Molecular determinants of cross-species transmission in emerging viral infections

SUMMARY

Several examples of high-impact cross-species transmission of newly emerging or re-emerging bat-borne viruses, such as Sudan virus, Nipah virus, and severe acute respiratory syndrome coronavirus 2, have occurred in the past decades. Recent advancements in next-generation sequencing have strengthened ongoing efforts to catalog the global virome, in particular from the multitude of different bat species. However, functional characterization of these novel viruses and virus sequences is typically limited with regard to assessment of their cross-species potential. Our understanding of the intricate interplay between virus and host underlying successful cross-species transmission has focused on the basic mechanisms of entry and replication, as well as the importance of host innate immune responses. In this review, we discuss the various roles of the respective molecular mechanisms underlying cross-species transmission using different recent bat-borne viruses as examples. To delineate the crucial cellular and molecular steps underlying cross-species transmission, we propose a framework of overall characterization to improve our capacity to characterize viruses as benign, of interest, or of concern.

INTRODUCTION

Viruses able to cross species barriers pose a continuous threat to global health. Many viruses possess the capacity to spill over into novel host populations, but currently, we cannot reliably identify the inherent viral properties that predict successful cross-species transmissibility and pathogenesis. Numerous ecological, environmental, viral, and within-host factors influence spillover potential (1–3). Here, we focus on the entry, replication, and innate antagonist mechanisms of key bat-borne zoonotic viruses from three viral families (Coronaviridae, Filoviridae, and Paramyxoviridae) to discuss characteristics that may enable cross-species transmission and subsequent replication and disease in a novel host. Categorizing novel viruses as viruses of concern, viruses of interest, or viruses expected to be benign, similarly to the evaluation of the evolution of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), could enable us to prioritize surveillance and the development of interventions (4).

A variety of factors must align for a virus to transmit, infect, replicate, and cause disease in a novel host species. Layers of barriers exist at different scales, and the ability of a virus to traverse these obstacles will vary (1–3). Here we specifically focus on directly transmitted viruses and not on vector-borne zoonotic diseases whose transmission cycle is more complex. At the ecological level, an infected natural host must be in a certain spatiotemporal proximity to the novel host in question and shed the virus at a transmissible dose. Virus prevalence and duration of shedding in the natural population, as well as the host’s density in an environment, determine how frequently potential transmission occurs (5, 6). The route of shedding, as well as the virus’ environmental stability once excreted, will influence transmission dynamics. Once shed, if viable virus encounters a route of entry into a novel host, the virus must overcome within-host barriers. At the surface level, viral-host receptor compatibility facilitates cellular entry. When viral genetic material is released in the cell, the virus relies on the host’s machinery to initiate replication and/or translation. Incompatibilities between viral proteins and essential host factors preclude the production of progeny virus. Hosts are equipped with defense mechanisms that detect pathogens and can create a microenvironment inhospitable to viral replication. Viruses must thwart a host’s innate antiviral defense to propagate within a host. For a virus to perpetuate in a novel species, viable virus must be shed in a dose high enough to facilitate transmission. Virus transmissibility and pathogenesis may be decoupled. If the latent period precedes symptom onset, the selective pressure linking transmissibility to pathogenesis is weakened. In contrast, if the latent period overlaps with or follows symptom onset, virulence may be under negative selective pressure (7). If infection is debilitating or lethal prior to shedding onset, sustained replication in the population is not expected. Social behaviors between infected and uninfected individuals can also influence transmissibility regardless of virulence (8). Adaptive immunity influences the dynamics of a novel virus within a population. In addition, the evasion of humoral and cellular immunity depends on the virus’ mutational tolerance in combination with the establishment of immunity within the novel population. With the expanding virome (9), utilizing the genotype of a particular virus to make predictions regarding the susceptibility and fitness of a virus in novel hosts is a question of considerable interest. Can we apply genotypic information to predict which receptors novel viruses are using or which cellular co-factors are required for entry? Or can predictions regarding tropism and species compatibility of novel viruses be made based on comparisons with currently known pathogens? Can we infer innate immune antagonism functions of novel viruses, and can this correlate with a possible host range of the pathogen?

Here, we focus our discussion on select bat-borne zoonotic viruses known to cross-species barriers. We review the within-host barriers these viruses encounter and address the gaps in predicting the range of species susceptible to novel viruses. We also provide a framework for answering the elusive question whether genetic information from a virus and host of interest can predict susceptibility and viral fitness within that host.

BAT-BORNE VIRUSES WITH CONTINUED CROSS-SPECIES TRANSMISSION POTENTIAL

Filoviridae, a family of non-segmented negative sense RNA viruses, includes four bat-associated genera: Cuevavirus, Dianlovirus, Orthoebolavirus, and Orthomarburgvirus (10, 11). Viruses from the Orthoebolavirus and Orthomarburgvirus cause severe to lethal disease in humans and other mammalian species (12–14). Ebola virus (EBOV), Bundibugyo virus, Sudan virus, and Taї Forest virus from Orthorebolavirus (15, 16), and Marburg virus (MARV) and Ravn virus (RAVN) from Orthomarburgvirus (17) cause disease in humans. EBOV and MARV outbreaks have imposed the greatest burden on human populations, causing 24 and 15 outbreaks, respectively (16, 17). Bombali virus (BOMV) (18), a novel Orthoebolavirus, Lloviu virus (LLOV) (19, 20), the only known Cuevavirus, and Mĕnglà virus (MLAV) (21), the only known Dianlovirus, have not been associated with human disease. Whether the lack of known human infections with BOMV, LLOV, and MLAV is due to lack of exposure, inability to replicate, or absence of pathogenesis is unknown.

Henipavirus (HNV), a genus within the Paramyxoviridae family of non-segmented negative sense RNA viruses, includes bat-associated viruses that differ in their threat to global health. Hendra virus (HeV) and Nipah virus (NiV) cause severe to lethal disease in humans and agricultural animals. In Australia, HeV regularly transmits from Pteropus spp. to horses, often resulting in lethal infection in the horse (22–25). Efficient HeV replication in horses allows them to act as amplifying and bridging hosts, and infected horses can transmit HeV to humans caring for these animals (23, 24). Direct spillover of HeV from bats into humans has not been documented. In Malaysia, NiV infection in pigs resulted in a subsequent outbreak in humans (26–28). The implementation of changes to pig agricultural practices prevented future transmission events, and NiV has not caused subsequent outbreaks in Malayisa (29, 30). A different strain of NiV found in Bangladesh and India regularly spills over into human populations (31–34). Unlike the Malaysian strain, NiV from Bangladesh is documented to spill over into humans directly through the consumption of contaminated date palm sap (35, 36). HeV and both strains of NiV both cause severe to lethal disease in humans (23, 24, 26, 37, 38). Other bat-associated HNVs, including Cedar virus (CedV) (39), Ghana virus (GhV) (40), and Angavokely virus (AngV) (41), pose either an unknown threat to humans (GhV and AngV) (40, 41) or are suspected to be innocuous (CedV) (42, 43).

Coronaviridae, a family of enveloped, non-segmented positive-strand RNA viruses, contains many viruses infecting amphibians, birds, and mammals. To date, seven of those viruses transmit to humans (44–50) and cause a variety of clinical manifestations from asymptotic to severe fatal disease. The four human coronaviruses (hCoV), OC43, hCoV-229E, hCoV-NL63, and hCoV-HKU1, are responsible for common cold-like infections in humans and have a putative rodent (HKU1 and OC43) or bat (229E and NL63) origin (51, 52). Three hCoVs of bat origin which are highly pathogenic to humans have been identified. In 2002–2003, severe acute respiratory syndrome coronavirus 1 (SARS-CoV-1) spilled into the human population from bats via an intermediate host, causing an epidemic with a case fatality estimate of 10% (53). Currently, SARS-CoV-1 is no longer circulating in the human population. Since 2012, Middle East respiratory syndrome coronavirus (MERS-CoV), another coronavirus with an ancestral reservoir in bats, continuously causes spillovers from dromedary camels to humans (54). These small epidemics have a case fatality rate of up to 35% and continue to emerge throughout the Middle East (55). In late 2019, the third highly pathogenic hCoV spilled into the human population, SARS-CoV-2, causing a pandemic with, to date, over 770 million infections worldwide and an estimated 7 million deaths so far (56). Currently, there is no definitive answer regarding putative intermediate hosts in the spillover of SARS-CoV-2 to humans. Continuing emergence of SARS-CoV-2 variants which are arising due to immune pressures in the human population is a new feature not observed for the other highly pathogenic hCoVs (57).

TRANSMISSION OF EMERGING VIRUSES

Filoviruses are expected to transmit between humans mainly through contact with infectious body fluids (12, 58). More recently, sexual transmission from recovered patients to their partners has been described (58–61). The initial route of transmission into humans is expected to occur through interaction with an infected carcass (62, 63), such as from a lethally infected gorilla, chimpanzee, or duiker, or an actively shedding bat or its infectious secretions (64, 65). The Egyptian rousette bat (ERB) (Rousettus aegyptiacus) is a natural reservoir host of MARV and RAVN (65). In these bat populations, infectious virus is found in saliva (66, 67), and transmission is suspected to occur through biting, injuries, and/or direct contact. Capturing or butchering an infected bat enables contact with infectious secretions (68). The reservoir hosts of pathogenic Orthoebolaviruses remain unknown (69), but bats are suspected to be the natural hosts based on the identification of other filoviruses across bat families, including BOMV in Molossidae (18), LLOV in Miniopteridae (19, 20), and MARV, RAVN, and MLAV in Pteropodidae (21, 64, 65).

Transmission of NiV between people has been reported but is likely inefficient and requires prolonged close contact; however, up to five generations of human-to-human transmission have been reported for NiV Bangladesh (70). A few infected people have been responsible for as many as 22 secondary cases (70), and 5% of NiV patients are responsible for 86% of confirmed human-to-human transmission events (33). To date, no human-to-human transmission of HeV has been reported. The transmission route of HeV and NiV between flying foxes is unknown, but infectious virus is found in the urine (71–78). Since pteropid bats roost arboreally in large groups (5, 79, 80), infectious urine-mediated transmission events could occur via aerosol, droplets, or orally (71). Horses in Australia likely become infected with HeV through ingestion or inhalation of infectious urine-contaminated grass or partially eaten fruits (78), respectively. Transmission of HeV from horses to humans occurs through contact with infectious bodily fluids (23, 24, 81, 82). In Bangladesh, NiV transmission from pteropid bats to humans mainly occurs through consumption of contaminated, raw date palm sap (35, 36, 83). Although Indian flying foxes (Pteropus medius) are suspected to source the outbreaks in India (34, 84–88), the route of transmission is unclear. Alternative to physiological basis of shedding, behavioral differences of bats, such as interacting with more people while infectious, may also contribute to transmission efficiency. Future studies are needed to identify host and behavioral factors that may regulate transmission efficiency of henipaviruses.

Most coronavirus human-to-human transmissions happen via the airborne route, mainly through droplets and aerosols (89). Although contact and fomite transmission appear to be possible routes of transmission, their relative contribution compared to airborne transmission remains debated (90). Interestingly, in the putative bat hosts, coronavirus transmission seems to be associated with the fecal-oral route (91). This aligns with the presence of coronavirus receptors in gastrointestinal and respiratory tissues of bats (92). The recent SARS-CoV-2 pandemic highlighted that the replication location of a virus plays an important part in virus transmission as SARS-CoV-1 and SARS-CoV-2 show similar surface and aerosol stability but showed very different transmission dynamics (93). Transmission of MERS-CoV from dromedary camels, the current animal reservoir, occurs through close contact with animal respiratory secretions (94). Introduction of SARS-CoV-1 into the human population has been described via close contact transmission from the intermediate host civet cats (95). Cross-species transmission from its original host and possible intermediate hosts of SARS-CoV-2 with regard to the emergence in humans remains debated (96).

CELLULAR RECEPTORS FOR VIRAL ENTRY

An essential function of viruses is cellular entry. For viruses to cross species barriers to infection, viruses typically bind to conserved receptors. Representative zoonotic filo-, henipa-, and coronavirus families all enter using receptors well conserved in their animal hosts (Fig. 1). The characterized receptor for Orthoebolavirus and Marburgvirus is NPC1 (97, 98). This endosomal receptor functions in cholesterol transport (99) and is conserved among mammals (100). The filovirus glycoproteins (GPs) interact with NPC1 domain C loops 1 and 2 differently (101), allowing for distinct species-specific patterns of susceptibility (102–104) (Box 1). For example, among bat species, polymorphisms in the NPC1 EBOV- and MARV-binding motifs predict competence for viral replication (102). The straw-colored fruit bat (Eidolon helvum) is resistant to EBOV entry (105), while pteropid bats are predicted to be resistant to MARV entry (102).

Fig 1.

Commonalities in viral entry. Similarities of paramyxo-, filo-, and coronaviruses during cell entry. Viruses of all three families use a common mechanism to enter the cells. While both paramyxoviruses and coronaviruses can use membrane fusion at the cell membrane, filoviruses exclusively use an endocytic pathway via macropinocytosis. Depending on the presence of cellular proteases, coronaviruses might prefer an endocytic entry over membrane fusion. Commonly, all these viruses make use of proteolytic cleavage of their glycoproteins to facilitate entry. To further facilitate entry and dissemination, viruses might use attachment factors such as DC-SIGN or C-type lectins. Only henipaviruses have no known attachment factors besides their respective receptors. Entry mechanisms for these viral families have been reviewed and discussed in detail previously (106–108).

Box 1. Molecular basis of host receptor-virus compatibility.

The protease cleaved glycoprotein (GP_CL) of bat-associated filoviruses interacts with the intraluminal domain C of Niemman Pick C1 (NPC1-C) within endosomes. The crystal structure of EBOV GP_CL-human NPC1-1C [Protein Data Bank (PDB): 5F1B] in complex has been resolved (109), allowing visualization of the key residues that facilitate their interaction. The protruding loops 1 (418–428, between β sheets 2 and 3) and 2 (501–506, between α helices 4 and 5) of NPC1-C interact with a hydrophobic pocket on GP1. Loop 1 primarily contacts GP through polar interactions, while loop 2 interacts more extensively within GP’s hydrophobic cavity. The loop 1 and 2 interacting regions of GP differ among the filoviruses, leading to variation in the optimal NPC1-binding motif. Relative to the EBOV GP sequence, the regions interacting with NPC1 are within three main amino acid stretches, 79–86, 111–115, and 141–148, with position 150 and 178 also contributing for some filovirus species. The variations in the NPC1 contacting residues suggest evolution for optimized interaction with their respective host’s NPC1. Sequence variation in mammalian NPC1 at loops 1 and 2 influences filovirus species-specific differences in entry. For example, Yaeyama flying fox (YFF) (Pteropus dasymallus yayeyamae) loop 1 has amino acid differences at residues 425–427, TET, relative to human, SGA, which prevent efficient interaction with MARV GP (102). Mutating human NPC1 to encode the equivalent YFF residues, TET, inhibits MARV entry. Molecular dynamic simulations demonstrated that this mutation is predicted to decrease binding affinity between YFF NPC1 and MARV (modeled using the RAVN GP structure in complex with a neutralizing antibody, PDB: 5UQY) but not with EBOV GP_CL (103). Additionally, a single amino acid within the straw-colored fruit bat’s (Eidolon helvum) loop 2, F502, is sufficient to impair entry of recombinant vesicular stomatitis virus (rVSV) psuedotyped with EBOV, Bundibugyo virus, or Taї Forest GP (105). Furthermore, E. helvum NPC1 interacts with and supports entry of MARV, Sudan virus, and LLOV GP pseudotyped rVSV. Inserting a mutation EBOV GP, V141A, at the site that primarily interacts with F502 restored interaction and rVSV-GP-EBOV entry.

Applying what is known about the molecular basis of NPC1 loop 1 and 2 interactions with cleaved GP1 allows generation of testable phylogenomic predictions about which species have compatible NPC1 sequences for the different filovirus GPs. Similar predictions can be made using the structures for ephrin B2 with henipavirus G (110–113), ACE2 with SARS-CoV-2 S (114), and DDP4 with MERS-CoV S (115).

Among the bat-associated henipaviruses, ephrins are the receptors (110, 111, 116). Ephrins are important during embryonic development in vertebrates, and the amino acid sequence and predicted structures are well conserved across species (117, 118). The differential use of ephrins among henipaviruses also contributes to the clinical manifestation. The pathogenic NiV and HeV use ephrin B2 and B3, which are expressed in the vascular (119) and central nervous (120) systems, respectively, contributing to the vasculitis and neurological symptoms observed during infection (38, 121). The apathogenic Cedar virus cannot interact with ephrin B3 (112). GhV can also enter using ephrin B2, although its pathogenic potential is unknown (113). Species-specific differences in tissue-dependent expression of the ephrin genes may contribute to differences in clinical manifestations and disease severity. Furthermore, tissue-dependent expression of the receptor may determine susceptibility as the lack of receptor expression at the site of exposure may prevent infection.

Coronaviruses make use of a variety of host cell receptors for binding and cellular entry from proteinaceous to glycan-based receptors. Structural studies have been crucial in identifying the interaction motifs between the coronavirus spike protein and its host cell receptors (122). Both convergent evolution displaying distinct molecular interaction and receptor-binding motifs (RBMs) for the same receptor and divergent evolution using similar RBMs for different cellular receptors have been observed among the spikes of coronaviruses (123, 124). An example of convergent evolution is SARS-CoV-2, SARS-CoV-1, and HCoV-NL63 usage of the enzyme receptor angiotensin-converting enzyme 2 (ACE2) but usage of different RBMs and distinct molecular interactions within their spike to bind to ACE2 (123). Conversely, MERS-CoV and SARS-CoV-1 use similar RBMs within spike but recruit different host cell receptors, DPP4 and ACE2, respectively (125, 126). Four protein receptors have been identified as entry factors for coronaviruses: ACE2 (HCoV-NL63, SARS-CoV, and SARS-CoV-2), DPP4 (MERS-CoV), APN (HCoV-229E and TGEV), and CEACAM1 (MHV) (125–130). Additionally, OC43 and HKU1 have been identified to use glycan-based receptors carrying 9-O-acetylated sialic acid (131). The role of glycans during entry of SARS-CoV-2 has been discussed extensively (132, 133). The interaction of the SARS-CoV-2 receptor-binding domain with glycans seems to support a role as attachment factors as discussed below and also presents as a therapeutic target.

The spike protein and its interaction with the cellular receptor are the main determinants of entry for coronaviruses and therefore also the main determinants for species susceptibility. While many species express ACE2 or DPP4, the entry receptors for SARS-CoV-1/SARS-CoV-2 and MERS-CoV, respectively, amino acid variation in these receptors often determines susceptibility of the host for virus entry and thereby prevents infection (134, 135).

CO-FACTORS AND ATTACHMENT FACTORS FOR VIRAL ENTRY

In addition to the receptor, viruses use various attachment factors and proteolytic enzymes to facilitate entry. The use of broadly conserved and/or redundant mechanisms of attachment can enhance entry efficiency and may enable receptor-independent entry. Filovirus GPs can exploit a variety of attachment factors including Tim-1 (136–140), Tim-4 (141), HER2 (142), heparin sulfate (143), and DC-SIGN (144–146). The increased attachment can facilitate both dissemination and entry. Although henipaviruses do not have described attachment factors aside from the entry receptor, the endothelial surface expression of galectin-1 can influence the formation of syncytia (147). Differential host expression of galectin-1 might then influence pathogenesis to henipaviruses that bind ephrin B2 through inhibiting syncytia formation (147, 148). Also, coronaviruses can make use of attachment factors such as C-type lectins (149). Despite the important role of the protein receptors for coronavirus entry, interactions with C-type lectins and sialic acids as possible co-receptors in the entry pathway have been shown or proposed for most of the hCoVs (132, 150, 151). This is in line with other virus families which have been shown to rely on sialic acid interactions during viral entry (152).

For viruses to produce infectious virions, proteolytic cleavage of the glycoprotein or fusion protein is needed for fusogenicity. Filovirus GP contains cleavage sites for furin and cathepsins B and L (153–156). Although furin cleavage is not essential for EBOV to cause disease in non-human primates (157), it is possible that incorporation of a furin cleavage site (FCS) could enhance entry into a broader range of cell types. In contrast with other paramyxoviruses, henipaviruses do not use trypsin- and furin-like proteases (158, 159), but they instead rely on cathepsins (160, 161) which are typically expressed more broadly and may expand the cell types which may support HNV replication. This could have downstream impacts on viral load and clinical manifestation. Coronaviruses use various co-factors in the form of proteases during entry to process the spike protein. Presence of secreted or plasma-membrane expressed serine proteases such as TMPRSS2 will cleave the exposed spike protein cleavage site, resulting in membrane fusion at the plasma membrane (106). In the absence of serine proteases, the virus is initially endocytosed, and cysteine proteases present in the endosomes, such as cathepsin L, will subsequently cleave the spike protein to complete viral entry (106). Based on the presence of specific proteases on the host cell, the site of coronavirus entry is determined, which results in differential exposure to cellular restriction factors within the cell (162). Furthermore, transmembrane lectins have recently been described as attachment factors for SARS-CoV-2 which promote ACE2-dependent entry especially in tissue with low ACE2 expression (149). The influence of protease processing of spike on transmissibility of SARS-CoV-2 has been shown by the FCS. If the FCS is present in spike, cellular furin cleaves and separates the S1 and S2 subdomains of spike when new virions leave the cells (163). This increases receptor engagement of novel virions and increases transmissibility (164). However, this increased cleavage destabilizes the spike protein, which could lead to reduced transmission and therefore results in selection pressure for stabilizing mutations such as Q613H, D614G, and H655Y (165).

INNATE ANTAGONISM OF VIRAL REPLICATION

The host’s key innate antiviral defense is the type I interferon (IFN-I) system. Upon viral infection, viral nucleic acids and replication intermediates (166–168) can trigger the activation of pathogen recognition receptors (PRRs) (169–171). Downstream of pathogen recognition, a signaling cascade is initiated, resulting in the activation of transcription factor interferon regulatory factor 3 (IRF3) and IRF7, which are key to the expression of IFN-I and early IFN-stimulated genes (ISGs) (172, 173). IFN-Is bind to the dimeric receptor, IFNAR, to initiate the activation of JAK1 and Tyk2 kinases which phosphorylate STAT1 and STAT2 (174). Phosphorylated STAT1 and STAT2 heterodimerize and interact with IRF9 to form the ISGF3 complex. Once translocated into the nucleus, ISGF3 binds to ISG promoters to induce their expression. ISGs are the effector molecules which facilitate an antiviral cellular environment. ISGs have different roles in impeding a virus’ progression through its replication cycle, whether blocking entry, uncoating, replication, transcription, or egress. Different viruses are susceptible to a subset of ISGs, depending on the host machinery the virus uses and the evolution of viral resistance (175–177).

Antagonism of IFN-I induction

A common feature of virulent zoonotic viruses is the ability to antagonize IFN-I induction and avoid recognition by cellular PRRs. Human pathogens from the filo-, henipa-, and coronavirus families target IFN-I induction redundantly (Fig. 2). Filovirus polymerase cofactor VP35 contains a double-stranded RNA (dsRNA)-binding motif (178–183) (Box 2) enabling titration of dsRNA away from the cytoplasmic RNA PRR RIG-I (178, 179, 184, 185). This masking of dsRNA prevents early IFN-I induction and provides the virus with a favorable cellular microenvironment for replication. Pathogenic henipaviruses prevent PRR recognition using accessory proteins V and W whose mRNA are transcribed from the phosphoprotein (P) gene using RNA editing (186). The V protein’s unique C-terminal domain blocks the activation and downstream signaling of cytoplasmic RNA PRR MDA5 (187–190). Coronaviruses have also evolved functions to mask their genetic material from PRRs by modifying the 5′-triphosphate using capping and methylation to prevent detection of genomic and subgenomic RNAs (191). Additionally, coronaviruses replicate in double-membrane vesicles which form the replication organelles and protect their dsRNA intermediates during replication from being detected by PRRs (192, 193). Next to antagonizing IFN-I induction at the PRR level, filovirus VP35 and henipaviruses V and W block IKKε- and/or TBK1-mediated IRF3 phosphorylation (194, 195). The matrix protein (M) of bat-associated henipaviruses has also been shown to block IRF3 activation (196). Similarly, coronaviruses like SARS-CoV-2 interfere with IFN-I induction by reducing expression of TBK1 and blocking RIG-I/MDA5 signaling in a mitochondrial antiviral-signaling protein-dependent manner using ORF7a/b protein (197, 198).

Fig 2.

Viral antagonism of host type I interferon induction and signaling. Viral proteins (colored hexagons: Orthoebolavirus in magenta, Orthomarburgvirus in orange, Henipavirus in cyan, and Coronavirus in gold) encode various antagonists that target one or more host proteins (purple shapes). (A) The host responds to cytoplasmic double-stranded RNA (dsRNA) through cytoplasmic dsRNA recognition receptors RIG-I and MDA-5. MDA5 and RIG-I then induce downstream signaling through the mitochondrial antiviral-signaling protein (MAVS), which recruits various kinases and ubiquitin ligases to facilitate the activation and phosphorylation of transcription factors that stimulate type I interferon (IFN-I) synthesis. (B) IFN-I signals through the dimeric IFN-I receptor (IFNAR1 and 2) to trigger the activation of tyrosine kinases JAK1 and Tyk2, which phosphorylate STAT1 and STAT2. The phosphorylated STATs interact with IRF9 to form the ISGF3 complex and translocate to the nucleus to induce the transcription of antiviral effector molecules, IFN-stimulated genes (ISGs).

Box 2. Molecular basis of viral innate immune antagonism.

Nipah and Hendra virus phosphoproteins (P) encode four proteins. P protein and accessory proteins V and W are transcribed through the same transcriptional start site and share their amino-terminal domain, but V and W encode distinct C-terminal domains due to an RNA editing-induced frameshift through the insertion of one or two nucleotides in the mRNA, respectively (186). A region within the shared N-terminal domain (amino acids 114–140) facilitates interaction with host interferon signaling transcription factor STAT1 (199, 200). Recombinant Nipah virus encoding P gene mutations to delete the STAT1 interacting domain (Δ114–140) or a point mutation at a key interacting residue (Y116E) significantly impaired type I interferon (IFN−) signaling in vitro and delayed the time to death of infected ferrets (200). Recently, the same P/V/W domain has been shown to facilitate interaction with STATs 2 and 4, but not STAT3 (201). Mutating the middle of STAT1’s SH2 domain (amino acids 610–646) to those of STAT3 abrogated interaction with P (201). Although no structure has been resolved for P, V, or W in complex with STAT1, 2, or 4, this initial molecular mapping provides a basis for identifying species-specific variation in sequence that could impact interaction and effectiveness of innate immune antagonism. Hypothesized differences in binding could be tested using high-throughput screens and were last tested using recombinant viral and host proteins in the context of infection.

Another mechanism viruses use to block IFN-I is through binding double-stranded RNA (dsRNA) in the host cytoplasm. Filovirus VP35 encodes a dsRNA-binding domain within the C-terminal central basic patch (amino acids 294–340) sufficient for attenuating IFN-I induction (178, 180–182, 202). Mutation of EBOV VP35 residues in direct contact with the dsRNA phosphodiester backbone (R312 and R322) or that facilitate intermolecular (R312 and R322) and intramolecular (K339 and I340) VP35 interaction impaired dsRNA binding and antagonism of IFN-I induction (180, 181). Other residues that interact with the dsRNA (P233, T237, F239, S272, D274, C275, I278, A306, K309, S310, and I340) had varying impacts on IFN-I production (181). Mutating EBOV to encode mutations that functionally impair VP35’s IFN-I block is attenuated in both non-human primate (203) and rodent (204) disease models. Prior to VP35’s crystallization, this type of RNA-binding domain had not been described and could not be predicted bioinformatically at the time. The identification and molecular characterization of novel domains is vital to predicting the function of novel viral proteins.

Antagonism of IFN-I signaling

Besides IFN-I induction, these viruses also block the signaling pathways of IFN-I. One example is the activation and/or nuclear translocation of STAT1 and STAT2. Pathogenic henipaviruses block the phosphorylation of STATs using the N-terminal motif shared among P, V, and W (195, 199, 205) (Box 2). While the V protein functions in the cytoplasm to block STAT phosphorylation, the W protein functions in the nucleus and blocks ISGF3 binding to ISG promoters (205). Though P contains the STAT-binding domain, it is thought that P preferentially engages in its polymerase co-factor function. VP24 of orthoebolaviruses, LLOV, and MLAV blocks nuclear transport of STAT1 through interaction with the nuclear importins in the karyopherin in the NPI-1 family (206–209). Furthermore, the orthomarburgvirus MLAV uses VP40 to antagonize JAK1 and prevent STAT1 phosphorylation (210, 211). The SARS-CoV-2 proteins nsp1 and nsp6 comparably interfere with STATs by depleting TYK2 and/or STAT2 in the case of nsp1 (212) or prevent phosphorylation of STAT1/STAT2 by nsp6 (213). Impeding the activation of STATs and IFN signaling prevents activation of IFN-stimulated genes which favors viral replication.

Implications of the IFN-I pathway during infection

The importance of innate antagonist viral proteins has been demonstrated using recombinant mutant viruses in vivo. Mutating VP35’s RNA-binding domain attenuates Ebola virus disease in non-human primates (203) and in guinea pigs (204) and mice (214) infected with rodent-adapted EBOV. Acquisition of a mutation in VP24 that prevents binding to karyopherin to block STAT1 nuclear translocation is essential for mouse-adapted EBOV to cause disease (214). Similarly, mutation of the VP40 gene to enable IFN-I signaling antagonism causes lethal disease in mice infected with mouse-adapted MARV (215). Ablation of the STAT1-binding domain in the P gene partially attenuates acute NiV disease, but infection was lethal in all infected ferrets (200). In contrast, deletion of accessory protein V in the NiV-Malaysia background resulted in non-lethal infection (216). The apathogenic henipavirus, Cedar virus, lacks the RNA editing site and was unable to block IFN-I induction efficiently (39, 42, 43). Similarly, SARS-CoV-1 caused lethal infection in STAT1 knock-out mice in an interferon independent manner while causing a non-lethal infection in BALB/c mice (217). In contrast, SARS-CoV-2 in vivo replication can be effectively controlled by interferons in certain mouse models (218).

Influence of innate antagonism on the adaptive immune response

The antagonism of the innate response, particularly in filoviruses, has repercussions on the establishment of an adaptive immune response. The innate antagonist functions of VP35 and VP24, as described in the preceding sections, prevent the maturation of dendritic cells (219–221), a primary early target of filoviruses. During infection, the antigen-presenting cells have decreased expression of co-stimulatory molecules, CD80 and CD86, and cytokines needed to activate lymphocytes (219–222). Infected myeloid cells also tend to produce pro-inflammatory cytokines to recruit additional susceptible myeloid cells facilitating virus amplification and dissemination. Furthermore, filovirus infection causes lymphopenia presumably through abortive infection that induces cell death (223). Although henipavirus infection is not associated with leukocyte infection, blocking of the innate immune system delays the onset of the adaptive immune response.

DISCUSSION

The emergence of bat-borne viruses has increased focus on the diversity and evolution of the animal virome (9), and many novel viruses have recently been identified in bats (18, 40, 41, 224–228) and other mammalian species (226, 228, 229). A pressing question is whether we can apply the increasing genetic information to predict efficiently and accurately the host range, pathogenic potential, and transmissibility of these novel viruses with unknown cross-species and zoonotic potential. Several studies have evaluated the phenotypic characteristics of novel coronaviruses (230–233) and paramyxoviruses (41, 234, 235). A primary limitation of this approach is inefficiency and limited capacity to evaluate the main phenotypic characteristics in a high-throughput fashion. These experiments rely on recovery of a virus isolate, animal model development, and a myriad of in vitro experiments. Time and resource investment to identify receptors, characterize the innate immune antagonist functions, and determine transmissibility is immense and not feasible to allow real-time phenotypic characterization for the number of novel viruses continuously being discovered. Developing bioinformatic and structure-based pipelines to assess molecular signatures alongside high-throughput in vitro methods could aid in predicting cross-species transmission potential of a higher number of novel viruses.

General genotype-to-phenotype pipeline

The general workflow to phenotypically characterize these novel viruses seems clear: full-genome sequencing, in combination with virus isolation; application of in vitro and in silico methods to identify the receptor, evaluate replication efficiency in different cell lines, establish an animal model, test sensitivity to known antivirals; and using a prototype pathogen approach (e.g., using SARS-CoV-2 as a prototype pathogen for a newly discovered betacoronavirus) to evaluate applicable countermeasures to the selected prototype and readily adaptable to the newly discovered pathogen (236). Several potential roadblocks exist along this pathway that may impair accurate predictions including a lack of usable animal models, unknown receptor identity, and focus on humans. Furthermore, the more distantly related a novel virus is to characterized viruses, the less likely we are to make useful predictions.

Classifications of novel pathogens from benign to concerning

To improve the classification of novel viruses into benign, interesting, or concerning based on their genetic sequence, a primary focus should be to better understand the molecular mechanisms of virus-host interactions in known viruses. Increased characterization of existing viruses will allow a better inference of that knowledge to inform and understand novel viruses and their genetic makeup. Initially, we are interested in predicting host range based on genomic sequences. The recent expansion of mammalian genomes available (237–239) bolsters the feasibility of phylogenomic prediction. A main limitation for this predictive model is that completely novel or distantly related viruses may not be categorized accurately due to data gaps. The lack of a known receptor for a virus family, such as pararubulaviruses, also inhibits inference. Host receptor-viral surface protein compatibility does not always translate to susceptibility. For example, ERB ephrins are predicted to interact with NiV G, but they do not support viral replication in vitro or in vivo (240). Additionally, ERB’s NPC1 is predicted to support EBOV entry (105) and support in vitro replication of multiple orthoebolaviruses (241), but ERBs do not support disseminated Orthoebolavirus replication, and infection can be quenched without mounting an adaptive immune response (242). These instances illustrate the importance of investigating host-virus interactions beyond receptor binding alone to predict host range.

Determination of innate antagonist proteins

To aid in inferring pathogenesis, we also suggest investigating whether a novel virus encodes innate antagonist proteins and their possible functionality. For example, the lack of RNA editing in the CedV P gene led to the experimentally supported prediction that this virus is apathogenic (39, 42, 43). In the case when no isolate is available, the predicted innate antagonists can be ectopically expressed or encoded into a related recombinant virus to assess function. While this ectopic expression commonly results in higher expression compared to expression during a natural virus infection, this enables the interrogation of gene function with the caveat of potential false positives. When a virus isolate is available, the suppression or activation of the immune system alongside viruses with known innate antagonist capacity can be evaluated in the context of infection. Reliance on cell lines can yield false-positive or false-negative results, depending on how well the cell line reflects the tissue the virus targets naturally. Furthermore, we are limited in cell line availability to a remarkably small number of animal model species and cell types. An opportunity to further develop our predictive skills is advancing basic research of the innate immune systems of non-model mammalian species. This research can provide key insights into species-specific regulation of innate antiviral defenses as well as facilitate advancements in conservation medicine for endangered or threatened species vulnerable to autoimmune or infectious diseases.

Considerations around genetic variation

Another important factor to consider is host genetic variation. The use of collaborative cross mice in the context of mouse-adapted EBOV infection uncovered distinct disease phenotypes ranging from tolerogenic to lethal with hemorrhagic fever presentation (243). Although the mechanism has not been elucidated, tolerant phenotypes were associated with genetic differences in Tie1 and Tie2, which encode endothelial cell-specific tyrosine kinases that are hypothesized to protect against hemorrhagic progression during infection (243). Furthermore, transcriptional differences between the tolerogenic and lethal populations could be used to create a model that could predict disease outcomes in human EBOV patients with 75% accuracy (244). Development of such machine learning-based models for other pathogens could inform whether novel viruses may cause disease based on transcriptional profiles. While this example highlights the influence of host genetic diversity, the genetic diversity within animals, such as collaborative cross mice available for in vivo experiments, likely only accounts for a minor part of the genetic diversity compared with the diversity observed in wildlife populations.

Additionally, it is important to evaluate the potential contribution of inborn errors of immunity and immune suppression on susceptibility and competence. Although rare, human mutations in key innate immune genes can lead to increased susceptibility to viral infections or severe disease associated with immune dysregulation (245–249). Recently, autoantibodies to IFN-I have been described to contribute to severe disease in SARS-CoV-2 patients (250–252), while co-morbidities can also enhance susceptibility to infection, as shown for diabetes in coronavirus infections (253, 254). Immunocompromised patients, due to medical procedures, immunotherapies, or chemotherapies could also be at increased risk of infection with viruses that may otherwise be predicted to be inert in humans (255, 256). The role of these individuals as initial susceptible patients, which could lead to subsequent adaptation of the virus to enable human infections and/or human-to-human transmission, remains underexplored.

Role of transmission

The final within-host factor to assess is capacity of transmission of infectious virus. A major factor in determining whether a novel virus can be sustained within a population is transmissibility of infectious virus. Predicting transmissibility of a novel virus is not as straightforward. A host may support disseminated infection and present with clinical symptoms without shedding infectious virus at levels sufficient to infect another individual. Many vector-borne viruses, including Japanese encephalitis virus and West Nile virus, can be pathogenic in humans but function as dead-end hosts (257, 258). Respiratory viruses, in contrast, such as human betacoronaviruses and seasonal influenza A and B viruses, transmit efficiently human-to-human. Although HeV and NiV can have a respiratory presentation with viral shedding in saliva and urine (33, 70), transmission is inefficient and unlikely to be sustained in the human population to the degree of SARS-CoV-2. Understanding the molecular mechanisms that influence shedding could have significant impact on inferring viruses with pandemic potential. The mechanisms that regulate transmission are likely virus- and species-specific, which impairs predictability of transmission.

Concluding remarks

At this stage, we are not equipped to predict the host range, pathogenic potential, or transmissibility of a novel virus based on sequence alone. With the development of machine learning models paired with solid molecular-level data for training, we can improve our capacity to characterize viruses within known virus families as benign, of interest, or of concern. Our largest gaps remain in understanding how host genetics and immune status may influence initial spillover events and adaptation, as well as predicting transmissibility. The gaps are more extensive for virus groups without characterized receptors, molecularly defined innate antagonist mechanisms, or disease mechanisms. Individual viruses genetically distant from known pathogenic viruses are even more difficult to predict. Technological innovations, such as structure-based protein function prediction and large-scale metagenomic sequencing and surveillance, combined with classic virological and molecular assessment of host-virus interactions, will be needed to narrow these gaps.