Machupo Virus with Mutations in the Transmembrane Domain and Glycosylation Sites of the Glycoprotein Is Attenuated and Immunogenic in Animal Models of Bolivian Hemorrhagic Fever

ABSTRACT

Several highly pathogenic mammarenaviruses cause severe hemorrhagic and neurologic disease in humans for which vaccines and antivirals are limited or unavailable. New World (NW) mammarenavirus Machupo virus (MACV) infection causes Bolivian hemorrhagic fever in humans. We previously reported that the disruption of specific N-linked glycan sites on the glycoprotein (GPC) partially attenuates MACV in an interferon alpha/beta and gamma (IFN-α/β and -γ) receptor knockout (R^−/−) mouse model. However, some capability to induce neurological pathology still remained. The highly pathogenic Junin virus (JUNV) is another NW arenavirus closely related to MACV. An F427I substitution in the GPC transmembrane domain (TMD) rendered JUNV attenuated in a lethal mouse model after intracranial inoculation. In this study, we rationally designed and rescued a MACV containing mutations at two glycosylation sites and the corresponding F438I substitution in the GPC TMD. The MACV mutant is fully attenuated in IFN-α/β and -γ R^−/− mice and outbred guinea pigs. Furthermore, inoculation with this mutant MACV completely protected guinea pigs from wild-type MACV lethal challenge. Last, we found the GPC TMD F438I substitution greatly impaired MACV growth in neuronal cell lines of mouse and human origins. Our results highlight the critical roles of the glycans and the TMD on the GPC in arenavirus virulence, which provide insight into the rational design of potential vaccine candidates for highly pathogenic arenaviruses.

IMPORTANCE For arenaviruses, the only vaccine available is the live attenuated Candid#1 vaccine, a JUNV vaccine approved in Argentina. We and others have found that the glycans on GPC and the F427 residue in the GPC TMD are important for virulence of JUNV. Nevertheless, mutating either of them is not sufficient for full and stable attenuation of JUNV. Using reverse genetics, we disrupted specific glycosylation sites on MACV GPC and also introduced the corresponding F438I substitution in the GPC TMD. This MACV mutant is fully attenuated in two animal models and protects animals from lethal infection. Thus, our studies highlight the feasibility of rational attenuation of highly pathogenic arenaviruses for vaccine development. Another important finding from this study is that the F438I substitution in GPC TMD could substantially affect MACV replication in neurons. Future studies are warranted to elucidate the underlying mechanism and the implication of this mutation in arenavirus neural tropism.

INTRODUCTION

Highly pathogenic mammarenaviruses can cause severe hemorrhagic fevers in humans. The Old World mammarenavirus, Lassa virus, is responsible for an estimated 300,000 infections and 5,000 deaths per year in West Africa (https://www.cdc.gov/vhf/lassa/). New World mammarenaviruses, including Junin virus (JUNV), Machupo virus (MACV), Guanarito virus (GTOV), Chapare virus (CHAPV), and Sabia virus (SABV), periodically emerge in parts of South America to cause localized epidemics with high case fatality rates (1, 2). The two most common New World arenaviruses, JUNV and MACV, are the causative agents of Argentinian hemorrhagic fever (AHF) and Bolivian hemorrhagic fever (BHF), respectively. AHF and BHF share a similar clinical course of illness, with patients progressing from an initial nonspecific prodromal phase into a biphasic hemorrhagic and neurologic disease with an overall case fatality rate ranging from 15 to 35% (3–5). Hemorrhagic manifestations include epistaxis, petechiae, hypotension, hematemesis, and mucosal hemorrhaging. Neurologic manifestations include tremors, delirium, muscle spasms, and coma (6, 7). Overall, there are no FDA-approved treatments or vaccines for any of these infections. Thus, there is an urgent need for new countermeasures for highly pathogenic arenaviruses.

The arenavirus genome consists of 4 viral genes, encoding 4 viral proteins, in a bisegmented, ambisense RNA genome (8). The large (L) segment encodes the RNA-dependent RNA polymerase (L) and the small RING finger matrix protein, Z. The small (S) segment encodes the nucleoprotein (NP) and the glycoprotein complex (GPC). GPC is cleaved by cellular proteases into stable signal peptide (SSP), GP1, and GP2, which all are subunits of GPC (9–11). Arenaviruses enter the cell through receptor-mediated endocytosis, binding to their cellular receptor via GP1, which is also the primary target of host-neutralizing antibodies (12, 13). Old World arenaviruses primarily use α-dystroglycan as a receptor (14), while pathogenic New World arenaviruses use transferrin receptor 1 (TfR1) (15, 16). Upon entry into the endosome, a pH-dependent membrane fusion process mediated by GP2 allows for the release of the viral genomic RNAs into the cytoplasm (17, 18), where early translation of NP and L protein begins. Later in the replication process, GPC and Z proteins are produced, and GPC is cleaved in the endoplasmic reticulum (ER) (19–22). Glycan cores are first posttranslationally added to GPC in the ER and are further trimmed and modified in the Golgi. This is followed by trafficking of the mature GPC to the cell surface, where virus assembly and release from the cell occurs.

Only one arenavirus vaccine, the live attenuated JUNV strain, Candid#1 (Cd1), has been approved for use only in Argentina (23, 24). This vaccine was created through the serial passaging of a lethal isolate of JUNV in animals and cultured cells, leading to the acquisition of a series of attenuating mutations. The strain was first passaged through guinea pigs twice, followed by 44 passages in suckling mouse brains (24, 25). During this passaging, the strain acquired a T168A substitution in GP1 that led to the loss of a glycan (21, 22). After the 44 passages in mouse brain, the resultant strain, XJ44, was attenuated in guinea pigs but retained virulence in suckling mice after intracranial inoculation. XJ44 was subsequently passaged in rhesus macaque lung fibroblast cells in vitro and acquired an additional substitution in the transmembrane domain (TMD) of GP2 (25). This Phe-to-Ile mutation (F427I) has been found to be critical for attenuation in the mouse model (25). Introduction of this single substitution alone into the wild-type JUNV sequence significantly attenuates the virus in the suckling mouse model of JUNV lethal infection, as 90% of mice survived after intracranial inoculation. However, this TMD substitution does not fully attenuate the virus, indicating that multiple mutations are necessary for complete attenuation (25).

Our previous work has demonstrated that a MACV chimera expressing the whole GPC of Cd1 is fully attenuated (26). Likewise, a MACV chimera containing the Cd1 GPC ectodomain is also partially attenuated, though reversions at two N-linked glycans (N83 and N166) were frequently observed (27, 28). Removal of these two N-linked glycans (MACV GPC_ΔN83/ΔN166) results in a mutant that is partially attenuated in interferon alpha/beta and gamma (IFN-α/β and -γ) receptor knockout (R^−/−) mice (28). While 100% of these mice survived infection, they all developed nonlethal neurological signs of disease. Meanwhile, a MACV mutant containing the analogous F438I TMD substitution was likewise partially attenuated in IFN-α/β and -γ R^−/− mice (29). However, 28% of mice infected with the MACV GPC_F438I mutants succumbed to disease, and all viruses isolated from the brains of succumbed mice reverted to the wild-type sequence.

In this study, we combined the TMD substitution along with the mutations disrupting specific glycosylation sites in MACV GPC. Herein, we report that the MACV mutant with glycan deficiencies at residues N83 and N166 and the F438I TMD substitution on GPC is fully attenuated in two different animal models. Furthermore, inoculation with this attenuated mutant provides protection from subsequent lethal challenge with wild-type MACV. This protection can likely be attributed to a strong MACV-specific humoral response. Finally, we provide evidence that the F438I substitution leads to a replication defect in neurons, likely contributing to the attenuation phenotype observed.

RESULTS

Introducing mutations to specific GPC glycan sites and the TMD F438I mutation leads to complete attenuation of MACV.

Our previous work demonstrates that disruption of glycosylation sites at GPC N83 and N166 partially attenuates MACV in IFN-α/β and -γ R^−/− mice (28). Though all mice survive challenge with this mutant, they all develop neurological signs of disease. On the other hand, while the F438I substitution is known to attenuate MACV, it is genetically unstable in vivo (29). We therefore combined the glycan removal mutations with the F438I substitution to test the hypothesis that the combination of these mutations would lead to a complete attenuation. Using our established MACV reverse genetics system (30), we rescued MACV mutants containing the F438I substitution in addition to glycan deficiency at either one or two sites (MACV GPC_ΔN166/F438I and MACV GPC_{ΔN83/ΔN166/F438I}) (Fig. 1A). For each of the rescued viruses, GPC protein expression and cleavage were confirmed via Western blotting (Fig. 1B). All of these viruses displayed similar growth kinetics in Vero cells (Fig. 1C). Furthermore, no reversions or unintended mutations were found in the GPC gene after 5 serial passages in Vero cells.

FIG 1.

Creation of MACV with glycan removals and transmembrane substitution. (A) Schematic representation of the genome of MACVs. Within GPC, SSP (stable signal peptide), ED (ectodomain), TMD (transmembrane domain), and CT (cytoplasmic tail) are shown. Ψ, N-glycosylation site. (B) Representative Western blotting of GPC in infected Vero cells at an MOI of 1. The number of putative N-glycosylation sites is shown in parentheses. (C) Viral replication of MACV GPC_F438I, MACV GPC_ΔN166/F438I, and MACV GPC_{ΔN83/ΔN166/F438I} was characterized in Vero cells (MOI, 0.01). The dashed line indicates the detection limit. Data shown are averages of triplicate wells with error bars indicating the SD.

We next studied virus infection in the IFN-α/β and -γ R^−/− mouse model. Groups of IFN-α/β and -γ R^−/− mice were inoculated with MACV GPC_F438I, MACV GPC_ΔN166/F438I, MACV GPC_{ΔN83/ΔN166/F438I}, or wild-type MACV Carvallo strain (originally isolated from a human patient’s spleen after a fatal case of BHF) and monitored for signs of disease for up to 49 days. Compared to the F438I substitution in the GPC TMD alone, removal of one or two specific glycans in addition to the F438I substitution further attenuates the virus, with the removal of two glycans resulting in the greatest attenuation. All mice infected with MACV GPC_{ΔN83/ΔN166/F438I} survived infection (Fig. 2A) without any observable clinical illness (Fig. 2B). Mice infected with MACV GPC_{ΔN83/ΔN166/F438I} maintained their body weights throughout the course of the study in contrast to animals infected with wild-type MACV (Fig. 2C). We also assessed virus load and viral RNA in serum, lung, liver, and brain samples of these animals at time of death/euthanasia or at 49 days postinfection (dpi). No virus was detected in any animals infected by MACV GPC_{ΔN83/ΔN166/F438I}, while virus was detected in at least one animal from every other group (Fig. 2D). Likewise, viral RNA was not detected in the brains of MACV GPC_{ΔN83/ΔN166/F438I}-infected animals (Fig. 2E), further demonstrating that MACV GPC_{ΔN83/ΔN166/F438I} is highly attenuated in IFN-α/β and -γ R^−/− mice.

FIG 2.

MACV with glycan removals and transmembrane substitution is attenuated in IFN-α/β and -γ R^−/− mice. (A) Survival rate of IFN-α/β and -γ R^−/−mice after intraperitoneal injection of 10⁴ PFU of the indicated strains and mutants (Carvallo, n = 8; MACV GPC_F438I, n = 12; MACV GPC_ΔN166/F438I, n = 11; MACV GPC_{ΔN83/ΔN166/F438I}, n = 11; PBS, n = 6). Statistically significant differences are indicated by asterisks (**, P < 0.01 by log-rank test). (B) Morbidity rate of the IFN-α/β and -γ R^−/− mice after infection. (C) Body weight changes were monitored on the indicated days. Error bars indicate the SEM. The data shown are pooled from two independent experiments. (D) Virus titers in brain, lung, liver, and serum. The minimum detection limit is shown as a dashed line. Data under detection limit are plotted as 1 PFU/g or mL. The solid bar represents the mean of the titers. (E) Viral RNA at time of euthanasia. Shown are the total number of animals per group, the number of animals per group that succumbed to infection and were humanely euthanized, the number of animals with detectable viral RNA (GPC) at time of death, and the number of animals with I438F revertant viral RNA. (F) Neutralization activity of mouse serum against wild-type MACV (Carvallo). (G) PRNT₅₀ against Carvallo using serum samples from mice infected with MACV GPC_F438I, MACV GPC_ΔN166/F438I, or MACV GPC_{ΔN83/ΔN166/F438I} at 49 dpi. (H) PRNT₅₀ against MACV Carvallo, MACV GPC_F438I, MACV GPC_ΔN166/F438I, and MACV GPC_{ΔN83/ΔN166/F438I} was determined using the serum samples (n = 7) collected on day 49 from MACV GPC_F438I-infected animals. Statistically significant differences are indicated by asterisks (*, P < 0.05; **, P < 0.01; ns, not significant by Welch's t test).

To determine if MACV GPC_{ΔN83/ΔN166/F438I} induces a neutralizing antibody response, we performed a 50% plaque reduction neutralization test (PRNT₅₀) assay for serum samples (Fig. 2F and Table 1). As expected, mice inoculated with MACV GPC_F438I, MACV GPC_ΔN166/F438I, and MACV GPC_{ΔN83/ΔN166/F438I} developed higher neutralizing antibody titers against wild-type virus than animals inoculated with wild-type MACV (Carvallo) (Fig. 2F and Table 1). It should be noted, however, that the survival time of wild-type MACV-infected animals was markedly shorter (average of 26.6 days of infection) than others, and therefore, a shorter survival period of wild-type MACV-infected animals may preclude maximal immune response. Furthermore, we analyzed sera collected at the same time after infection (49 dpi) for neutralizing antibody titers against wild-type MACV (Carvallo) (Table 1). Notably, the neutralization antibody titer to wild-type MACV (Carvallo) in the MACV GPC_{ΔN83/ΔN166/F438I}-infected animals was significantly higher than that in the MACV GPC_F438I-infected animals (P < 0.05, by Welch’s t test) (Table 1 and Fig. 2G). This suggests that loss of specific N-linked glycans might enhance GPC immunogenicity. To evaluate the effect of the N-linked glycans at N83 and N166 on viral sensitivity to antibody neutralization, we compared the neutralization sensitivity of different recombinant MACVs (rMACVs) to the sera harvested from MACV GPC_F438I-infected animals at 49 dpi. MACV Carvallo was the least susceptible to neutralization (73.1), followed by MACV GPC_F438I (146.3), followed by MACV GPC_ΔN166/F438I (217.4), and finally leaving MACV GPC_{ΔN83/ΔN166/F438I} as the most susceptible to neutralization (≥646.0) (Table 1 and Fig. 2H). MACV GPC_{ΔN83/ΔN166/F438I}, which has two glycan defects, exhibited significantly higher susceptibility than other viruses. These results are largely in agreement with previous findings that the number of glycans on GPC is inversely correlated with the susceptibility of arenaviruses to neutralizing antibodies, likely due to glycan shielding of neutralizing epitopes (31, 32).

TABLE 1.

PRNT₅₀ geometric mean titers in mice infected with MACV mutants

Strain or mutant	PRNT50 geometric mean titer for serum group:
	Carvallo	GPCF438I	GPCΔN166/F438I	GPCΔN83/ΔN166/F438I	PBS
Carvallo	23.8a (N = 6, 26.6 dpi)b	80.1 (12, 39.9)	112.7 (11, 46.8)	128.6 (10, 49)	<30 (6, 49)
Carvallo (with serum collected at 49 dpi)	NAc	73.1 (7, 49)	104.5 (10, 49)	128.6 (10, 49)	<30 (6, 49)
GPCF438I	NA	146.3 (7, 49)	NA	NA	NA
GPCΔN166/F438I	NA	217.4 (7, 49)	≥727.5 (10, 49)	NA	NA
GPCΔN83/ΔN166/F438I	NA	≥646.0d (7, 49)	NA	≥846.3 (10, 49)	NA

^aA titer of <30 was calculated as 15 for geometric mean if samples contain <30. The titer was shown as <30 when all the samples were <30.

^bNumbers in parentheses indicate the numbers of samples tested and the average sample collection days postinfection.

^cNA, not available.

^dThe serum dilution factor for PRNT is from 1:30 to 1:960. If a group includes samples that neutralized the virus even at the 1:960 dilution, they are listed as “≥Ave” because they may be higher than the average shown.

MACV GPC_{ΔN83/ΔN166/F438I} is attenuated in outbred guinea pigs.

MACV GPC_{ΔN83/ΔN166/F438I} was fully attenuated even in immunocompromised mice, which are normally highly susceptible to MACV. Therefore, we sought to test whether this mutant is also attenuated in an immunocompetent guinea pig model and could provide protection against lethal challenge with wild-type MACV. The Carvallo strain of MACV used in our previous mouse experiments is reportedly not fully lethal in the outbred Hartley guinea pig model (33). Consistent with this report, we also found that the lethality of Carvallo strain infection of outbred guinea pigs is highly variable, ranging from 0 to 80% across different studies (data not shown). A recent study reported that the Chicava strain of MACV exhibited 100% lethality in Hartley guinea pigs (34). Therefore, we used the Chicava strain of MACV in our guinea pig studies. Groups of guinea pigs were challenged intraperitoneally with 10,000 PFU of either MACV GPC_{ΔN83/ΔN166/F438I} or MACV Chicava strain and monitored for disease signs. Body weights and temperatures were collected daily for the first 30 days postinfection. Consistent with our mouse experiments, 100% of guinea pigs infected with MACV GPC_{ΔN83/ΔN166/F438I} survived challenge, while 100% of animals infected with the Chicava strain of MACV succumbed to challenge by day 22 (Fig. 3A). Guinea pigs infected with the Chicava strain of MACV developed mild fevers as early as 5 dpi and began to lose weight at 15 dpi, reaching humane euthanasia criteria by days 19 to 22. Some animals also developed additional clinical signs, including vomiting and hind limb paralysis, with the latter a clear manifestation of neurological pathology. In contrast, none of the animals infected with MACV GPC_{ΔN83/ΔN166/F438I} developed any clinical signs (Fig. 3B). These animals steadily gained weight over the course of the study (Fig. 3C), and none of the animals developed an elevated temperature (Fig. 3D).

FIG 3.

MACV with glycan removals and transmembrane substitution is attenuated in Hartley guinea pigs. (A) Survival rate of animals injected intraperitoneally with 10⁴ PFU of either MACV GPC_{ΔN83/ΔN166/F438I} (n = 10) or WT MACV Chicava (n = 4). Data are representative of two independent experiments. (B) Morbidity rate of animals after infection. Data are representative of two independent experiments. (C) Body weight changes were monitored daily for the first 30 days postinfection (n = 4 per group). Error bars represent the SEM. (D) Temperature was monitored via BMDS transponder daily for 30 days postinfection (MACV GPC_{ΔN83/ΔN166/F438I}, n = 4; wild-type (WT) MACV Chicava, n = 8). Error bars represent the SEM. (E) Virus titers in brain, liver, lung, spleen, and kidney. The minimum detection limit is shown as a dashed line. Data under detection limit are plotted as 1 PFU/g or mL. The solid bar represents the mean of the titers.

Though no clinical illness was observed for any animal infected with MACV GPC_{ΔN83/ΔN166/F438I}, we wanted to determine whether live virus and/or viral RNA were present in any organs. At 30 dpi, 3 animals infected with MACV GPC_{ΔN83/ΔN166/F438I} were humanely euthanized, and blood and organs were collected. We could not detect live virus in the brain, liver, spleen, lung, or kidney of any animal infected with MACV GPC_{ΔN83/ΔN166/F438I} (Fig. 3E). Furthermore, no viral RNA could be detected in any organ (data not shown). In contrast, animals infected with the Chicava strain of MACV all had detectable virus in at least one of the five organs tested at the time of euthanasia. While some of these animals were able to clear the virus in some organs, all animals infected with the Chicava strain of MACV had detectable live virus in the brain. In summary, these results demonstrate that MACV GPC_{ΔN83/ΔN166/F438I} (Carvallo backbone) does not cause detectable disease in an immunocompetent animal model.

Immunization with attenuated MACV GPC_{ΔN83/ΔN166/F438I} protects guinea pigs from lethal challenge with heterologous MACV strain.

Since MACV GPC_{ΔN83/ΔN166/F438I} is fully attenuated in both immunocompromised mice and immunocompetent guinea pigs, we next investigated whether this mutant MACV could protect animals from lethal challenge with the heterologous MACV Chicava strain. In the MACV GPC_{ΔN83/ΔN166/F438I} group, guinea pigs were first immunized with MACV GPC_{ΔN83/ΔN166/F438I} and then challenged with 10,000 PFU of the MACV Chicava strain 60 days later. As a control, a new group of age-matched naïve guinea pigs was challenged with 10,000 PFU of the Chicava strain of MACV (the control group). Animals were monitored for clinical signs of disease as well as weight and temperature changes for 30 days postchallenge. Again, 100% of naïve control animals succumbed to MACV Chicava strain challenge between days 19 to 22, with similar disease development to what was observed and reported previously. All animals immunized with MACV GPC_{ΔN83/ΔN166/F438I} survived challenge (Fig. 4A) and did not develop any discernible clinical signs of disease (Fig. 4B). These animals maintained their weight throughout the course of the study (Fig. 4C) and never developed fevers (Fig. 4D), in contrast to the unimmunized control animals.

FIG 4.

Guinea pigs challenged with wild-type MACV after inoculation with MACV GPC_{ΔN83/ΔN166/F438I}. (A) Survival rate of animals inoculated with MACV GPC_{ΔN83/ΔN166/F438I} (n = 7) or naïve animals (n = 6) challenged 60 days postinoculation with an intraperitoneal injection of 10⁴ PFU of wild-type MACV (Chicava). Data are representative of two independent experiments. (B) Morbidity rate of animals after infection. Data are representative of two independent experiments. (C) Body weight changes were monitored daily for the first 30 days postinfection. Error bars represent the SEM. Data are pooled from two independent experiments. (D) Temperature was monitored via BMDS transponder daily for 30 days postinfection. Error bars represent the SEM. Data are pooled from two independent experiments. (E) Virus titers in brain, liver, lung, spleen, and kidney. The minimum detection limit is shown as a dashed line. Data under detection limit is plotted as 1 PFU/g or mL. The solid bar represents the mean of the titers.

Thirty days after the MACV Chicava challenge, animals inoculated with MACV GPC_{ΔN83/ΔN166/F438I} were humanely euthanized for the collection of blood and tissue samples. No live virus was detected in the brain, liver, spleen, lung, or kidney of any of the animals in the MACV GPC_{ΔN83/ΔN166/F438I} group (Fig. 4E). Consistently, viral RNA was undetectable (data not shown). In contrast, virus could be detected in the brains of all animals in the control group. High viral titers were observed in other organs in the control group as well.

We performed a histopathological analysis of the liver, spleen, and brain of MACV-challenged Hartley guinea pigs. Portal and lobular inflammatory lymphohistiocytic lesions were observed in the livers of lethally challenged animals without immunization (Fig. 5, left panels). Mildly decreased white pulp cellularity was observed in the spleens of succumbed control animals but not in those inoculated with MACV GPC_{ΔN83/ΔN166/F438I}. Perivascular cellular infiltrates were observed in the brains of the control animals, but not in those inoculated with MACV GPC_{ΔN83/ΔN166/F438I} (Fig. 5) (34). Overall, these results demonstrate that MACV GPC_{ΔN83/ΔN166/F438I} administration conferred protection from lethal MACV challenge.

FIG 5.

Histopathology from MACV-challenged Hartley guinea pigs. Major target organs of MACV were collected at time of euthanasia, 20 to 22 days after inoculation with Chicava without immunization (left panels), 90 days after inoculation with MACV GPC_{N83/N166/F438I} and 30 days following wild-type Chicava challenge (middle panels), or 30 days after inoculation with MACV GPC_{N83/N166/F438I} followed by no challenge (right panels). In the liver, portal (white arrowhead) and lobular (black arrowhead) inflammatory lymphohistiocytic lesions were observed in lethally challenged animals (inset shows lobular focus of inflammation). In the spleen, mildly decreased white pulp cellularity (white asterisk) was observed in lethally challenged animals. In the brain, perivascular infiltrates were observed in lethally challenged animals (white arrow and inset). Hematoxylin and eosin stain; magnification, ×10.

MACV-specific humoral response is generated after immunization with MACV GPC_{ΔN83/ΔN166/F438I}.

The humoral immune response is known to be critical for combating New World arenavirus infections (35, 36), so we next wanted to determine whether MACV GPC_{ΔN83/ΔN166/F438I} elicited a humoral response that mediates protection. A total of 7 animals were inoculated with MACV GPC_{ΔN83/ΔN166/F438I} and, 60 days later, challenged with heterologous MACV Chicava strain. As controls, 3 naïve animals were challenged with Chicava strain of MACV. Blood was taken from 3 animals at 30 days and from another 4 animals at 60 days postinoculation with MACV GPC_{ΔN83/ΔN166/F438I}. Blood was also taken from 3 naïve control animals. We first performed PRNT assays to assess the serum harvested before challenge for neutralizing antibody titers against wild-type MACV (both Carvallo and Chicava strain) as well as the homologous MACV GPC_{ΔN83/ΔN166/F438I} mutant. Six of the seven animals inoculated with MACV GPC_{ΔN83/ΔN166/F438I} developed a strong neutralizing antibody response to the homologous, glycan-deficient MACV, which increased over time. However, only one of these seven animals developed detectable neutralizing antibodies against either the homologous or heterologous wild-type MACV strain before challenge (Fig. 6A). After challenge with the lethal Chicava strain, some, but not all, animals did develop detectable neutralizing antibodies to wild-type Carvallo virus (Fig. 6C). However, the neutralization antibody levels to the Chicava strain were very low or were not detectable in these animals. The neutralizing antibody titers against MACV GPC_{ΔN83/ΔN166/F438I} were consistently higher after Chicava challenge (Fig. 6C). This is consistent with previous data demonstrating that MACV mutants lacking specific glycans on GPC are more susceptible to antibody neutralization (28). We further performed enzyme-linked immunosorbent assays (ELISAs) to measure the level of MACV GP1- and NP-specific antibodies with purified MACV (Carvallo) GP1 and NP proteins. All but one animal had detectable MACV NP or GP1 binding antibodies, though titers were highly variable among individual animals (Fig. 6B). Titers increased after challenge with the Chicava strain (Fig. 6D). Overall, these results demonstrate that most of the animals inoculated with the glycan-deficient MACV GPC_{ΔN83/ΔN166/F438I} developed a measurable humoral response; however, the neutralizing antibody response to MACV Chicava strain was below the detection limit in the majority of animals.

FIG 6.

Humoral response following inoculation with attenuated MACV. (A) PRNT₅₀ titers of guinea pig serum collected at either 30 days or 60 days postinoculation with MACV GPC_{ΔN83/ΔN166/F438I}. Serum samples collected at 30 or 60 days postinoculation were collected from two different groups of animals. Neutralizing activity against Carvallo, Chicava, and GPC_{ΔN83/ΔN166/F438I} mutants of MACV was assessed. The solid line represents the geometric mean, while the dashed lines indicate the upper and lower limits of detection. (B) MACV-specific antibodies against either GP1 or NP measured via ELISA after 30 or 60 days postinoculation with MACV GPC_{ΔN83/ΔN166/F438I}. An ELISA titer was considered positive when the optical density at 450 nm (OD₄₅₀) value was higher than at least two standard deviations of the mean of naïve animal serum run concurrently. (C) PRNT₅₀ titers of guinea pig serum collected at either time of death or at 30 days postchallenge with wild-type MACV. (D) MACV-specific antibodies measured via ELISA at time of death or at 30 days postchallenge with wild-type MACV.

MACV containing the F438I substitution in the GPC TMD replicates poorly in neurons.

It is clear that the F438I substitution in MACV GPC abrogates neurological pathology in our IFN-α/β and -γ R^−/− mouse model. However, the mechanism by which the F438I substitution in the GPC TMD attenuates MACV in vivo remains unknown. To address this question, we compared the replication of wild-type MACV (Carvallo), MACV GPC_F438I, MACV GPC_ΔN166/F438I, and MACV GPC_{ΔN83/ΔN166/F438I} in mouse astrocytoma (C8D1A) and neuroblastoma (Neuro-2A) cell lines at a multiplicity of infection (MOI) of 0.01. While we observed no significant replication differences for these viruses in Vero cells (Fig. 1C) or mouse astrocytes (C8D1A) (Fig. 7A), all mutants containing the F438I substitution replicated to significantly lower titers than wild-type MACV in mouse neurons (Neuro-2A) (Fig. 7A). MACV GPC_F438I demonstrated the most severe defect in viral replication, while removal of glycans appeared to partially restore replication. We further confirmed this observation in human astrocytoma (132N1) and neuroblastoma (IMR-32) cell lines. Again, we observed no replication differences in 132N1 astrocytes, but we detected significantly impaired replication of all mutants in human IMR-32 neurons (Fig. 7B). These data demonstrate, for the first time, that the F438I substitution affects MACV infection in cultured neurons, providing one potential mechanism for the attenuation in vivo.

FIG 7.

Replication of MACV containing the F438I transmembrane substitution in neurological cell lines. (A) Viral replication of MACV GPC_F438I, MACV GPC_ΔN166/F438I, and MACV GPC_{ΔN83/ΔN166/F438I} was characterized in C8D1A and Neuro-2A cells (MOI, 0.01). The dashed line indicates the detection limit. Data shown are averages of triplicate wells with error bars indicating the SD. (B) Viral replication of MACV GPC_F438I, MACV GPC_ΔN166/F438I, and MACV GPC_{ΔN83/ΔN166/F438I} was characterized in human-derived cell lines 1321N1 astrocytes and IMR-32 neurons (MOI, 0.01). The dashed line indicates the detection limit. Data shown are averages of triplicate wells, with error bars indicating the SD.

DISCUSSION

In this study, we characterized a MACV mutant containing the TMD F438I substitution and mutations at glycosylation sites on GP1. Our previous research revealed that neither the F438I substitution nor the glycosylation site mutations alone are sufficient for full attenuation in IFN-α/β and -γ R^−/− mice (28, 29). This work demonstrates that the combination of the F438I substitution in the TMD with the glycan deficiency at N83 and N166 of GPC led to full attenuation in our IFN-α/β and -γ R^−/− mouse model. Interestingly, both glycans must be absent for complete and stable attenuation. Lack of the glycan at N166, even along with the F438I mutation, is not sufficient for full attenuation in mice: 1 of 11 mice developed disease and ultimately succumbed to infection. Sequencing analysis identified the I438F reversion in the GPC TMD of the MACV isolated from the mouse that succumbed. We also observed neurological manifestations such as imbalance in this animal. The emerged I438F revertant might be the cause of the neurological disease and the death of the animal, but we cannot rule out contributions from other unrecognized factors without future additional studies. As the TMD F438I substitution was acquired during repeated passaging of JUNV, it is not surprising that it is not stable and prone to reversion in other arenaviruses. This reversion occurred not only in vivo in MACV but also in vitro in LASV (25, 29). Nevertheless, we did not observe any reversion of MACV GPC_{ΔN83/ΔN166/F438I} in vitro, either in Vero or Neuro-2A cells. We could not isolate live virus from any animal infected with MACV GPC_{ΔN83/ΔN166/F438I}, so it is unclear whether reversion occurred in vivo. Further studies attempting to isolate live virus from animals earlier in the course of infection will be needed to assess the genetic stability of MACV GPC_{ΔN83/ΔN166/F438I} in vivo.

We did not conclusively determine that neutralizing antibodies play a significant role in mediating either the attenuation of MACV GPC_{ΔN83/ΔN166/F438I} or the protection from lethal challenge observed during this study. We did observe that guinea pigs inoculated with MACV GPC_{ΔN83/ΔN166/F438I} developed high titers of neutralizing antibodies against MACV GPC_{ΔN83/ΔN166/F438I}. However, interestingly, none of the animals developed detectable neutralizing antibodies against either the wild-type Carvallo or Chicava strain of MACV by 30 dpi, and only one animal had developed low titers of neutralizing antibodies against the wild-type strains by 60 dpi. Differences in GPC structure between wild-type MACV and MACV GPC_{ΔN83/ΔN166/F438I} may account for this observation. Previous studies have determined that neutralizing antibodies against JUNV do not neutralize MACV due to a loop 10 feature that blocks the receptor binding domain (RBD) (37, 38). Glycans likely play a similar role in blocking the RBD from neutralization, as we observe that MACV GPC_{ΔN83/ΔN166/F438I} is more easily neutralized by antibodies due to the loss of glycans and that the wild-type viruses are less susceptible to neutralization, likely due to glycan shielding (31, 32, 39). Nevertheless, despite the lack of detectable neutralizing antibodies against MACV Chicava strain before and after challenge in the majority of animals, all MACV GPC_{ΔN83/ΔN166/F438I}-inoculated guinea pigs survived without disease signs. While many of these animals (5 out of 7) did develop a neutralizing antibody response against wild-type Carvallo virus 30 days after challenge with Chicava virus, 2 animals did not develop any detectable neutralizing antibodies throughout the course of the study, likely indicating that protection was not mediated by neutralizing antibodies alone in these animals.

All except one animal developed high MACV-specific antibodies, suggesting that antibody-dependent cellular cytotoxicity (ADCC) and other nonneutralizing antibody functions contribute to protection. Nonneutralizing antibodies against the GP of ebolavirus and the hemagglutinin (HA) of influenza have been demonstrated to offer protection (40–42). The role of nonneutralizing antibodies during arenavirus infection has also been noted as significant in several studies. Nonneutralizing binding antibodies directed against LCMV have been demonstrated to accelerate viral clearance from infected mice (43, 44). For LASV, a vaccine candidate consisting of vaccinia virus containing LASV, NP RNA provided some protection despite the absence of GPC (45). In addition, an inactivated rabies vaccine vector expressing LASV GPC failed to elicit a neutralizing antibody response in vaccinated mice, yet it did elicit high titers of nonneutralizing antibodies capable of stimulating ADCC (46). Another study demonstrated that, while full-length IgG antibodies that neutralize JUNV in vitro protect guinea pigs from lethal challenge with JUNV, F(ab')2 fragment antibodies that still successfully neutralize JUNV in vitro do not offer protection for guinea pigs infected with JUNV (47). This indicates that elimination of infected cells via Fc-mediated effector functions is critical. For MACV-specific antibodies, ADCC activity has been observed (13), but further studies will be needed to elucidate the role of nonneutralizing antibodies during New World arenavirus infection.

Cellular immunity may also play a role, as one of our animals had no detectable MACV-specific antibodies of any type before lethal challenge with the Chicava strain, but nevertheless survived. This animal later developed MACV-specific antibodies after the challenge (Fig. 6D), confirming that the animal was infected and yet survived challenge. Because these animals were outbred, variability in their response to MACV infection is expected. Only one-third of human BHF patients develop severe disease (4), so the impact of other host factors cannot be ruled out. Nevertheless, further studies examining the potential contributions of cellular immunity during MACV infection would be beneficial.

Another important observation from our study was the important contribution of the F438I substitution to the attenuation of the virus in neurons. In our previous study, 100% of MACV GPC_ΔN83/ΔN166-infected IFN-α/β and -γ R^−/− mice developed neurological signs of disease such as imbalance. In this study, addition of the F438I substitution produced a mutant incapable of causing observable neurological signs of disease in both IFN-α/β and -γ R^−/− mice and Hartley guinea pigs. For both of our animal models, 100% of naïve animals infected with wild-type MACV developed detectable viral titers in the brain, even late in infection after the virus had been cleared from other organs. Meanwhile, no live virus or viral RNA could be detected in the brains of any animal inoculated with MACV GPC_{ΔN83/ΔN166/F438I}. We cannot confirm in our animal models whether MACV GPC_{ΔN83/ΔN166/F438I} can invade the nervous system but is incapable of replicating in the brain or whether it is incapable of spreading to the nervous system altogether. It is also possible that poor replication of MACV GPC_{ΔN83/ΔN166/F438I} allows for earlier immune system clearance and prevents the virus from establishing an infection in nervous tissues, independent of its ability to invade or replicate in the nervous system. Future studies using an intracranial mode of inoculation would better address these issues. However, we did confirm that MACV GPC_{ΔN83/ΔN166/F438I} replicates poorly relative to wild-type virus in both mouse and human neurons in vitro. In our study, the combination of the glycan removals and the F438I substitution led to an overall significant loss of replication in neurons, which may partially explain the attenuation observed. Even if the attenuation observed in our animal study was not due to any loss of replication or invasion of the nervous system, this is the first known report of attenuated replication in vitro of an arenavirus mutant containing the F438I transmembrane substitution. It is possible that the replication defect we observed in neurons may also be present in other cell types. Future studies using primary cell lines and other cell types such as dendritic cells or macrophages may address this possibility.

The mechanism behind the replication defect of MACV GPC_{ΔN83/ΔN166/F438I} and MACV GPC_F438I in neurons remains unclear. One previous report indicated that the F427I substitution in JUNV promotes cell-to-cell fusion at a neutral pH, as opposed to the acidic pH normally required for fusion, which led to reduced JUNV minigenome replication (48). The authors speculated that the F427I substitution alters the GP in such a way that it remains in a postfusion conformation. However, other studies have reported that the pH needed for fusion remains unchanged in JUNV mutants containing F427I alone (49). Nevertheless, these reports also demonstrate that a substitution at F427 can restore the phenotype of SSP mutants that lower the required pH for fusion (17, 49, 50). The F427 residue has been demonstrated to genetically interact with certain charged residues in the SSP (49, 51). In addition to a potential defect in entry/fusion, a potential conformational mismatch with SSP due to the F427I substitution could lead to a reduction in GPC protein trafficking as well. Interestingly, we observed a partial restoration of replication in MACV GPC_{ΔN83/ΔN166/F438I} compared to MACV GPC_F438I. One previous report has indicated that glycan removal in the LCMV GPC also leads to enhanced replication in neurons (52). This could be due to a loss of steric hindrance from the N166 site, allowing for easier access to the cellular receptor. Future studies should focus on determining the exact mechanism for attenuation of these viruses that is so specific to neurons.

Overall, the results from our study demonstrate that a combination of mutations at glycosylation sites and the TMD of the arenavirus GPC is necessary for complete attenuation. We have demonstrated that MACV GPC_{ΔN83/ΔN166/F438I} is attenuated in two animal models, a key requirement in the development of potential vaccine candidates. Furthermore, immunization with MACV GPC_{ΔN83/ΔN166/F438I} is sufficient to protect guinea pigs from a lethal dose of wild-type MACV (Chicava). These results emphasize the necessity of including multiple attenuating mutations when designing attenuated arenavirus mutants. This is particularly important for the creation of potential vaccine candidates, as multiple mutations reduce the chance of genetic reversion and thus increase the safety of live attenuated vaccines. Taken together, our results demonstrate the feasibility of this approach in the rational design of future attenuated arenavirus mutants. While there are safety concerns over the use of live attenuated viruses as vaccines, it might be possible to generate genetically stable mutants that are highly attenuated and safe for use in diagnostic, antiviral, and molecular research outside high-containment facilities.

MATERIALS AND METHODS

Cells and viruses.

Vero cells (ATCC CCL-81), Neuro-2A (ATCC CCL-131), C8D1A (ATCC CRL-2541), 1321N1 (MilliporeSigma), IMR-32 (ATCC CCL-127), and BHK-21 cells (ATCC CCL-10) were maintained in minimal essential medium (MEM) (GE Healthcare Life Sciences) supplemented with 10% fetal bovine serum (FBS) (Life Technologies) and 1% penicillin-streptomycin (P/S) (Life Technologies) at 37°C with 5% CO₂. HEK-293T cells (Life Technologies) were maintained in Dulbecco's modified Eagle’s medium (DMEM) (GE Healthcare Life Sciences) with 10% FBS (Life Technologies) and 1% P/S (Life Technologies) at 37°C with 5% CO₂. The recombinant MACV Carvallo strain (30), MACV GPC_F438I (29), MACV GPC_ΔN83/ΔN166 (28), MACV GPC_ΔN83/F438I, MACV GPC_ΔN166/F438I, and MACV GPC_{ΔN83/ΔN166/F438I} were rescued by using a reverse genetics system previously described (30, 53). In our previous study, the mutation of either P85S or A168S/T was detected in 100% of the MCg1 (which has the ectodomain of Cd1 GPC in a MACV backbone)-infected animals at 42 dpi (28). Therefore, to disrupt the N-linked glycosylation site, two amino acids (N and S/T) of the sequon were mutated to minimize the possibility of reversion. The wild-type Chicava strain was a generous gift from Thomas Ksiazek. We used the first or second passage of viruses for all infection experiments in this study. For in vitro infection, confluent monolayers of Vero, C8D1A, or Neuro-2A cells were infected with viruses at an MOI of 0.01 for virus growth curves or MOI of 1 for protein expression profiling. After 1 h of incubation at 37°C with 5% CO₂, the medium was replaced with MEM containing 2% FBS and 1% P/S. The supernatants or cells were collected at the indicated days. Virus titers of serum samples, homogenized organs, or supernatant from cell culture were determined by plaque assay with tragacanth gum as previously described (27). All experiments with live viruses except Cd1 were performed in the biosafety level 4 (BSL4) laboratories at the Galveston National Laboratory (GNL) in accordance with institutional safety guidelines, NIH guidelines, and U.S. federal law.

Animal studies.

IFN-α/β and -γ R^−/− mice were bred and maintained in the animal BSL2 (ABSL2) facilities in the GNL at the University of Texas Medical Branch at Galveston. Five- to 9-week-old IFN-α/β and -γ R^−/− mice were challenged by intraperitoneal injection with recombinant MACVs (10,000 PFU) and monitored for 49 days postinfection (dpi). All animal experiments were performed twice in independent studies. Serum, brain, and liver samples were collected for virus titration and virus RNA copies when animals were euthanized or dead. Animals were humanely euthanized at the end of study (49 dpi) or if they became paralyzed or lost more than 20% of body weight. The Institutional Animal Care and Use Committee at the University of Texas Medical Branch at Galveston approved the study protocol (1208050A).

Six- to 8-week-old Hartley guinea pigs (Charles River Laboratories) were challenged by intraperitoneal injection with 10,000 PFU of recombinant MACV GPC_{ΔN83/ΔN166/F438I} (Carvallo backbone) or wild-type MACV Chicava strain. Animals were monitored for up to 90 days postinfection. Weights were taken daily for the first 30 days postinfection. Temperatures were recorded daily for the first 30 days postinfection using a Bio Medic Data Systems (BMDS) transponder following subcutaneous insertion of a transponder chip in each animal. Blood was collected via the vena cava at 60 dpi and at the time of euthanasia. At 60 dpi, surviving animals were rechallenged with 10,000 PFU of heterologous wild-type MACV (Chicava) intraperitoneally. Animals were humanely euthanized at 30 or 90 dpi or if they became paralyzed or lost more than 15% of their body weight. At the time of euthanasia, brain, liver, spleen, lung, and kidney samples were collected for virus titration, viral RNA content, and histopathology. The Institutional Animal Care and Use Committee at the University of Texas Medical Branch at Galveston approved the study protocol (2006067).

RNA extraction, cDNA synthesis, and sequence analysis.

RNAs were extracted from homogenized organs or lysed cell lines by using TRIzol reagent (Life Technologies) and Direct-zol RNA miniprep kits (Zymo Research, Irvine, CA) as previously described (26). Reverse transcription was performed by using the SuperScript III first-strand synthesis system (Life Technologies) with random primers according to the manufacturer's protocol. cDNAs for GPC open reading frame (ORF) were amplified by PCR with 5′-CGCACCGGGGATCCTAGGCGATTC-3′ and 5′-CCTCTCAGCCTTCTATTTCTACCC-3′. cDNAs for whole virus genome were amplified by PCR into two and three DNA fragments for viral S segment with previously described primers (27) and L segments with the following primers, respectively: 5′-CGCACCGGGGATCCTAGGCGTAAC-3′ and 5′-TAGGAACTGTGCCAGAAAGG-3′, 5′-TCTCTAACGCACTTGCTACC-3′ and 5′-TTGGATGTGCTGTGGTGAAC-3′, and 5′-GAGGATGTTGCCTAACTC-3′ and 5′-CGCACCGAGGATCCTAGGCGACAC-3′. To read the sequence of the 5′ and 3′ ends, the 5′ RACE system for rapid amplification of cDNA ends (Life Technologies) and the 3′ RACE system for rapid amplification of cDNA ends (Life Technologies) were used according to the manufacturer's protocol. After purification using a QIAquick PCR purification kit (Qiagen), PCR products were directly sequenced using an ABI Prism 3130xl DNA sequencer (Life Technologies).

Western blotting.

Western blotting was performed as previously described (21). Briefly, infected or transfected cells were harvested in the 2× Laemmli sample buffer (Bio-Rad) containing 5% β-mercaptoethanol and boiled at 95°C for 5 min. The samples were loaded in the wells of 4 to 15% Mini-Protean TGX gels (Bio-Rad). Then, proteins were transferred to polyvinyl difluoride (PVDF) membranes using the Trans-Blot Turbo transfer system (Bio-Rad). After blocking with phosphate-buffered saline (PBS) containing 0.05% Tween 20 (PBS-T) and 5% skim milk (blocking buffer) for 1 h at room temperature, the membranes were reacted with primary antibodies in blocking buffer at 4°C overnight. After three washes with PBS-T, the membranes were reacted with secondary antibody in blocking buffer for 2 h at room temperature. After three washes with PBS-T, the horseradish peroxidase (HRP) signal was visualized by enhanced chemoluminescence (ECL Western blotting system; Amersham). The primary polyclonal antibodies targeting the cytoplasmic tail of MACV GP2 were created by immunization of synthetic peptides (KYPRLKKPTIWHKR) (ProSci) in rabbit.

Plaque reduction neutralization test.

Serum samples were heat inactivated at 56°C for 30 min. The serum samples were diluted and mixed with an equal volume of diluent containing 80 PFU of each virus. After incubation for 1 h at 37°C, the mixture was applied onto Vero cell monolayers. After an incubation for 1 h at 37°C, the inoculum was replaced with tragacanth overlay (1.2% tragacanth gum mixed with equal volume of Temin's 2× MEM containing 4% FBS and 2% P/S) and incubated for 7 to 8 days. Then, the plates were fixed and stained with 1% crystal violet in 10% formalin. All serum samples were diluted at 1:30 to 1:960 and tested in the PRNT₅₀ assay. The PRNT₅₀ titer was determined based on the lowest dilution at which 50% of virus was neutralized. Samples that could neutralize over 50% virus at 1:960 dilution were recorded as ≥960. If a group included a sample with neutralization antibody titer >960, the titer of that sample was counted as 960 when calculating the average. The average of that group was presented as “≥Ave,” as the actual mean value may be higher than the average shown.

Enzyme-linked immunosorbent assay.

Purified MACV (Carvallo strain) GP1 or NP (Immune Technology) were used as antigens. ELISA plates were coated with 25 ng of antigen per well and incubated overnight at 4°C. Wells were then blocked with 5% skim milk suspended in PBS. After 1 wash with PBS-T, wells were incubated with serially diluted serum for 1 h at room temperature. Wells were then washed 5 times with PBS-T before incubation with secondary antibody (1:10,000 diluted goat anti-guinea pig IgG H&L [HRP]; Abcam) for 1 h at room temperature. Following 5 more washes with PBS-T, 3,3′,5,5′-tetramethylbenzidine liquid substrate, supersensitive, for ELISA (Sigma) was added to visualize the reaction. After 30 min of incubation at room temperature, 1 M phosphoric acid was added to stop the reaction. Optical density at 450 nm (OD₄₅₀) was then measured. A positive titer was considered to be anything above two standard deviations of the value for the corresponding naïve control serum.

Histopathological analysis.

Tissues were collected from dead or euthanized animals and fixed in 10% buffered formalin for a minimum of 5 days. Tissues were then cut, paraffin embedded, and sectioned into 5-μm slices for standard staining with hematoxylin and eosin (H&E).

Statistical analysis.

Data were analyzed using Dunnett’s post hoc test following a one-way analysis of variance (ANOVA), log-rank analysis, Welch’s t test, and the Mann-Whitney U test. Results were considered to be statistically significantly different when the P value was <0.05.