding GPC can be inserted into a producer cell genome to yield a system in which authentic viral GP processing enables propagation of nonspreading VRPs.

The primary target cells of LASV, macrophages and dendritic cells, fail to activate upon infection [18, 19]. LASV NP mediates inhibition of innate immunity recognition [20] via its C-terminal exonuclease domain [12, 13]. Additionally, arenavirus Z protein has been designated as an inhibitor of RIG-I signaling. While some researchers found that this Z protein function is a feature of LASV and other pathogenic arenaviruses regardless of origin [21], others suggested that the RIG-I inhibitory function is found in New World arenaviruses of South America but not in Old World arenaviruses such as LASV [22]. We reasoned that removing the ability of the VRPs to antagonize the innate immune response could improve vaccine immunogenicity.

To examine this idea, we rescued 4 different VRPs: wild type (WT)–WT, which has no changes in the L and S segments; WT-Exo(N), which has no change in the L segment and the S segment-encoded NP exonuclease made null by mutations D389A/G392A [12, 13]; MopZ-WT, which has no change in the S segment and replaces LASV Z with Mopeia virus Z, which was reported not to inhibit RIG-I signaling [21]; and MopZ-Exo(N), which has the changes in the S segment and Z protein described above. Immunofluorescence staining was used to study IFN regulatory factor 3 (IRF-3) relocalization in VRP-inoculated A549 cells. As seen in Figure 4A, IRF-3 remained in the cytoplasm of cells with WT-WT or MopZ-WT replication. In contrast, frequent nuclear accumulation of the transcription factor was seen in cells with WT-Exo(N) or MopZ-Exo(N) replication. Inactivating the VRPs with UV light prior to inoculation prevented the observed changes in IRF-3 relocalization, showing that VRP replication, not soluble factors in the inocula, trigger the response.

To confirm these results, guinea pig GPC-16 cells were inoculated with the VRPs in parallel with LASV Josiah, and activation of the innate immune response was monitored by IFN-β and CCL5/RANTES qRT-PCR analysis. As seen in Figure 4B, WT-WT–inoculated cells displayed low activation levels comparable to those of the virus control. NP exonuclease mutations enabled a more robust response, as previously reported using recombinant viruses [23]. However, replacing LASV Z with Mopeia virus Z did not enhance the cellular response to VRPs. Based on these results, WT-WT and WT-Exo(N) VRPs were chosen as the vaccine candidates to be tested in the guinea pig model of Lassa fever.

LASV Josiah causes a lethal infection in inbred strain 13 guinea pigs [24], and this model, recently characterized in detail elsewhere [25], has been used extensively to test LASV vaccine candidates [26–30] and therapeutics [31, 32]. To evaluate the protective efficacy and critically test the scalability of the candidates, we chose a vaccine dose that could be delivered without concentrating the VRPs. Groups of 5 or 4 strain 13 guinea pigs were mock vaccinated or vaccinated subcutaneously with 5 × 10⁵ focus-forming units of WT-WT VRPs or WT-Exo(N) VRPs. No clinical signs, fever, or weight loss were associated with the injections (Figure 5), indicating that the VRPs caused no disease in an animal model susceptible to Lassa fever.

After 28 days, prechallenge blood samples were collected, and the guinea pigs were challenged subcutaneously with 1 × 10⁴ focus-forming units of LASV Josiah. Elevated temperatures and body weight loss were seen from day 11 onward in mock-vaccinated animals, which succumbed to infection 20–23 days after infection (Figure 6). All animals vaccinated with WT-WT or WT-Exo(N) VRPs survived the infection without exhibiting fever, weight loss, or clinical signs during the 42-day observation period (P < .01 for the difference in survival between the recipients of each vaccine and the control group).

All tissue samples (ie, blood, liver, lung, spleen, kidney, adrenal glands, heart, ovary, epididymis, vagina, testes, eye, and brain) collected from animals in the mock-vaccinated group after euthanasia revealed LASV RNA (Figure 7A). In contrast, none of the vaccinated animals had detectable LASV RNA in any of the examined tissues at the end of the observation period.

Two of 5 euthanized animals in the mock-vaccinated group had detectable IgM antibodies against LASV at the time of euthanasia, as determined by ELISA (Figure 7B). None of the vaccinated animals exhibited anti-LASV IgM ELISA titers before challenge (ie, 28 days after vaccination) or after challenge (ie, 42 days after asymptomatic infection). No IgG ELISA titers were observed in the mock-vaccinated or WT-WT–vaccinated animals either before or after challenge, but all animals of the WT-Exo(N) group developed IgG ELISA titers against LASV by the end of the postchallenge observation period. A more sensitive immunofluorescence study of the serological responses detected IgG against LASV NP in 2 of 5 WT-WT–vaccinated animals before challenge, with no such prechallenge antibodies in the mock-vaccinated or WT-Exo(N) groups (Supplementary Figure 1). Using this assay, postchallenge seroconversion could be detected in all animals of the study.