Filoviruses: Ecology, Molecular Biology, and Evolution

Abstract

The Filoviridae are a family of negative-strand RNA viruses that include several important human pathogens. Ebola virus (EBOV) and Marburg virus are well-known filoviruses which cause life-threatening viral hemorrhagic fever in human and nonhuman primates. In addition to severe pathogenesis, filoviruses also exhibit a propensity for human-to-human transmission by close contact, posing challenges to containment and crisis management. Past outbreaks, in particular the recent West African EBOV epidemic, have been responsible for thousands of deaths and vaulted the filoviruses into public consciousness. Both national and international health agencies continue to regard potential filovirus outbreaks as critical threats to global public health. To develop effective countermeasures, a basic understanding of filovirus biology is needed. This review encompasses the epidemiology, ecology, molecular biology, and evolution of the filoviruses.

1. INTRODUCTION

The discovery of filoviruses dates back five decades when cases with a hemorrhagic fever syndrome were investigated in Marburg, Germany and Belgrade, Yugoslavia (now Serbia). The source of infection was associated with handling African green monkeys imported from Uganda. This outbreak resulted in a total of 32 human cases and 7 fatalities across Europe. The etiologic agent was subsequently identified and characterized; it was named Marburg virus (MARV) after the location of the first cases (Siegert et al., 1967; Slenczka, 1999). Nine years later two outbreaks of hemorrhagic fever occurred almost simultaneously in southern Sudan (now the Republic of South Sudan) and northern Zaire (now the Democratic Republic of the Congo, DRC). Studies revealed a new virus which was named Ebola virus (EBOV) after a nearby river in northern Zaire. The virus was related but distinct from MARV (WHO, 1978a,b). A few years later, molecular characterization revealed distinct ebolaviruses as the causative agents for the two outbreaks with EBOV causing disease in northern Zaire and the Sudan virus (SUDV) being responsible for disease in southern Sudan (Richman et al., 1983).

Despite infrequent smaller outbreaks and episodes of viral hemorrhagic fevers caused by the above-described filoviruses throughout sub-Saharan Africa, the next discovery of a new ebolavirus dates to 1989 in Reston, Virginia, USA. Macaques imported from the Philippines showed signs of hemorrhagic fever during quarantine and a new filovirus, named Reston virus (RESTV) after the geographic location of first isolation, was coisolated from these macaques together with simian hemorrhagic fever virus (Jahrling et al., 1990). Despite verified human infections, today RESTV is considered apathogenic for humans and epizootic in the Philippines and perhaps greater Southeast Asia (Miranda and Miranda, 2011).

In 1994, another different ebolavirus, Taï Forest virus (TAFV), was discovered in a single nonfatal case of viral hemorrhagic fever in Côte d’Ivoire. Until today, this remains the only described human TAFV case. A massive, years-long ecological investigation in the Taï Forest did not reveal the reservoir for TAFV (Formenty et al., 1999; Le Guenno et al., 1995). The newest ebolavirus to be discovered was Bundibugyo virus (BDBV) in Uganda in 2007 (Towner et al., 2008; Wamala et al., 2010) where it caused an outbreak of hemorrhagic fever in humans.

All known marburgviruses as well as EBOV, SUDV, and BDBV are considered pathogens of significant public health importance in Africa. Outbreaks are especially notable for their high case fatality rates, commonly ranging from 25% to 90%, depending on the virus species (Muyembe-Tamfum et al., 2012). The recent West African EBOV epidemic has dramatically emphasized the public health relevance of these pathogens with unprecedented case numbers and fatalities (approximately 30,000 cases and 11,000 fatalities) (WHO, 2016). For the first time, it has added a global public health perspective to filoviruses with multiple introductions of exposed people and confirmed cases through returning travelers and evacuations of health care workers from West Africa (WHO, 2016).

In 2011, sequences of a novel filovirus, related but distinct from ebolaviruses and marburgviruses, were detected in specimens derived from dying bats in Spain (Negredo et al., 2011). An isolate is missing, but the putative virus was named Lloviu virus (LLOV) and is classified in its own genus Cuevavirus within the Filoviridae. LLOV has not been associated with human infections.

2. TAXONOMY

The Filoviridae family is grouped within the order Mononegavirales, a diverse taxon of unsegmented negative-sense single-stranded RNA viruses. Other virus families within this order include Bornaviridae, Mymonaviridae, Nyamiviridae, Paramyxoviridae, Pneumoviridae, Rhabdoviridae, and Sunviridae. The taxonomy of filoviruses was substantially changed in 2010 to reflect advances in the field and has since been revised (Bukreyev et al., 2014; Kuhn et al., 2010). Filoviruses share a long, filamentous morphology and comprise three distinct genera: Ebolavirus, Marburgvirus, and Cuevavirus (Fig. 1). The Ebolavirus genus is the most diverse, consisting of five recognized species: Zaire ebolavirus, Sudan ebolavirus, Reston ebolavirus, Taï Forest ebolavirus, and Bundibugyo ebolavirus (represented by EBOV, SUDV, RESTV, TAFV, and BDBV, respectively). The Marburgvirus genus consists of only a single species, Marburg marburgvirus, that has been further delineated into the variants MARV and Ravn virus (RAVV). The relatively new Cuevavirus genus also consists of a single species, Lloviu cuevavirus (represented by LLOV). This putative classification of LLOV is based on genomic sequence data only as a virus isolate could not be obtained (Negredo et al., 2011). In 2015, novel filovirus sequences were detected in Chinese fruit bats, but a virus has not been isolated and a whole-genome alignment has yet to be obtained (He et al., 2015). A basic phylogeny of the family Filoviridae is provided in Fig. 1.

Fig. 1.

Phylogenetic tree of the Filoviridae family. This tree was generated based on the nucleotide sequences for the GP gene coding regions. All sequences were acquired from GenBank (LLOV, NC_016144.1; MARV, NC_001608; SUDV, NC_006432.1; RESTV, NC_004161.1; EBOV, NC_002549; TAFV, NC_014372.1; BDBV, NC_014373.1; RAVV, NC_024781).

3. VIRUS ECOLOGY

Filoviruses only sporadically infect human and nonhuman primates but principally circulate in nonhuman hosts. The exact mechanism of zoonotic transmission is unknown, but unprotected exposure to the reservoir, handling contaminated bushmeat, and receiving bites from an infected animal are possible routes (Azarian et al., 2015; Chippaux, 2014). MARV, RAVV, BDBV, TAFV, EBOV, and SUDV circulate in the tropical regions of sub-Saharan Africa, whereas RESTV has been identified in the Philippines and sequences of LLOV in Spain. Although EBOV has not been isolated outside Africa, filovirus-specific antibodies cross-reactive with EBOV have been found in Rousettus leschenaulti bats in Bangladesh and in Pongo pygmaeus orangutans on Borneo (Nidom et al., 2012; Olival et al., 2013).

Furthermore, in China, new filovirus sequences with phylogenetic identities intermediate between EBOV, RESTV, and LLOV have been detected (He et al., 2015). These findings suggest that there are hitherto unidentified filovirus species circulating in Asia. At present, all filoviruses are considered serious human pathogens apart from RESTV, which is thought to be apathogenic although unambiguous evidence for human infection is lacking, and the nonisolated LLOV and Chinese filoviruses, for which pathogenicity in humans is unknown.

Over the past few decades, there has been an increasing effort to identify the zoonotic origin of filoviruses. This task has been complicated by the remote and ecologically rich jungle habitats where outbreaks are known to occur, as the diverse number of mammalian species in these regions makes it difficult to identify putative reservoir hosts. One commonly accepted view is that animal-to-human transmission events are exceedingly rare, but virus spread between humans after an initial introduction is efficient. Recent evidence from the 2013 to 2016 EBOV epidemic supports this hypothesis, as genomic surveillance data postulate that a single animal-to-human transmission was responsible for the epidemic, even though an increased rate of human-to-human transmission gave rise to an accumulation of largely nonsynonymous mutations (Gire et al., 2014). The 1995 Kikwit EBOV outbreak is also thought to be caused by a single reservoir-to-host introduction, but in contrast to the 2013–2016 epidemic, no sequence variation in the EBOV glycoprotein (GP) gene was found possibly due to a more limited number of transmission chains in humans (Rodriguez et al., 1999). Other outbreaks, including a series of recurrent outbreaks in Gabon and the Republic of the Congo (RC), were also characterized by shorter transmission chains of multiple EBOV lineages that affected both human and nonhuman primates with little or no evidence of mutations in the end host (Leroy et al., 2004). Lastly, the MARV outbreak in Durba/Watsa in 1998–2000 was characterized by multiple simultaneous introductions of distinct genetic MARV lineages into the human population with shorter transmission chains in the end host (Bausch et al., 2006).

3.1. Ebolaviruses

It is well known that EBOV, and perhaps also other ebolaviruses, affect other primates besides humans. Wild gorilla (Gorilla gorilla) and chimpanzee (Pan troglodytes) populations have experienced significant declines as a result of EBOV outbreaks (Bermejo et al., 2006; Formenty et al., 1999; Rouquet et al., 2005). The severe pathogenicity of EBOV in nonhuman primates and the sporadic nature of infection suggest that primates are dead-end hosts like humans. Beginning with the detection of EBOV genomic material in several Old-World fruit and insectivorous bat species, bats have come under increasing suspicion as filovirus carriers (Hayman et al., 2010, 2012; Leroy et al., 2005; Ogawa et al., 2015; Pourrut et al., 2009). However, ebolaviruses have yet to be isolated from any wild bat or animal species. Table 1 summarizes evidence and proof for the detection of filoviruses in bats. Evidence of EBOV and RESTV has been found in numerous bat species spanning a wide geographical area, including Ghana, Gabon, RC, DRC, Zambia, Bangladesh, Philippines, and China. In contrast, BDBV, SUDV, and TAFV have not been surveyed to the same degree (Table 1). An overview of EBOV transmission is proposed in Fig. 2.

Table 1.

Overview of Filovirus Detection in Wild Bat Species

Virus/Bat Species	Location	Detection Method	References
BDBV/
Eidolon helvum	Zambia	Antibodies	Ogawa et al. (2015)
EBOV/
Eidolon helvum	Ghana, Zambia	Antibodies	Hayman et al. (2010); Ogawa et al. (2015)
Epomops franqueti	Gabon, RC	Antibodies	Leroy et al. (2005); Pourrut et al. (2009)
Epomophorus gambianus	Ghana	Antibodies	Hayman et al. (2012)
Hypsignathus monstrosus	Gabon, RC	Antibodies	Leroy et al. (2005); Pourrut et al. (2009)
Micropteropus pusillus	Gabon, RC	Antibodies	Pourrut et al. (2009)
Mops condylurus	Gabon, RC	Antibodies	Pourrut et al. (2009)
Myonycteris torquata	Gabon, RC	Antibodies	Leroy et al. (2005); Pourrut et al. (2009)
Rousettus aegyptiacus	Gabon, RC	Antibodies	Pourrut et al. (2009)
R. leschenaultia	Bangladesh	Antibodies	Olival et al. (2013)
RESTV/
Acerodon jubatus	Philippines	Antibodies	Jayme et al. (2015)
Miniopterus schreibersii	China, Philippines	Antibodies, RNA	Jayme et al. (2015); Yuan et al. (2012)
Chaerephon plicata	Philippines	RNA	Jayme et al. (2015)
Cynopterus brachyotis	Philippines	RNA	Jayme et al. (2015)
Cynopterus sphinx	China	Antibodies	Yuan et al. (2012)
Eidolon helvum	Zambia	Antibodies	Ogawa et al. (2015)
Hipposideros Pomona	China	Antibodies	Yuan et al. (2012)
Miniopterus australis	Philippines	RNA	Jayme et al. (2015)
Myotis ricketti	China	Antibodies	Yuan et al. (2012)
Pipistrellus pipistrellus	China	Antibodies	Yuan et al. (2012)
Rousettus amplexicaudatus	Philippines	Antibodies	Taniguchi et al. (2011)
R. leschenaultia	China, Bangladesh	Antibodies	Olival et al. (2013); Yuan et al. (2012)
SUDV/
Eidolon helvum	Zambia	Antibodies	Ogawa et al. (2015)
TAFV/
Eidolon helvum	Zambia	Antibodies	Ogawa et al. (2015)
MARV and RAVV/a
Eidolon helvum	Zambia	Antibodies	Ogawa et al. (2015)
Epomops franqueti	Gabon, RC	Antibodies	Pourrut et al. (2009)
Hypsignathus monstrosus	Gabon, RC	Antibodies	Pourrut et al. (2009)
Miniopterus inflatus	DRC	RNA	Swanepoel et al. (2007)
Rhinolophus eloquens	DRC	Antibodies, RNA	Swanepoel et al. (2007)
Rousettus aegyptiacus	DRC, Gabon, RC, Uganda	Virus isolation, antibodies, RNA	Amman et al. (2014); Pourrut et al. (2009); Swanepoel et al. (2007); Towner et al. (2009)
LLOV/
Miniopterus schreibersii	Spain	Antibodies, whole-genome sequencing	Negredo et al. (2011)

^aN.B.: Both MARV and RAVV variants have been detected by Swanepoel et al. (2007) via PCR, but in cases where antibodies were used, due to antibody cross-reactivity, the subspecies variant is not always possible to determine

Fig. 2.

Model for the proposed transmission cycle of EBOV. (A) Reservoir bat species are thought to infect nonhuman primates such as Gorilla gorilla and Pan troglodytes. It has been widely hypothesized that bats directly infect humans through an unknown mechanism, although bites may be a possible route of transmission. Consumption of nonhuman primate bushmeat has been well documented as a source of past EBOV outbreaks. (D) Human-to-human transmission occurs via direct contact with bodily fluids and through airborne droplets. Fomites and aerosols are thought to be less probable routes of transmission.

RESTV was isolated from domesticated pigs (Sus scrofa) in the Philippines in 2008 (Barrette et al., 2009). More recently, RESTV RNA has been found in pigs near Shanghai, China (Pan et al., 2014), but domesticated swine are more likely to serve as intermediate or amplifying hosts rather than as the reservoir (Barrette et al., 2009; Pan et al., 2014). Pigs experimentally infected with RESTV remain asymptomatic (Marsh et al., 2011). This stands in contrast to experimental EBOV infection in pigs, which can cause severe respiratory symptoms and viral shedding (Kobinger et al., 2011). RESTV causes severe disease in nonhuman primates (Jahrling et al., 1996), but is not known to cause serious disease in humans (CDC, 1990a,b). As with EBOV, there is evidence that RESTV is also present in bats (Jayme et al., 2015; Taniguchi et al., 2011; Yuan et al., 2012) (Table 1).

3.2. Marburgviruses

Like EBOV and SUDV, members of the Marburgvirus genus cause disease with high rates of lethality in human and nonhuman primates. MARV and RAVV variants have been isolated from Egyptian fruit bats (Rousettus aegypticus), appearing to be a major natural reservoir for marburgviruses (Amman et al., 2014; Pourrut et al., 2009; Towner et al., 2009) (Table 1). This is supported by the observation that seasonal changes in MARV circulation among R. aegypticus correlate with the risk of human infection (Amman et al., 2012). R. aegypticus have also been shown to orally shed MARV when experimentally infected, despite showing no signs of overt morbidity (Amman et al., 2015). This may indicate that MARV has evolved to proliferate in Rousettus host populations and that bat bites may be a route of zoonotic infection. Besides fruit bats, insectivorous bats, such as Miniopterus inflatus and Rhinolophus eloquens, have also been implicated as carriers (Swanepoel et al., 2007). Like the ebolaviruses, marburgviruses have also been identified in the urban-dwelling straw-colored bat (Eidolon helvum) (Ogawa et al., 2015).

3.3. Cuevaviruses

LLOV, the only known cuevavirus, was first identified after a spate of bat die-offs on the Iberian Peninsula. In the Lloviu Cave in Northern Spain, colonies of Schreiber’s bats (Miniopterus schreibersii) and greater mouse-eared bats (Myotis myotis) yielded numerous carcasses available for microbiological and toxicological analysis (Table 1). Although no gross pathology was observed, histopathological signs of viral pneumonia were present. All the dead M. schreibersii were positive for LLOV RNA, whereas no LLOV was detected in healthy M. schreibersii taken from the same location (Negredo et al., 2011). Although little else is known about LLOV, its pathogenicity in M. schreibersii may suggest the existence of an alternative reservoir host.

4. FILOVIRAL PARTICLES

4.1. Morphology

Filovirus particles are long and filamentous. The particles may form elongated rods or curve one or more times, giving them a string or torus-like appearance (Fig. 3B). These distinctive shapes originally gave rise to the name Filoviridae, after the Latin filum, meaning thread. Although the virion filaments have a consistent diameter of about 80 nm, the length of particles is variable. The median virion length for an infectious particle has been determined to be about 800 and 1000 nm for MARV and EBOV, respectively (Geisbert and Jahrling, 1995). All seven structural filovirus proteins are incorporated into the virion as shown in Fig. 3A. Only the soluble glycoprotein (sGP) and the small soluble glycoprotein (ssGP) of ebolaviruses are entirely nonstructural and are discussed in further detail later. From these protein–protein, protein–RNA, and protein–membrane interactions, the complex filovirus morphology is generated. The nucleoprotein (NP) and virion protein (VP) 30 interact strongly with RNA; this complex together with the polymerase L and VP35 forms the ribonucleoprotein complex (RNP). This, in turn, is enclosed by an outer envelope derived from the host-cell plasma membrane, lined with the matrix protein VP40. Protruding from the envelope surface are membrane-anchored glycoprotein trimers, studded across the virion. A brief summary of protein function is provided in Table 2.

Fig. 3.

The filovirus particle. (A) Schematic of filovirus particle. GP trimers are embedded in the membrane, projecting spike-like structures away from the virion. The matrix underneath the membrane is composed primarily of VP40. The ribonucleoprotein complex (RNP) is composed of L, NP, VP30, and VP35. (B) EBOV electron micrographs. The left panel is a TEM image of EBOV. The virus is budding from the cellular membrane, forming pleiomorphic virions outside the cell (scale bar = 1 μm). The right panel is a 2D cryo-EM image of EBOV. Note the thread-like turns characteristic of filovirus morphology (scale bar = 100 nm).

Table 2.

Overview of Filovirus Proteins

Genome Position	Protein	Protein Functions	Molecular Weight (kDa)
1	Nucleoprotein (NP)	Major nucleoprotein, component of the RNP	90–104
2	Virion protein 35 (VP35)	Polymerase cofactor, component of the RNP, interferon antagonist	35
3	Virion protein 40 (VP40)	Matrix protein, virion assembly, budding, interferon antagonist for MARV only	35–40
4	Glycoprotein (GP)	Viral entry, receptor binding, membrane fusion	150–170
	Soluble glycoprotein (sGP) (EBOV only)	Antigenic subversion, restores endothelial barrier function	50–55
	Small soluble glycoprotein (ssGP) (EBOV only)	Unknown	50–55
5	Virion protein 30 (VP30)	Minor nucleoprotein, component of the RNP	27–30
6	Virion protein 24 (VP24)	Nucleocapsid condensation, virion assembly, interferon antagonist for EBOV only	24–25
7	RNA polymerase (L)	RNA-dependent RNA polymerase, component of the RNP	~270

4.2. Genome Structure

All filovirus genomes comprise a negative-sense single-stranded RNA molecule, approximately 19 kilobases (kb) in length. Filoviruses generally encode seven genes, arranged in a consistent order: NP, VP35, VP40, glycoprotein (GP), VP30, VP24, and RNA-dependent RNA polymerase (L) (Sanchez et al., 1993). An exception is LLOV, which is believed to encode only six mRNA transcripts, including one bicistronic transcript that expresses both VP24 and L (Negredo et al., 2011). Each gene is flanked by conserved transcription initiation and termination sites with minimal variation (Bukreyev et al., 1995; Feldmann et al., 1992; Groseth et al., 2002; Ikegami et al., 2001; Sanchez et al., 1993). A single long intergenic region has been noted in both the Ebolavirus genus (between VP30 and VP24) and the Marburgvirus genus (between GP and VP30) (Crary et al., 2003; Mühlberger et al., 1992; Sanchez and Rollin, 2005). Other genes are either separated by short intergenic regions, comprising four to seven highly conserved nucleotides, or overlap (Bukreyev et al., 1995; Feldmann et al., 1992; Groseth et al., 2002). The overlapping genes vary between virus species and may be used to identify specific isolates (Sanchez et al., 1993). A schematic representation of various filovirus genome structures is provided in Fig. 4.

Fig. 4.

Filovirus genome organization. This schematic of four representative filovirus genomes depicts the arrangement of genes and intergenic regions. Notably, there is an appreciable diversity with regard to the number of gene overlaps. MARV exhibits only a single-gene overlap between VP30 and VP24, whereas LLOV exhibits four gene overlaps. SUDV, BDBV, and TAFV (not shown) have three overlaps, arranged in a pattern similar to EBOV. The sizes of the overlaps have been slightly exaggerated for clarity.

In addition to intergenic regions, there are also extragenic regions located on the 3′ and 5′ ends of all filovirus genomes. The 3′ end contains a leader sequence that is approximately 60 bp in length. The 5′ end contains a trailer sequence that varies in length between species, but may be as large as 676 bp in the case of EBOV. The terminal ends of the genome have conserved sequences with high complementarity that form stem-loop structures (Crary et al., 2003; Mühlberger et al., 1992; Sanchez and Rollin, 2005). RNA secondary structure in the 3′-leader region mediates binding specificity of the viral transcription factor VP30 and is required in EBOV for activating transcription (Biedenkopf et al., 2016; Schlereth et al., 2016).

Each gene contains a single open reading frame (ORF) apart from LLOV VP24/L, which is bicistronic, and GP of the ebolaviruses, which contain three overlapping ORFs (Ikegami et al., 2001; Mehedi et al., 2011; Negredo et al., 2011). Filovirus ORFs are flanked by untranslated regions, ranging in length from 57 to 684 nucleotides (Groseth et al., 2002; Ikegami et al., 2001; Sanchez and Rollin, 2005; Towner et al., 2006). The presence of long untranslated regions is an unusual characteristic of the filovirus genome, a feature that is only shared by the henipaviruses of the Paramyxoviridae family (Wang et al., 2001). The precise importance of these untranslated elements is currently unknown, but their considerable size and distribution across all filoviruses may suggest a functional role.

4.3. Protein Structure and Function

4.3.1. Nucleoprotein Complex

Filoviruses contain major (NP) and minor (VP30) nucleoproteins which interact with viral genomic RNA to form the primary structure of the RNP (Elliott et al., 1985) (Fig. 3). When expressed in mammalian cells, recombinant NP forms cytoplasmic inclusion bodies and nonspecifically associates with host-cell RNA, forming helices (Noda et al., 2010). Moreover, 450 amino acids located on the N-terminus of EBOV NP have been shown to facilitate spontaneous assembly of NP tubes that may form a type of scaffolding for the RNP (Noda et al., 2010; Watanabe et al., 2006).

VP30 also directly binds RNA and preferentially binds viral genomic RNA at a stem-loop structure near the leader sequence. VP30 has two homo-oligomerization domains and forms hexamers via an N-terminal domain (Hartlieb et al., 2003). The crystal structure of the C-terminal region has been solved, showing a role in transcription and nucleocapsid association (Hartlieb et al., 2007). VP30 also possesses a zinc-finger motif that is highly conserved across all filoviruses and binds to RNA more readily in the presence of Zn²⁺ (Modrofet al., 2003). In terms of structural importance, it has been suggested that VP30 acts as a bridge between NP and L in the RNP (Groseth et al., 2009). In vitro studies using an artificial replication system originally showed that coexpression of NP, VP35, and L is necessary and sufficient for transcription and genome replication of MARV, whereas VP30 was additionally required for EBOV transcription (Mühlberger et al., 1998, 1999). More recently, the crystal structure of the NP–VP30 complex has been solved and the complex was shown to be essential for the transcription of EBOV genomic RNA with the binding strength modulating the rate of RNA synthesis (Kirchdoerfer et al., 2016). When expressed alone in mammalian cells, VP30 does not form inclusions like NP, but is distributed throughout the cytoplasm (Groseth et al., 2009). EBOV VP30 also serves as a trans-acting factor for RNA editing of the GP gene (Mehedi et al., 2013). EBOV VP30 has been shown to act as a phosphorylation-dependent transcription activator and replication inhibitor (Biedenkopf et al., 2013), whereas MARV VP30 appears to lack this property (Mühlberger et al., 1998).

4.3.2. Polymerase Complex

The L protein, an RNA-dependent RNA polymerase, is the enzymatic component of the RNP, for which VP35 is an essential cofactor (Mühlberger et al., 1992). When expressed alone in mammalians cells, the L protein forms small perinuclear inclusions. L does not directly interact with NP, and VP35 has been identified as the bridging factor between these two RNP components in vitro (Groseth et al., 2009). Both the template binding and catalytic sites of the polymerase complex have been identified (Feldmann et al., 1992). As might be expected, the enzymatically active portions of the L protein are highly conserved to maintain appropriate polymerase functions. However, certain portions of the sequence, especially near the C-terminus, exhibit some interspecies variation. For instance, the length of the MARV L gene (about 7.7 kb) is substantially longer than the EBOV (6.8 kb) or LLOV L gene (6.6 kb) (Fig. 4).

4.3.3. Matrix Protein

The matrix protein VP40 is the most abundant protein in the filovirus particle (Han et al., 2003). VP40 does not possess transmembrane domains, but exhibits a strong affinity for membranes and associates with the viral envelope. VP40 directly facilitates the envelopment of the RNP by the cellular membrane and is essential for viral budding (Jasenosky and Kawaoka, 2004). VP40 is the primary structural component of the matrix and forms hexameric and octameric ring structures in vitro (Jasenosky et al., 2001; Timmins et al., 2003). Oligomerization has been shown to be a prerequisite for the budding activity of VP40 (Hoenen et al., 2010b). Recently, biochemical and biophysical studies have demonstrated an intrinsic ability of VP40 to penetrate into lipid bilayers and initiate membrane curvature leading to virus particle formation (Adu-Gyamfi et al., 2012; Soni et al., 2013).

4.3.4. Viral Surface Spike

All filoviruses express the trimeric structural GP on the particle surface that is integral to viral entry and pathogenesis. GP is the best characterized of all the filovirus proteins, in part because of its functional importance and high immunogenic potential. To assemble into the functional GP, a series of processing steps must occur. First, ebolavirus GP transcripts must undergo transcriptional editing resulting in what is called the “8A” or “full-length” transcript (Sanchez et al., 1996; Volchkov et al., 1995). This unique feature is not present in marburgviruses. For all filovirus GPs, this full-length transcript gets translated into an ~680 amino acid (aa) long protein, representing the uncleaved precursor GP0. While in the ER, the N-terminal signal peptide is cleaved off cotranslationally and intramolecular disulfide binding occurs at key cysteine residues. As GP is transported through the ER and Golgi, it gets heavily glycosylated with N-linked and O-linked glycans that potentially mask antigenic epitopes in particular in the mucin-like domain (Reynard et al., 2009; Volchkov et al., 1998b). Next, GP0 is moved to the trans Golgi, where the cellular protease furin (or furin-like proteases) cleaves GP0 at a conserved cleavage site into the amino-terminal GP1 (approx. 470 aa) and carboxyl-terminal GP2 (approx. 175 aa) (Volchkov et al., 1998a,b). Interestingly, this cleavage event is not essential for virus replication or viral entry (Neumann et al., 2002; Wool-Lewis and Bates, 1999). GP1 contains a divergent mucin-like domain near the C-terminus and a region with several conserved cysteine residues near the N-terminus; it also contains the receptor-binding domain. GP2 also contains conserved cysteine residues, as well as a fusion peptide, heptad repeats, a Type 1 transmembrane domain (Feldmann et al., 1991; Volchkov et al., 1998b), and a pattern homologous to the conserved immunosuppressive motif found in oncogenic retroviruses (Volchkov et al., 1992). In vitro studies have shown that for pathogenic filoviruses (but not RESTV) this motif inhibits human CD4 and CD8 T cell cycle progression and suppresses the host cytokine response (Yaddanapudi et al., 2006). Postcleavage, the polypeptides are oriented such that the N-terminus ends appose one another, comprising GP1,2 heterodimers connected by means of a disulfide bridge (Malashkevich et al., 1999). Three GP1,2 heterodimers trimerize to form a GP peplomer, the spiked protrusions on the surface of the virion. This trimerization is facilitated by the coiled coils of heptad repeat sequences present on GP2, and the trimeric GP is embedded in the envelope membrane via the transmembrane domain in GP2 (Weissenhorn et al., 1998a,b). While the majority of GP is membrane bound, it has been shown that shed forms of GP exist: shed GP1 due to disulfide bond instability (Volchkov et al., 1998a) and GP1,2ΔTM, the trimeric GP without the transmembrane domain due to cleavage by the metalloprotease TACE (Dolnik et al., 2004). Although all filovirus GPs are presumed to undergo the processing described earlier, little data exist for LLOV GP (Maruyama et al., 2014). In recent years, much has been learned about the role of GP, particularly regarding EBOV entry, which is discussed further in Section 5.

4.3.5. Nonstructural Glycoproteins

One distinct feature of ebolaviruses that is lacking with all marburgviruses is transcriptional editing of the GP ORF and the resulting production of three different proteins: full-length GP (8A), soluble GP (sGP, 7A), and small, soluble GP (ssGP, 6A, or 9A). The N-terminal region (approx. 300 aa) including the signal peptide of all three proteins is identical; however, the C-terminal sequences are unique. Like GP0, pre-sGP is cleaved at the furin cleavage site resulting in sGP and delta-peptide (Volchkov et al., 1999; Volchkova et al., 1999). For delta-peptide a modulatory role has been proposed by inhibiting virus entry into target cells (Radoshitzky et al., 2011). Both sGP and ssGP form parallel homodimers via disulfide bonding at cysteine residues (Barrientos et al., 2004; Mehedi et al., 2011). Regarding protein function, sGP is thought to play a role in pathogenesis through multiple mechanisms. First, sGP has the antiinflammatory property of restoring endothelial barrier function, which is antagonized by GP1,2 and VP40 (Wahl-Jensen et al., 2005). Second, epitopes common to both sGP and structural GP allow sGP to competitively bind host antibodies in a process termed “antigenic subversion” (Cook and Lee, 2013; Mohan et al., 2012). This potential evasion of the host immune response may, in part, be responsible for the severity of disease, as patients with acute infection have large amounts of sGP circulating in the blood (Sanchez et al., 1999). However, in the guinea pig animal model, elimination of sGP expression using reverse genetics did not significantly attenuate virulence and lethality of EBOV infection (Hoenen et al., 2015b). The function of ssGP is not understood, but unlike sGP, ssGP does not restore endothelial barrier function (Mehedi et al., 2011).

4.3.6. Interferon Antagonists

Both ebolaviruses and marburgviruses have been found to potently antagonize the host interferon (IFN) response through similar as well as distinct mechanisms. VP35 and VP24 have been identified to interfere with the host innate immune response during EBOV infection; VP35 and VP40 have been associated with this function for MARV.

EBOV and MARV VP35 antagonize the host innate immune response by actively blocking IFN through interaction with IFN regulatory factor 3 (IRF3) and IRF7 (Basler et al., 2003; Prins et al., 2009). Cocrystal structures of the EBOV VP35 RNA-binding domain and double-stranded (ds)RNA revealed that VP35 coats the dsRNA during replication, thereby circumventing pattern recognition receptors such as those found in the RIG-I pathway (Bale et al., 2013; Leung et al., 2010). Interestingly, VP35 is analogous to the phosphoprotein present in Pneumoviridae and Paramyxoviridae, but VP35 seems only weakly phosphorylated (Becker et al., 1998; Elliott et al., 1985). EBOV VP35 is distributed throughout the cytoplasm and forms a bridge structure, connecting NP and L, when expressed in vitro (Groseth et al., 2009).

EBOV VP24 is associated with the outside of RNPs in the virus particle and blocks the nuclear accumulation of tyrosine-phosphorylated STAT1 by binding to karyopherin α1, the nuclear localization signal for STAT1 (Leung et al., 2006). Additionally, VP24 is able to inhibit IFN-stimulated phosphorylation of p38-α in certain cell lines, an immunosuppressive mechanism found in other viruses (Halfmann et al., 2011; Ishida et al., 2004). Furthermore, VP24 has been shown to localize near the cellular membrane and perinuclear region where it regulates viral RNA synthesis. An N-terminal domain confers the ability of VP24 to oligomerize in vitro (Han et al., 2003). It has been demonstrated that a knockdown of VP24 hinders nucleocapsid assembly (Mateo et al., 2011). More recently, VP24 was described to play a role in genome packaging (Watt et al., 2014) likely through interaction with NP (Banadyga et al., 2017).

MARV VP40 directly inhibits Jak1 and STAT tyrosine phosphorylation and, therefore, blocks efficiently IFN and interleukin 6 signaling (Valmas et al., 2010).

5. VIRAL LIFE CYCLE

5.1. Entry and Uncoating

The entry mechanism for filoviruses has been studied for several decades mainly using EBOV. While the GP is essential for the attachment, uptake, and receptor binding, this process does not follow a classical entry mechanism. Over the last decade, light has been shed on the rather complicated sequence of events critical for EBOV entry into target cells (Fig. 5). However, some steps remain controversial among filovirus researchers.

Fig. 5.

Filovirus replication cycle. (A) A mature virus particle outside the cell. (B) Virus attaches to the cell via GP binding to attachment factors, e.g., C-type lectins. (C) The virus is subsumed by the cell largely via macropinocytosis. (D) An endosome forms. GP is cleaved by cysteine proteases and engages its receptor NPC1 (purple), initiating membrane fusion. (E) The genome is deposited into the cytosol. Positive-sense transcripts and antigenomes are generated via the polymerase complex. (F) Viral RNA transcripts are translated by host ribosomes. (G) The replication complex uses the antigenomes as a template for progeny genomes. (H) Progeny genomes are formed. (I) VP40 and VP24 facilitate virion assembly and budding from the cell. Filovirus GP expressed on the cellular surface is incorporated into the virus envelope and the mature virus is released from the cell. The SEM image is of a Vero E6 cell with large amounts of EBOV budding from its cellular membrane (scale bar = 2 μm).

In a first step the virus particle attaches to the cell surface binding to C-type lectins like DC-SIGN (Alvarez et al., 2002; Simmons et al., 2003) and other proteins in the past mistakenly proposed as universal EBOV receptors. Binding to these molecules enhances viral entry at least in certain cell types, but they do not function as the EBOV receptor. After the virus particle is attached to the cell surface, it is taken up mainly via macropinocytosis (Aleksandrowicz et al., 2011; Nanbo et al., 2010; Saeed et al., 2010), although some studies demonstrated that other pathways contribute to virus uptake and propose that multiple pathways may be used together (Aleksandrowicz et al., 2011; Hunt et al., 2012; Sanchez, 2007). Regardless, the virus particle ends up in the acidic environment of endosomes. There, it has been shown that proteolysis of the EBOV GP1 is required for viral entry. Specifically, a role for cathepsin B and a likely accessory role for cathepsin L (both endosomal host cysteine proteases) have been established in vitro (Chandran et al., 2005). In vivo experiments have shown that neither cathepsin B nor L knockout mice displayed a delayed disease phenotype (Marzi et al., 2012), indicating that in the absence of these specific proteases other, functionally similar proteases like thermolysin could cleave GP1 (Brecher et al., 2012). However, this cleavage event is essential to the entry mechanism as it exposes the structurally buried fusion loop which can now engage the filovirus receptor, the cholesterol transporter Niemann-Pick C1 (NPC1) (Carette et al., 2011; Côté et al., 2011). Specific sequences in the luminal domain C of NPC1 have been identified as essential for GP binding and viral entry (Krishnan et al., 2012). The crystal structure of the NPC1–GP binding complex has highlighted several regions of interest, such as the area of interaction between NPC1 and the GP α1 helix (Wang et al., 2016). This is in line with the finding that variations surrounding the GP α1 helix sequence have been shown to contribute to differences in host species tropism (Martinez et al., 2013; Ng et al., 2015; Urbanowicz et al., 2016).

Finally, cleaved GP1 binds to NPC1 and triggers fusion between the viral and endosomal membrane. Subsequently the viral RNP is released into the cytoplasm and production of progeny virus is initiated. It is believed that all filoviruses follow the entry mechanism described for EBOV as shown by Davey et al. (2017).

5.2. Transcription and Replication

The mechanism of filovirus replication has been studied in detail using artificial replication systems such as minigenomes (Mühlberger et al., 1998, 1999). All filoviruses encode the proteins required to form an RNA-dependent RNA polymerase complex, which produces a positive-sense antigenome that serves as a template for replication of the negative-sense genome. In both MARV and EBOV, it has been shown that NP, VP35, and L are necessary and sufficient to support genome replication. This is similar to paramyxovirus and rhabdovirus replication, which require only NP, P, and L proteins to be replication competent (Conzelmann, 2004; Whelan et al., 2004). Key protein interactions such as NP–VP35 and VP35–L have been established, highlighting the important structural role VP35 plays in the viral polymerase complex (Becker et al., 1998; Groseth et al., 2009). It has been shown that the proteins involved in genome replication change localization over the course of cellular infection: first they are distributed in cytoplasmic inclusions that are enlarged near the nucleus, then broken into smaller pieces, and subsequently localized near the plasma membrane (Nanbo et al., 2013). In EBOV, replication begins at a bipartite promoter region that obeys the rule of six and flanks the transcription initiation sequence for the NP gene (Weik et al., 2005).

Filoviruses encode seven monocistronic transcriptional units (genes), except for LLOV, which is thought to possess only six including one bicistronic unit. Each unit is associated with a corresponding transcriptional promoter region. As stated previously, filoviruses have overlapping ORFs. There is a surprising difference in the protein requirement for transcription in MARV and EBOV. For MARV, NP, VP35, and L are sufficient to initiate transcription (Mühlberger et al., 1998). In EBOV, however, a fourth protein, VP30, is an essential cofactor that acts to circumvent a hairpin loop in the noncoding 3′ region of the genome (Weik et al., 2002).

Over the past decade, much has been learned about the mechanisms responsible for regulating transcription and genome replication, specifically regarding VP30 and VP24. It is known that VP30 stabilizes VP35–L RNA binding. As a result, it has been hypothesized that the phosphorylation state of VP30 modulates the conformation of the RNP to form either a transcriptase or a replicase complex (Biedenkopf et al., 2013). VP24 is also believed to regulate RNA synthesis by affecting the conformation of RNP. Electron microscopy analysis of VP24 shows that it associates with the exterior surface of the RNP and that this association condenses RNP into the form found in viral particles (Beniac et al., 2012; Bharat et al., 2012). In vitro systems with overexpression of VP24 show inhibition of RNA synthesis (Hoenen et al., 2010b; Watanabe et al., 2007). Thus, it is believed that the presence of VP24 changes the RNP from a form compatible with transcription/replication to a packaging-competent form (Banadyga et al., 2017; Watt et al., 2014).

5.3. Virion Assembly

While viral RNA synthesis occurs in inclusion bodies (Hoenen et al., 2012; Nanbo et al., 2013), virion assembly occurs in the cytosol. When expressed in vitro, VP40 alone forms virus-like particles (VLPs). These VLPs exhibit smaller diameters (53–80 nm) than infectious particles (about 80 nm), indicating that other viral proteins contribute to virus morphology (Kolesnikova et al., 2004; Noda et al., 2002). VP40 is formed as a monomer and is subsequently oligomerized into hexameric and octameric rings by means of an N-terminal domain (Ruigrok et al., 2000; Timmins et al., 2003). The conformation of the metastable monomeric form of VP40 is determined by interactions with the lipid membrane, as mediated by the C-terminal domain (Jasenosky et al., 2001; Scianimanico et al., 2000). The hexameric form of VP40 results when the C-terminal domain extends via a flexible linker from the ring structure, allowing stable binding of VP40 (Dessen et al., 2000; Scianimanico et al., 2000). The octameric form depends on four homodimers binding RNA trinucleotides of the sequence 5′-UGA-3′ and cannot form in the absence of RNA (Gomis-Rüth et al., 2003; Timmins et al., 2003). In EBOV, VP40 octamerization is essential for infectivity (Hoenen et al., 2005, 2010a).

VP24 is also important for proper assembly as it plays a role regulating viral RNA synthesis. While VP24 was early on suggested to function as a minor matrix protein due to its location in the matrix space (Becker et al., 1998), recent electron microscopy studies have shown that VP24 is indeed associated with the outside of viral RNPs resulting in rigid RNP forms as observed in EBOV particles (Beniac et al., 2012; Bharat et al., 2012). These and other findings suggest that VP24 is essential for RNP production capable of replication, as it controls RNP condensation from a flexible form which can undergo transcription and replication to a rigid form that is packaged into virus particles (Banadyga et al., 2017; Bharat et al., 2012; Hoenen et al., 2006).

5.4. Budding

Several studies have identified the actin cytoskeleton to be important for the transport of virus particle components to the site of virus budding. Progeny RNPs travel from the inclusion bodies to the sites of budding using the actin fibers (Schudt et al., 2015). Actin has also been shown to direct the transport of VP40 to the budding sites (Adu-Gyamfi et al., 2012). The mature GP trimers are already embedded in the plasma membrane. VP40, the major driver for particle formation and membrane deformation, needs to oligomerize in order to form the proper particle matrix (Hoenen et al., 2010a). VP40 has the intrinsic ability to penetrate into lipid bilayers and initiate membrane curvature which is consistent with virus particle formation (Adu-Gyamfi et al., 2012; Soni and Stahelin, 2014).

6. EVOLUTION

The evolutionary origin of filoviruses is a matter of some controversy, as the timescale of viral evolution is widely disputed (Holmes, 2003; Patel et al., 2011; Sharp and Simmonds, 2011; Wertheim and Kosakovsky Pond, 2011). Some analyses estimate that the filoviruses emerged several thousand years ago (Carroll et al., 2013; Suzuki and Gojobori, 1997). Other estimates propose a much older emergence event on the order of several million years ago, as evidenced by filovirus-like elements incorporated into mammalian genomes, with Ebolavirus and Cuevavirus genera diverging from the Marburgvirus genus since the early Miocene (Taylor et al., 2010, 2011, 2014). If true, it remains unknown what effects filoviruses may have had on mammalian evolution in the distant past. A long-term coevolutionary relationship with bats has been hypothesized on the basis of selective pressure on NPC1 and the presence of endogenous viral elements (Ng et al., 2015; Taylor et al., 2011). Differences in the intraspecies diversity of the filoviruses have also been noted as measured by the time estimates for the most recent common ancestor (MRCA). Members of the Marburgvirus marburgvirus species exhibit considerable sequence diversity and are thought to have an MRCA about 700 years ago. This stands in contrast to Zaire ebolavirus and Reston ebolavirus, which are believed to have undergone genetic bottlenecks within the past several decades (Carroll et al., 2013). In any case, the modern evolutionary rate of filoviruses ranges from approximately 4.6 × 10⁻⁵ nucleotide substitutions/site/year for Sudan ebolavirus to 8.2 × 10⁻⁴ nucleotide substitutions/site/year for Reston ebolavirus (Carroll et al., 2013). The variation in this rate of change reflects a difference in selective pressure as it relates to the ecological niche of each virus species.

Regarding the pathogenic filoviruses, spillover events into human populations seem to occur infrequently. A genetically diverse filovirus population circulates in largely unknown natural reservoirs and seems to fluctuate over time (Amman et al., 2012; Ogawa et al., 2015). Some human outbreaks, such as the MARV outbreaks in Durba (1998–2000) and the EBOV outbreaks in Gabon and RC (2000–2004), have been characterized by multiple incursion events occurring simultaneously or over a relatively short period of time, respectively (Bausch et al., 2006; Leroy et al., 2004). Other outbreaks, such as the MARV outbreak in Angola, the EBOV outbreak in Kikwit (1995), and the SUDV outbreak in Gulu (2000 – 2001), yielded clinical samples that showed a surprisingly low amount of genetic diversity throughout the entire course of the outbreak (Rodriguez et al., 1999; Towner et al., 2004, 2006). This suggests that the outbreaks were due to a single zoonotic transmission, followed by low rates of viral evolution during limited transmission chains in humans.

The recent EBOV epidemic in West Africa (2013–2016) is also thought to be derived from a single zoonotic incursion followed by an unfamiliar sustained human-to-human transmission. The index case is believed to be a young boy living in Guéckédou, Guinea who is thought to have acquired EBOV from a local bat in December 2013 (Baize et al., 2014). Although the exact timeline and geographic distribution remain uncertain, it is believed that the West African strain diverged from Central African strains around 2004 (Gire et al., 2014). As a result of improved sequencing technology, it is increasingly feasible to monitor changes in viral genomes over the course of a single outbreak (Carroll et al., 2015). This makes it possible to easily identify species-specific adaptations and raises the question of whether filoviruses adapt or attenuate to humans over time. Regarding host adaptation, it has been shown in vitro that amino acid substitutions observed during the West African outbreak altered EBOV GP to increase tropism for human cells and reduce tropism for bat cells by effecting changes in N-glycosylation, the GP signal peptide, NPC1 binding, and various epistatic interactions (Bedford and Malik, 2016; Diehl et al., 2016; Urbanowicz et al., 2016). Regarding human-to-human transmission, some investigators have indicated that the specific changes in the EBOV Makona genome reflect adaptive evolution (Liu et al., 2015), whereas other groups have found no evidence for human adaptation (Azarian et al., 2015; Hoenen et al., 2015a; Olabode et al., 2015). Despite some initial concerns about an increased rate of mutation (Gire et al., 2014), extended analyses have suggested that the mutation rate of West African EBOV isolate Makona displays a similar mutation rate than those isolates causing previous Central African outbreaks (Hoenen et al., 2015a). Extensive human-to-human transmission, however, resulted in an accumulation of synonymous and nonsynonymous mutations with hitherto unknown importance.

7. FUTURE PERSPECTIVES

Although much has been learned about filoviruses over the past two decades, many important gaps in our knowledge remain. Filovirus ecology is still poorly understood and, with the exception of marburgviruses, no natural reservoirs have been identified. Therefore, selection of the appropriate species for surveillance is consequently challenging. Greater effort should be directed toward identifying the precise mechanism of zoonotic transmission, so public health practices to reduce potential exposure can be established. Thanks to an improved understanding of how GP affects immunogenicity and cellular tropism, vaccines and therapeutics may be specifically tailored to viral targets to lend the greatest efficacy. Similar studies need to be intensified to better understand the functions of other viral proteins for virus life cycle and pathogenicity to define new targets for intervention. Comparative analyses of pathogenic and nonpathogenic filoviruses will also be valuable for the purposes of evaluating isolate-specific health risks. Moving forward, it will also be important to continue the search for novel filoviruses, especially in East Asia, to evaluate their pathogenic potential. Understanding the broad evolutionary forces driving changes in filovirus sequence and host distribution will be instrumental in identifying optimal regions for filovirus surveillance.