Andes Virus Genome Mutations That Are Likely Associated with Animal Model Attenuation and Human Person-to-Person Transmission

ABSTRACT We performed whole-genome sequencing with bait enrichment techniques to analyze Andes virus (ANDV), a cause of human hantavirus pulmonary syndrome. We used cryopreserved lung tissues from a naturally infected long-tailed colilargo, including early, intermediate, and late cell culture, passages of an ANDV isolate from that animal, and lung tissues from golden hamsters experimentally exposed to that ANDV isolate. The resulting complete genome sequences were subjected to detailed comparative genomic analysis against American orthohantaviruses. We identified four amino acid substitutions related to cell culture adaptation that resulted in attenuation of ANDV in the typically lethal golden hamster animal model of hantavirus pulmonary syndrome. Changes in the ANDV nucleocapsid protein, glycoprotein, and small nonstructural protein open reading frames correlated with mutations typical for ANDV strains associated with increased virulence in the small-animal model. Finally, we identified three amino acid substitutions, two in the small nonstructural protein and one in the glycoprotein, that were only present in the clade of viruses associated with efficient person-to-person transmission. Our results indicate that there are single-nucleotide polymorphisms that could be used to predict strain-specific ANDV virulence and/or transmissibility. IMPORTANCE Several orthohantaviruses cause the zoonotic disease hantavirus pulmonary syndrome (HPS) in the Americas. Among them, HPS caused by Andes virus (ANDV) is of great public health concern because it is associated with the highest case fatality rate (up to 50%). ANDV is also the only orthohantavirus associated with relatively robust evidence of person-to-person transmission. This work reveals nucleotide changes in the ANDV genome that are associated with virulence attenuation in an animal model and increased transmissibility in humans. These findings may pave the way to early severity predictions in future ANDV-caused HPS outbreaks.


RESULTS
Cell culture passaging of Andes virus strain ARG results in virulence attenuation in vivo. Andes virus strain ARG (ANDV/ARG) is one of a select few available strains isolated directly from the rodent reservoir, long-tailed colilargos (20). To our knowledge, it is also the only ANDV strain directly sequenced from rodent material (passage 0 [p0]). We hypothesized that cell culture passaging attenuates ANDV/ARG. To test this hypothesis, we passaged ANDV/ARG p9, described previously as causing 100% lethality in golden hamsters at 10 days after exposure (14), an additional 10 times in grivet Vero E6 cells (to p19). In a side-by-side comparison, all golden hamsters exposed via intramuscular injection of ANDV/ARG p9 uniformly reached euthanasia criteria, as expected, whereas 33.3% of those exposed to ANDV/ARG p19 recovered, and mock-exposed control animals uniformly survived (Fig. 1). Kaplan-Meier comparison of survival curves and log rank tests (Mantel-Cox [P , 0.0001; chi-square = 18.47], trend variation with the number of passages [P = 0.0101; chi-square = 6.610], and Gehan-Breslow-Wilcoxon [P = 0.0005; chi-square = 15.13]) demonstrated that these survival differences are statistically significant. ANDV/ARG RNA was consistently detected in golden hamster lung samples (3.8 Â 10 6 to 1.7 Â 10 10 RNA copies per 100 mg of perfused tissue) in the ANDV/ARG p9 and p19 cohorts but not in the mock-exposed control cohort.
ANDV/ARG is therefore not directly related to the other ANDV strains associated with person-to-person transmission. Interestingly, BASV clusters separately from ANDV sensu stricto together with LECV and ORNV; and, based on the analysis of the S, M, and L segments (Fig. S1), ANDV/CHI-9717869 appears to be the more ancestral strain. Furthermore, the analysis also shows that ANDV/ARG genetic distances to other strains reflect their geographic distribution (Fig. 2B).
Sequencing of passaged variants of Andes virus strain ARG reveals sites of adaptation associated with attenuation in the golden hamster model. To identify genotypic differences associated with golden hamster model outcome phenotype, we sequenced the S, M, and L genomic segments of ANDV/ARG p0 (sampled from the longtailed colilargo). The resulting isolate was seeded in Vero E6 cells for analysis of p3, p9 (14), and p19, as well as lung-tissue homogenates from golden hamsters exposed to ANDV/ ARG p9. We also included a human blood sample of Buenos Aires virus (BASV/BA02-C1S) from a case of HPS in La Plata, Provincia de Buenos Aires, in 2002 (21). We obtained complete genomic sequences for all segments (.98.3% coverage) for all ANDV strains, except for the L segment from the p0 strain (46.1% coverage). Sequences are available in GenBank under accession no. OP555720 to OP555735.
The comparative analysis revealed only a few nucleotide changes over passages (Table 1), and p0 and p3 sequences were identical. By p9, three single-nucleotide polymorphisms (SNPs) were observed: two in the S segment (S46N in the nucleocapsid [N] open reading frame [ORF] and V20I in the small nonstructural protein [NSs] [7] ORF) and one in the L segment (I1295M in the large protein [L] ORF). By p19, four additional SNPs were observed: two in the S segment, including a synonymous change at nucleotide position G57A and a nonsynonymous change at G103A (A21T), and two that were derived by nonsynonymous substitution, including one in the M segment (S97P), and one in the L segment (P1675S); also, one   Pathogenic Determinants in the Andes Virus Genome mSphere no reversions were detected in the genomic sequences of ANDV/ARG p9 in the lungs of golden hamsters exposed to ANDV/ARG p9. (Note that no data were collected from the lungs of golden hamsters exposed to ANDV/ARG p19). Sequencing of Andes virus strain ARG reveals virulence markers compared with pathogenic and nonpathogenic strains of Andes virus utilized in the golden hamster model. To identify potential genotypic virulence markers in the ANDV/ARG genome, we initially focused on 23 specific SNPs that had been described between the golden hamster attenuated ANDV/CHI-7913 compared to golden hamster lethal ANDV/CHI-9717869 (17). We also mapped five additional SNPs between those genomes, as the NSs ORF was not included in the original comparison (17). In 23 of those 28 positions, ANDV/ ARG p0 shared nucleotide bases with attenuated ANDV/CHI-7913. ANDV/CHI-97177869 and ANDV/ARG shared only position 11 of the Gn glycoprotein, position 938 of the Gc glycoprotein, and positions 20 and 37 of the NSs ORFs (Table 2). ANDV/ARG differ from both ANDV/CHI-7913 and ANDV/CHI-9717869 at genome position 46 of the N ORF.
Next, we focused on comparing the amino acid changes that arose during ANDV/ARG passaging with the differences in virulence observed in the golden hamster model. We identified five: A21T in the N ORF, V20I in the NSs ORF, S97P in the Gn glycoprotein, and I1295M and P1675S in the L ORF.
A21T, which appeared only in ANDV/ARG p19, occurs in a region known to participate in orthohantavirus NSs homotypic interactions (22). Additionally, we identified a second amino acid change in the N ORF (S46N), which was encoded only by ANDV/ARG p9 (Table 1 and  Table 2). Interestingly, in the same N ORF, Simons et al. reported an ANDV-specific kinaserecruitable hypervariable domain (HVD) in the N ORF by comparison of ANDV/CHI-7913 with other American orthohantaviruses and demonstrated its importance in regulating interferon (IFN) signaling (23). The HVD, which consists of 44 residues (nucleotides 252 to 296), encodes six characteristic amino acids (at positions A253, K262, N273, H286, T289, and T296) that are determinants of the phosphorylation of S386 in the N ORF, which is posited as a virulence determinant (23). Although we confirmed that S386 and five of the six residues are conserved among all ANDV and ANDV-like viruses ( Table 2 and Table 3), A253 is exclusive for ANDV, whereas P (BASV and LECV) or L (ORNV) is found in ANDV-like viruses; Q is found in MAPV, RIOMV, and LANV; and P is found in SNV and CHOV.
The recently discovered ANDV NSs antagonizes the type I IFN response by inhibiting mitochondrial antiviral-signaling protein (MAVS) signaling by binding MAV without disrupting MAVS-TBK-1 (22). In the presence of ANDV NSs, the ubiquitinylation of MAVS is reduced.      The V20I SNPs in the NSs ORF were observed arising in the ANDV/ARG p9 strain by the same mutation at nucleotide position 179 in the S segment. (Note that the NSs ORF is at position 11 compared with the N ORF.) ANDV/CHI-9717869 and ANDV/CHI-7913 differ in the NSs ORF at five amino acid positions (5, 20, 33, 35, and 37) ( Table 2 and Table 3). The S97P Gn change, found only in ANDV/ARG p19, could not be associated with any functional change. The site has only been reported as part of an antibody epitope (24). The S residue is conserved in BASV and LANV Gn proteins, whereas other orthohantavirus Gn proteins (LECV, ORNV, MAPV, RIOMV, SNV, and CHOV) have an A at that position ( Table 2).
M1295 and S1675 in the L ORF were encoded by ANDV/ARG p9 and p19 strains, respectively, but those genomic regions could not be associated with any functionality. M1295 has not been observed previously in nature; other orthohantaviruses have I1295 (MAPV, RIOMV, LANV, and CHOV) or Y1295 (SNV) ( Table 2). S1675 has not been observed in L ORFs of other orthohantaviruses, and the P1675 position appears entirely conserved ( Table 2).
Sequencing of Andes virus strain ARG reveals potential transmissibility markers when comparing Andes virus strains with differences of efficiency in person-to-person potential. The presence of outbreak-related determinants associated with person-toperson transmission was assessed by comparing genomic sequences of ANDV strains from clades clearly associated with person-to-person transmission and ANDV strains and ANDVlike viruses (BASV, LECV, and ORNV) that had not (Table 3). Interestingly, BASV, the most closely related ANDV-like virus ( Fig. 2A and B), has also been implicated in secondary transmissions but with limited efficiency (21,25).
Only one mutation in the M segment (resulting in T641I) was unique to person-to-person-associated clade 2 strains. Only one mutation in the S segment (resulting in A253N) was exclusively present in ANDV, whereas S386 is conserved among ANDV strains and ANDV-like viruses. Our analysis did not include the L segment of ANDV-like viruses because those sequences remain unavailable.
Five M ORF positions were unique to ANDV genomes compared with genomes of American orthohantaviruses (amino acid residue positions 499, 569, 570, 641, and 1133) ( Table 2). Four positions (569, 570, 641, and 1133) are also shared by ANDV-like viruses ( Table 3). S97 is encoded by all ANDV strains and BASV. V499I was present in some PTP strains and also in ANDV-like strains, whereas the T641I change was only encoded by ANDV strains from the clade associated with person-to-person transmission. However, only the latter (T641I) had been mapped in the vicinity of the absolutely conserved pentapeptide WAASA cleavage site, where signal peptidases cleave Gn and Gc (26). Since position 641 maps to a region that provides a signal to cellular peptidases, this change might affect the cleavage's efficiency. Signal peptides share several characteristic features determined by their amino acid composition (27), including a tripartite architecture with a positively charged N terminus and a hydrophobic segment that determines the strength of the signal. T641 changes from a polar noncharged amino acid (T) to a nonpolar (I) amino acid.
In comparison with the bulk of described ANDV isolates, the recently discovered ANDV NSs ORF presents seven sites of variation. Three are unique to ANDV/CHI-9717869 (Q5, E33, and L35) and two are unique to ANDV/CHI-7913 (I20 and D37). Intriguingly, we identified two SNPs in the NSs ORF at positions 40 (Q40R) and 47 (N47S) that were present only in the clade 2 strains (e.g., ANDV/Epuyén/18-19 and ANDV/Epilink/96) associated with personto-person transmission. Both NSs ORF changes need to be functionally evaluated for their effect on MAVS signaling.

DISCUSSION
Passaging in cell culture, especially when involving different hosts, usually results in virus adaptation, often affecting their virulence (19,28,29). However, ANDV/ARG p0 and p3 genome sequences were identical, and very few mutations were accumulated in the p9, p19, and hamster strains. The two amino acid substitutions (A21T and S46N) in N mapped to the intramolecular coiled-coil structure in the N-terminal region (a1 and a2), an exceptionally well-conserved region implicated in antibody recognition, formation of the ribonucleoprotein complex, and genome encapsidation (30)(31)(32)(33). Interestingly, one adaptation appears to involve a change in NSs, which has been recently related to IFN regulation. Only a single nucleotide change (T337C) was found in the M segment during late passaging (p19). This is unexpected since the M segment encodes Gn and Gc, two of the most variable regions of the genome in evolutionary terms.
Interestingly, we could also correlate some of the changes with differences in virulence in a small-animal model. ANDV/ARG p9 is uniformly lethal in hamsters (14). However, ANDV/ARG p19 was significantly less lethal (66.4%). Compared side by side, the ANDV/ARG p9 and p19 only diverged in five encoded residues (A21T and N46S in N, I20V in NSs, S97P in Gn, and P1675S in RNA-directed RNA polymerase [RdRp] encoded by the S, M, and L segments, respectively). Nevertheless, based on previous knowledge of functional domains, only the changes in N had been associated with viral replication. Structural studies of the N-terminal region of SNV and ANDV demonstrated that basic residues interact with the N core to stabilize interprotomer N association and formation of ribonucleoprotein (RNP) complexes (32). The A21T change likely affects that region, which is exceptionally well conserved among orthohantaviruses. The region is a target of the most cross-reactive antibodies against orthohantavirus, immunodominant, and proposed to have important effects regarding N polymerization, RNP complex formation, and subcellular localization of the assembly sites (30)(31)(32). We hypothesize that A21T and other changes in N ( Table 2 and Table 3) may affect N oligomerization dynamics. The importance of this area as a potential determinant of pathogenesis might be underscored by the observed differences in the region at positions 31 (A31T) and 38 (D38E) ( Table 3) that define ANDV-like viruses (i.e., BASV, LECV, and ORNV). The changes, all located at the bend between the two parallel coiled regions, could potentially affect the structure of the region. On the other hand, these two changes are only encoded by ANDV-like viruses, but not by LANV, MAPV, RIOMV, or SNV (Table 2). Thus, if these markers are associated with virulence, they would act via changes in the structure and not necessarily by SNP differences. Although the A21T change observed in late passages of ANDV/ARG is intriguing, A21 is conserved in BASV, LECV, and ORNV, but not SNV (Table 2). Collectively, this could indicate that structural changes in this area could be driving virulence differences instead of SNPs. Nonetheless, the limited animal data presented here could be improved with the use of imaging, advanced histological analyses, and other multiomics technologies. Support from developing countries and funding agencies for the study of these neglected pathogens is urgently needed as the resources for these types of studies in low-or middle-income countries (LMICs) are lacking.
Moreover, our analysis confirmed that the amino acid position S386, previously posited by Simons as a determinant of virulence (23), is conserved by ANDV, ANDV-like viruses (BASV, LECV, and ORNV), and LANV. In the N HVD, all ANDV strains share the described signature six residues, which are not found in any other orthohantavirus N HVD (Table 2). However, only five residues are shared with the three ANDV-like viruses, whereas A253 seems to be an exclusive ANDV marker (Table 3). We therefore suggest that A253 is an ANDV-exclusive virulence determinant and that the S386 modification and the five remaining HDV residues are virulence determinants for all viruses currently classified in the species Andes orthohantavirus (i.e., ANDV and ANDV-like viruses). However, based on the differences in lethality in hamsters among ANDV/CHI-9713 (nonlethal in golden hamsters), LANV (nonlethal in Brandt's hamsters [Mesocricetus brandti (Nehring, 1898)]) (34), MAPV (moderately lethal in golden hamsters), and ANDV/CHI-9717869 and ANDV/ARG (highly lethal in golden hamsters), the change in S386 does not appear to be a virulence determinant in hamsters. The N ORF has been associated with multiple functions associated with pathogenesis and virulence. The efficiency of orthohantavirus replication is inversely proportional to the ability of infected cells to activate MxA expression (35). The MxA protein is a critical component of the antiviral state induced by type I IFN (36). In turn, MxA protein binds to N, forming an MxA-N protein complex in a yet-to-be-defined manner (37). Moreover, the N protein also has a role in regulating the antiviral state. For instance, ANDV N hinders autophosphorylation of TBK1, resulting in the inhibition of interferon regulatory factor 3 (IRF3) phosphorylation and RIG-I/MDA5-directed type I IFN induction (38). Additionally, N can affect protein kinase R (PKR) dimerization (39), thereby preventing PKR phosphorylation, which is essential for its enzymatic activity. PKR inhibits virus replication (40). Bunyaviral NSs are nonessential for virus replication, but they are pathogenesis determinants by acting as IFN antagonists (41). As a case in point, ANDV/CHI-9717869 NSs antagonize the type I IFN induction pathway (22). We therefore hypothesize that the two changes observed in NSs of ANDV strains associated with person-to-person transmission might enhance IFN antagonist potential. Moreover, the number of changes in ANDV/CHI-9717869 compared with ANDV/CHI-7913 and ANDV/ARG might explain the differences in lethality in the golden hamster animal model.
In the M segment, the amino acid change T641I is also shared among ANDV strains associated with person-to-person-transmission but not among ANDV-like viruses. However, the change is also found in ANDV/NRC-6/18, which has not been associated with person-toperson transmission, and it is absent in ANDV/NRC-3/18, which has been involved in an event of secondary transmission (Table 3). T641 is located in the signal peptide of Gc, in the region preceding the hyperconserved cleavage site WAASA. Because host protease binding sites are guided by the signal from this region (22), we hypothesize that this change might affect the dynamics and speed of ANDV glycoprotein retention and trafficking. Signal peptides share several characteristic features determined by their amino acid composition (42), including a tripartite architecture with a positively charged N terminus and a hydrophobic segment that determines the strength of the signal. The comparative data presented here need to be complemented experimentally. For instance, recombinant vesicular stomatitis Indiana viruses expressing orthohantavirus glycoproteins could be used to partially compensate for the current lack of reverse-genetics systems for orthohantaviruses (43).
The phylogenetic analysis showed that ANDV/ARG is closely related to variants causing disease in humans and groups according to their geographic origin. ANDV/CHI-7913 is most closely related to ANDV/ARG, more than sequences obtained from patients reported in the region of endemicity. ANDV/CHI-9717869, on the other hand, is the most genetically divergent and remote geographically. Indeed, ANDV/ARG and ANDV/CHI-7913 share the most positions compared to ANDV/CHI-9717869. Thus, the decision to use ANDV/CHI-9717869 as the accepted exposure stock for medical countermeasure assessment needs to be revised, as this strain is a clear outlier that might not be representative of wild-type circulating strains.
Taken together, the results of our study indicate that determination and subsequent comparison of wild-type, cell-culture-passaged, and animal model-derived ANDV-and likely other orthohantavirus genome sequences-may allow predictions regarding their overall virulence and transmissibility, possibly informing countermeasure approaches. To strengthen such predictions, additional sequence information from yet-to-be-characterized ANDV strains and completion of genomic sequences of ANDVlike viruses are warranted.
Pathogenicity assessment. An established lethal animal model of ANDV infection, using golden hamsters [Mesocricetus auratus (Waterhouse, 1839)] (12), was leveraged to compare the previously established pathogenicity of the ANDV strain ARG (ANDV/ARG p9) (17) and to assess the pathogenicity of ANDV/ARG p19. Eight 12-week-old golden hamsters (four males and four females, obtained from the Instituto Nacional de Producción de Biológicos in Buenos Aires) were exposed intramuscularly to 100 mL of mock inoculum (phosphate-buffered saline [PBS]). Nine 12-week-old golden hamsters (four males and five females) were exposed intramuscularly to 100 mL of PBS containing 10 5 focus-forming units (FFU) of ANDV/ARG p19. Exposed golden hamsters were placed individually in ventilated cages and monitored daily up to 33 days postexposure. Food and water were available ad libitum. All animal experiments were performed in an accredited animal biological safety level 3 ABSL-3 biocontainment laboratory in compliance with institutional guidelines and Argentinian national law no. 14,346, which regulates experiments involving animals and adheres to principles stated in the Guide for the Care and Use of Laboratory Animals (44) of the National Research Council. An Institutional Animal Care and Use Committee approved all procedures involving animals.
RT-qPCR. Lung specimens were obtained from all golden hamsters following standard necropsy protocols. Total RNA was extracted from lung specimens using TRIzol, as described previously (45). RT-qPCR using ANDV genomic small (S) segment primers was performed following published procedures (46). Two microliters of each RNA sample were amplified in duplicate assays with a CFX detection system (Bio-Rad, Hercules, CA, USA), using TaqMan RT-PCR master mix (Quanta Biosciences, Gaithersburg, MD, USA), according to the manufacturers' instructions. A primer set designed to detect the human RNase P gene was used to ensure that samples were free of PCR inhibitors and that RNA extractions were homogeneous.
Genomic and phylogenetic analyses. Virus genome sequencing was performed using three ANDV cell culture passages (early [p3], intermediate [p9], and late [p19]), cryopreserved lung tissue from a naturally ANDV-infected long-tailed colilargo (p0), and lung tissues obtained from golden hamsters exposed to ANDV/ ARG p9. Also included in the analysis was a blood clot sample from a HPS patient (case C1-s, survivor, 14 years old) associated with secondary transmission of Buenos Aires virus (BASV) in Central Argentina (21).
Total RNA was extracted from cell culture supernatants, lung tissues, and clinical samples utilizing TRIzol. Virus genome sequencing was performed as previously described (11,47). Briefly, a targeted baitenrichment approach was used to enrich transcriptome sequencing (RNA-seq) libraries for sequencing on the MiSeq platform (Illumina, San Diego, CA, USA). Orthohantavirus sequences from each genomic segment (S, M, and L) were collected (Table S1) and aligned using MAFFT v.7.397, implemented in ClustalW version 2.0 (48). The initial data set consisted of coding-complete sequences obtained in this work and listed in Table S1. Other American orthohantavirus sequences from GenBank were also included. The resulting alignments were visually inspected to identify synonymous and nonsynonymous changes. Phylogenetic trees were reconstructed using IQ-TREE v.1.6.12 (49) with automatic model selection (50). Branch supports were assessed by 1,000 ultrafast bootstraps (49).
Data availability. Sequencing data are publicly available through GenBank under accession no. OP555720 to OP555735.