Receptor-mediated cell entry of paramyxoviruses: Mechanisms, and consequences for tropism and pathogenesis

Research in the last decade has uncovered many new paramyxoviruses, airborne agents that cause epidemic diseases in animals including humans. Most paramyxoviruses enter epithelial cells of the airway using sialic acid as a receptor and cause only mild disease. However, others cross the epithelial barrier and cause more severe disease. For some of these viruses, the host receptors have been identified, and the mechanisms of cell entry have been elucidated. The tetrameric attachment proteins of paramyxoviruses have vastly different binding affinities for their cognate receptors, which they contact through different binding surfaces. Nevertheless, all input signals are converted to the same output: conformational changes that trigger refolding of trimeric fusion proteins and membrane fusion. Experiments with selectively receptor-blinded viruses inoculated into their natural hosts have provided insights into tropism, identifying the cells and tissues that support growth and revealing the mechanisms of pathogenesis. These analyses also shed light on diabolically elegant mechanisms used by morbilliviruses, including the measles virus, to promote massive amplification within the host, followed by efficient aerosolization and rapid spread through host populations. In another paradigm of receptor-facilitated severe disease, henipaviruses, including Nipah and Hendra viruses, use different members of one protein family to cause zoonoses. Specific properties of different paramyxoviruses, like neurotoxicity and immunosuppression, are now understood in the light of receptor specificity. We propose that research on the specific receptors for several newly identified members of the Paramyxoviridae family that may not bind sialic acid is needed to anticipate their zoonotic potential and to generate effective vaccines and antiviral compounds.

In the last decade, many new paramyxoviruses have been identified. These viruses, transmitted mainly through airborne routes, are similar to pathogens that have caused the loss of hundreds of millions of human lives (1,2). Animal paramyxoviruses continue to cause crippling economic losses when they spread in domesticated mammals and in poultry, they are threatening the extinction of certain species, and they often cause debilitating diseases and significant mortality in companion animals (2,3).
This review focuses on insights recently gained on the processes by which two of the paramyxoviruses most relevant for human health, measles virus (MeV) 3 and Nipah virus (NiV), infect their hosts. Whereas most paramyxoviruses enter epithelial cells of the airway using sialic acid as a receptor, causing mild disease, MeV and NiV cross the epithelial barrier and cause severe disease. The receptors that allow entry of these viruses and the cells and tissues that support their growth were only recently unambiguously identified and characterized, providing insights into viral tropism and mechanisms of pathogenesis.

The Paramyxoviridae
A few months ago, growth of the Paramyxoviridae family to 72 members prompted the International Committee on Virus Taxonomy to restructure it into four subfamilies and 16 genera (4). Rather than illustrating the new classification, Fig. 1 focuses on the genetic relationships of the currently most relevant paramyxoviruses. These include MeV that still causes about 140,000 deaths annually (WHO Key Facts, https://www. who.int/news-room/fact-sheets/detail/measles) 4 and is targeted for eradication by the World Health Organization (2). Eradication has been successful for the animal morbillivirus rinderpest (RPV), which had major economic impact on cattle rearing and was lethal for wild species of even-toed ungulates (5). The emerging henipaviruses, Hendra virus (HeV) and NiV, have a broad mammalian host range, including humans and domestic animals, causing severe and often fatal respiratory and neurological diseases. High case fatality rates and a lack of approved therapeutics or vaccines have earned these viruses This work was supported by a Mayo Clinic Center for Biological Discovery grant (to R. C.). The authors declare that they have no conflicts of interest with the contents of this article. 1 Supported by the Mayo Clinic Graduate School of Biomedical Sciences. 2 To whom correspondence should be addressed. E-mail: cattaneo.
roberto@mayo.edu. the highest biosafety classification (level 4). Medically relevant paramyxoviruses also include mumps virus (MuV) and the human parainfluenza viruses (HPIV1-4), which are among the most prevalent human viruses known. Several animal viruses are also shown in Fig. 1. Newcastle disease virus (NDV), an avulavirus, is an avian pathogen that can be destructive for the poultry industry. Four other animal viruses are especially relevant for fundamental research: parainfluenza virus 5 (PIV5) is a rubulavirus that has been instrumental to understanding the molecular biology of the family and the interactions with the cellular innate immune response. Sendai virus (SeV) is a respirovirus that infects mice, providing a convenient model for pathogenesis. Canine distemper virus (CDV) is a morbillivirus that infects many carnivores, including ferrets, and is used to characterize the pathogenesis of morbilliviruses. Cedar virus (CedV), a henipavirus isolated from bats, is important for the study of receptor interactions and virus emergence.
The Paramyxoviridae are enveloped negative-strand RNA viruses that share different characteristics with two other families of negative-strand RNA viruses (1). Their envelope glycoproteins have similar structure and function as those of the Orthomyxoviridae, including the influenza viruses. However, the Paramyxoviridae genome is nonsegmented, sharing a similar organization and gene expression strategy with the Rhabdoviridae like rabies virus and vesicular stomatitis virus. Before focusing on the mechanisms of paramyxovirus cell entry and on the consequences of receptor-specific cell entry for tropism and pathogenesis, we briefly review their genome structure and replication mechanisms. This is important because, whereas receptor recognition is the main determinant of paramyxovirus tropism, post-entry mechanisms are crucial for efficient virus spread.

Genomes and replication
Fig. 2 illustrates the RNA genome (top) and a particle (bottom) of MeV, a typical paramyxovirus (2,4). The MeV genomic RNA, as that of all negative-strand RNA viruses, serves two functions. First, it is the template for synthesis of mRNAs. And second, it is the template for the synthesis of the antigenome positive-strand RNA (shown in Fig. 2A). Negative-strand RNA viruses encode and package their own RNA polymerase, which transcribes mRNAs once the virus envelope has fused with the plasma membrane of the infected cell. Viral replication, which occurs after transcription, requires the continuous synthesis of viral proteins. Newly replicated antigenomes serve as template for amplification of negative-strand genomic RNAs, which are the templates for secondary transcription (7).
Genomes of Paramyxoviridae are about 15,000 -19,000 bases in length and contain six or more genes in a conserved order. The 15,894-nucleotide (nt) MeV genome begins with a 52-nt 3Ј region, the leader, and ends with a 40-nt region, the trailer. These control regions flank the six contiguous transcription units (genes), which are separated by three untranscribed nucleotides. For MeV there are six genes coding for eight proteins, in the order (positive strand): 5Ј-N-P/V/C-M-F-H-L-3Ј ( Fig. 2A). The P gene of paramyxoviruses uses overlapping reading frames to code for up to four proteins, P, C, V, and W. The polymerase (known as large protein, L) transcribes the viral genome with a sequential "stop-start" mechanism. It accesses the genome through an entry site located near its 3Ј end. It then transcribes the first gene (N) with high processivity; caps, methylates, and polyadenylates the N mRNA; and reinitiates P mRNA synthesis. The frequency of reinitiation is less than 100%, resulting in a gradient of transcript levels: N is transcribed at the highest levels and L at the lowest (8).
RNA genomes, which are protected by helically arranged N proteins, have characteristic size and shape: 190-Å diameter, 49.5-Å pitch, and about 1 m in length for MeV (9). Recent structural studies have revealed how the 3Ј end of the genomic RNA becomes accessible to the viral polymerase complex (10), constituted of one copy of the L polymerase and four copies of the P co-factor (11). Another replication complex component, the C protein, interacts with P and improves polymerase processivity and accuracy (12,13). The third and fourth proteins expressed by the P gene, V and W, are nonstructural proteins that control innate immunity by antagonizing both type I interferon signaling and interferon production (14 -18). In addition, these proteins, which are post-entry determinants of virus tropism, may modulate tissue-specific virus gene expression (14).

Envelope
Enveloped particles of Paramyxoviridae have been observed to be pleomorphic or spherical (Fig. 2B), depending on the methods used for purification. MeV particles are particularly heterogeneous in size. Their diameter is in the 120 -300-nm range, but they can vary from 100 to 1000 nm in length, implying that their cargo volume differs. Indeed, large particles can contain multiple genomes, as deduced initially from sedimentation and UV inactivation studies and then confirmed by genetic complementation experiments (19). The envelope includes two glycoproteins, the tetrameric attachment protein and the trimeric fusion (F) protein (Fig. 2C). The F-protein of Paramyxoviridae causes membrane fusion at neutral pH, although exceptions to this rule have been described (6). The F and H oligomers form "spikes" that extend ϳ8 -12 nm from the surface of the particle membrane (20). The matrix (M) protein bridges the envelope with the nucleocapsid. In MeV, M is observed as a two-dimensional paracrystalline array associated with the inner leaflet of the plasma membrane (21).
The paramyxovirus attachment proteins that use neuraminic acid (␣2,3-linked sialic acid) as a receptor are known as hemagglutinin/neuraminidases (HN), because they also have receptor-cleaving neuraminidase activity. The attachment proteins of MeV and of the other morbilliviruses are named hemagglutinins (H), whereas those of the henipaviruses are called glyco-proteins (G). In truth, H is a misnomer. Originally, the attachment protein of an attenuated MeV strain was named H because cells infected with this virus absorb the erythrocytes of New World primate species. However, only the attachment protein of that attenuated strain, and not those of WT strains of MeV or other morbilliviruses, has this function. Eventually, it was shown that hemabsorption depends on binding a specific protein, CD46, rather than sialic acid (22,23). Binding to CD46, which is expressed on the erythrocytes of New World monkeys but not on those of humans (24), is thus implied in MeV attenuation (25, 26).

Receptors, tropism, and pathogenesis
Most paramyxoviruses that transmit through airborne routes (Fig. 3A) enter epithelial cells of the airways using protein-linked sialic acid as receptor and cause relatively mild disease. However, some paramyxoviruses cross the epithelial barrier and cause systemic, severe disease. We discuss the tropism and pathogenesis of three representative viruses: human parainfluenza virus (HPIV3), which causes acute respiratory disease but does not spread systemically; NiV, which spreads systemically and can cause lethal disease in humans (27); and MeV, which replicates in receptor-defined ecological niches, causing both acute respiratory disease and delayed immunosuppression (28) (Fig. 3B). Particles can contain multiple genomes, as represented by the three genomes in this particle. C, enlarged representations of the H tetramer and F trimer. The stalk of the H tetramer is modeled on the solved NDV HN structure (87,117). A green cylinder represents the membrane distal region of the stalk that was not in the solved structure. A blue star denotes a kink in the parallel four-helix bundle structure of the stalk. Four blue hexameric heads represent the six-bladed ␤-propellers of the receptor-binding domains. The heads are connected to the stalk by flexible dimeric linkers (green/blue) and four monomeric connectors (purple) (133,134). The MeV F trimer ectodomain structure (PDB code 5YXW) (154) is shown on the right with monomers represented as blue, purple, and orange. F-protein residues critical for receiving the fusion triggering signal from H are shaded red. Interrupted lines represent the unstructured segments of the ectodomains.

HPIV3 binds sialic acid and causes acute respiratory disease
HPIV3 represents the respiroviruses (Fig. 1). The HPIV3 attachment protein, HN, uses ␣2,3-linked sialic acid as a receptor. We do not know to what type of moieties this receptor is attached, but we think that it could be on many different proteins. This oligosaccharide is predominantly found on nasopharyngeal, bronchiolar, bronchial, and tracheal epithelial cells and alveolar pneumocytes. There is some indication of ocular distribution of ␣2,3-linked sialic acid as well, but very little is known about the exact levels (29). Tissue distribution of the receptor accounts for the symptoms caused by HPIV3 infections, which include not only mild respiratory disease but also bronchiolitis, bronchitis, and pneumonia in infants and children.
HPIV3 infection starts with virus entering through the nasal route ( Fig. 3A), where it infects nasopharyngeal cells. As virus replicates and amplifies, it moves on to infect bronchiolar, bronchial, and tracheal epithelial cells and eventually spreads to alveolar pneumocytes in the lungs (30). HPIV3 replication peaks within the first week of infection ( Fig. 3B, top). Different animal models have suggested that HPIV3 replication can start as early as 1 day after exposure and peaks between 2 and 5 days (31)(32)(33). HPIV3 is usually cleared by the immune system, and infection is resolved within 7-10 days; however, in immunocompromised individuals, it can lead to death (34 -36).

NiV binds ephrin-B2 or ephrin-B3 and can cause lethal systemic disease
NiV represents the henipaviruses (Fig. 1). NiV, like HeV, is endemic in certain pteropid fruit bat species, which serve as reservoir hosts (37). NiV and HeV have an exceptionally broad species tropism, and natural or experimental infection has been documented to span six mammalian orders. Affected species include pigs, horses, cats, dogs, guinea pigs, mice, hamsters, ferrets, squirrel monkeys, African green monkeys, and humans (38). NiV causes severe and often fatal respiratory and neurological diseases. The NiV attachment protein, G, utilizes ephrin-B2 and ephrin-B3, as receptors for cell entry. Ephrin-B2 is Middle, NiV infects lung epithelial cells, spreads to the vascular system, and, during late stages, infects different organs, causing multiple-organ failure. Relapse-encephalitis may occur in the brain up to 13 months post-infection. Bottom, MeV infects alveolar macrophages and dendritic cells, which transfer the virus to draining lymph nodes. After extensive replication in lymphatic tissues, MeV spreads to the upper-airway epithelia. Rarely, MeV infects the brain, and its persistence can cause lethal disorders years after acute infection. JBC REVIEWS: Receptor-mediated cell entry of paramyxoviruses expressed in arteries, capillaries, and bronchial and pulmonary alveolar type II epithelial cells (39 -41). Ephrin-B3 is predominantly found in the central nervous system (CNS) and at lower levels in the vasculature (40,42). High CNS expression of NiV receptors may account for its neurological disease potential, which is often lethal.
NiV enters the host through the nasal or oronasal route, from respiratory secretions or contaminated food (Fig. 3A) (43)(44)(45)(46). It initially replicates in ephrin-B2-positive bronchial and pulmonary alveolar type II epithelial cells (43) and lung endothelial cells, as early as 2 days post-infection (47). It then enters the bloodstream by infecting arterial endothelium, arterial smooth muscles, and pericytes that also express ephrin-B2; in a nonhuman primate model, replication in blood vessels is observed starting 4 days post-infection (47). Within the first week of infection, NiV disseminates to different organs, including spleen, kidneys, heart, and liver, as documented in different animal models (Fig. 3B, center) (47)(48)(49)(50). NiV can also directly enter the CNS via the olfactory nerve and possibly the hematogenous route (44,52). Once inside the brain, it infects ephrin-B3-and/or ephrin-B2-expressing neurons and other parenchymal cells within the CNS, causing encephalitis. This eventually leads to neurological disease (51), multiple organ failure, and death of the host. Survivors of acute NiV infection may suffer a lethal relapse-encephalitis 8 -13 months after initial disease resolution (Fig. 3B) (53)(54)(55).

MeV binds two proteins, causing both respiratory disease and immunosuppression
MeV is a human morbillivirus, a genus that also includes several important animal pathogens (Fig. 1). Even though disease severity caused by different morbilliviruses varies, the underlying pathogenic mechanisms and clinical manifestations are very similar. The attachment protein of all morbilliviruses binds two receptors, the signaling lymphocytic activation molecule (SLAM) and nectin-4. SLAM is found on immune cells, namely immature thymocytes, activated and memory T cells, naive and activated B cells, macrophages, and dendritic cells (56 -59). Nectin-4 is expressed at high levels by epithelial cells of the nasopharynx, trachea, and epidermal keratinocytes (60) and localizes to the adherens junctions (61).
MeV enters the body through the nasal route and infects SLAM-positive alveolar macrophages and dendritic cells (Fig.  3A) (62,63). These cells ferry the infection through the epithelial barrier and spread it to the local lymph nodes (64,65). The cellular distribution of SLAM overlaps with the susceptibility of different cell types to WT MeV infection (66). Another argument for the central role of SLAM in morbillivirus tropism is that CDV and RPV also enter immune cells through the cognate SLAM protein (67). Indeed, genetically modified MeV and CDV unable to enter cells through SLAM are attenuated in primate and ferret models, respectively (68,69).
Massive amplification of MeV and the other morbilliviruses in lymph nodes and primary immune tissues sets the stage for synchronous, massive invasion of tissues expressing the morbillivirus epithelial receptor, nectin-4 (70,71). Contrary to initial assumptions, morbilliviruses enter epithelia from the basolateral side, being delivered there by infected immune cells (72,73). Indeed, nectin-4 is located on the basolateral side of epithelial cells, and airway infection is restricted to epithelia that express nectin-4, including the trachea and the upper respiratory tract (Fig. 3B, bottom) (70,71). Because the trachea is the anatomical location most useful to support particle aerosolization, this two-phase process of amplification in immune cells followed by massive, synchronous invasion of the trachea accounts for the extremely contagious nature of MeV infections (74 -77).
Weeks after contagion, MeV can cause rare neurological complications, like primary measles encephalitis, in about 1 of 1000 cases. Moreover, 5-10 years after resolution of the primary infection, persistent MeV infections cause lethal subacute sclerosing panencephalitis in about 1 in 10,000 cases, or possibly at a higher incidence when very young children are infected (78,79). A neural receptor accounting for these pathologies has been sought but not yet identified (80). Alternative mechanisms that may account for neuro-invasion include MeV delivery to the brain by SLAM-expressing infected immune cells and a newly discovered form of cytoplasm transfer occurring between nectin-4 -expressing epithelial cells and nectin-1expressing neurons (81).

Receptor binding
Cell entry of all paramyxoviruses is mediated by the concerted action of attachment protein (HN, H, or G) tetramers with F-protein trimers (Fig. 2C). After receptors bind one or more attachment protein heads, the heads move and change their conformation, leading either to a structural change in the stalk or the exposure of parts of the stalk, which elicits refolding of F-trimers (82)(83)(84)(85)(86)(87)(88). This causes fusion of the particle envelope with the plasma membrane, a process that occurs at neutral pH (76,89,90).
The attachment proteins of paramyxoviruses are type II membrane glycoproteins comprised of a short cytoplasmic tail, a membrane-spanning segment, and an ectodomain with a stalk and a six-blade ␤-propeller headgroup that contacts the receptors (Fig. 2C, left). The F-proteins are type I membrane glycoproteins that form trimers, initially folding to a metastable, prefusion conformation (Fig. 2C, right), and upon activation refolding to catalyze membrane fusion. Proteolytic cleavage of F into two subunits is required for fusion activity. The MeV F-protein is cleaved by furin-like proteases in a trans-Golgi compartment as it is trafficked to the cell surface (91). Whereas most paramyxovirus F-proteins are similarly processed prior to arrival at the plasma membrane, the henipavirus F-proteins reach the cell surface as inactive precursors. Henipavirus F-activation requires endocytosis, which allows F-cleavage by an endosome-resident protease, cathepsin L (92)(93)(94).
Receptor-binding affinities- Table 1  These striking differences in binding affinity are explained in part by how many receptors the different attachment proteins need to engage. Due to the abundance of sialic acid expressed on the cell surface as a terminal component of sugar chains, the HN low affinity translates into high virus-binding avidities (101). In contrast, only several H-protein interactions with the cognate receptors may suffice to promote strong binding of morbillivirus particles to cells. The extremely strong binding affinity of the henipavirus G-proteins to certain ephrin receptors (K d ϭ 0.1 nM for NiV G with ephrin-B2 and K d ϭ 0.24 nM for CedV with ephrin-B1) suggests that perhaps only a single receptor interaction may be sufficient to stabilize henipavirus particle binding to cells. The range of binding affinities exhibited by paramyxoviruses may reflect, in part, the relative abundance of their respective receptors.
Receptor-binding modes-Crystal structures of receptorbound attachment proteins, further validated by biochemical and functional studies, have revealed much about the receptorbinding modes of paramyxoviruses (100, 102-107). Fig. 4A presents a top view of the attachment protein heads of HPIV3, HeV, and MeV and illustrates their modes of receptor binding. The overall structure of the three attachment proteins is conserved, but important differences in receptor-binding geometries exist.
HPIV3 HN interactions with sialic acid occur toward the central funnel of the ␤-barrel (Fig. 4A, left) (102), within an active site that conserves essential catalytic residues found in other neuraminidases (108). Neuraminidase function is reduced at neutral pH, enabling attachment and entry, but is activated at lower pH, enabling sialic acid removal during particle budding.
The HeV G-protein residues interacting with ephrin-B2/-B3 are also located in the central pocket of the ␤-barrel (Fig. 4A,  center) (100, 103, 104). Interestingly, this binding mode is rem-iniscent of that of HN with sialic acid, and could have evolved from it, and completely different from the protein receptor using morbilliviruses. However, a closer analysis of the binding interface reveals a flexible, exposed loop of the ephrin receptors (G-H loop, Fig. 5) (96, 103, 104) that fits snugly into a cavity of the henipavirus G-binding pocket in an induced-fit lock-andkey mechanism (Fig. 5A) (42,100). This intimate interaction between the residues of the G-H loop and the G-protein receptor-binding cavity may account for the strong receptor-binding affinities.
Ephrins differ in their ability to serve as henipavirus receptors, primarily due to differences in their G-H loop sequences (97). CedV, the most recent henipavirus isolate, binds ephrin-B1, in addition to ephrin-B2, but not ephrin-B3 (109,110). The crystal structure of the CedV G-head domain in complex with ephrin-B2 revealed that the binding cavity is comprised of three pockets (Fig. 5B, P1-P3), whereas the HeV G-binding cavity has four pockets (Fig. 5C, P1-P4) (100). This structural alteration allows accommodation of residues with larger side chains in the pocket P3, accounting for the altered receptor specificity of CedV (100,111).
Whereas the mode of receptor binding to henipavirus G-proteins retains vestiges of sialic acid binding to HN, receptor binding to MeV H must have evolved independently from the original interaction localized to the central cavity: all three MeV receptors interact with one side of the ␤-barrel (Fig. 4A, right) (105-107). All three binding sites overlap, but nectin-4 and CD46 functionally interact with a groove between two ␤-sheets (␤4-␤5) (107), whereas SLAM does not penetrate the groove, but rather lies over it and interacts with residues in blades ␤5 and ␤6 (98).
Because all three receptors bind distinct but overlapping sites on the MeV H-protein, this precise location may be critical for function. This hypothesis was tested by positioning 6-histidine tags in exposed loops of all blades of the MeV H-protein head and testing their function by pulling on the tags utilizing specific membrane-bound antibodies. Indeed, only pulling on those tags located near the binding sites of the natural receptors triggered membrane fusion (86).

JBC REVIEWS: Receptor-mediated cell entry of paramyxoviruses
Altogether, these studies indicate that different types of receptors (proteins or carbohydrates) can trigger the paramyxovirus fusion apparatus by binding its attachment protein with vastly different affinities. Receptors contact the heads of the HN and G attachment proteins from the top, through the central funnel of the ␤-barrel, but contact the head of the H-protein through a lateral groove. Nevertheless, all input signals are converted to the same output, F-trimer triggering leading to membrane fusion.

Membrane fusion mechanisms
The cell entry processes of all paramyxoviruses evolved from the same core mechanism. These processes rely on modular, exchangeable components: the heads of the G and H attachment proteins can be exchanged, provided that the contact of their stalk with the cognate F-trimer is maintained (112,113). It is also possible to trigger membrane fusion by appending a foreign binding domain to the attachment proteins and pulling on it with its cognate receptor, as shown for MeV H (114, 115) and NiV G (116). However, triggering occurs only when the contact of the attachment protein stalk with F-trimers remains intact. Thus, the attachment protein stalk has a central role in membrane fusion triggering.
Structure and function of the attachment protein stalks-The stalks of the HN-, H-, and G-proteins have many similarities and some interesting differences. Crystal structure analyses of HN-stalks indicate that they are organized into parallel fourhelix bundles (117)(118)(119), and biochemical evidence suggests that the henipavirus G-and morbillivirus H-protein stalks organize into similar structures (82, 85, 87, 88, 120 -123). The PIV5 four-helix bundle stalk structure shows adjacent segments of 11-mer and 7-mer hydrophobic repeat regions, with a kink at their junction (Fig. 2C, left, star) (118). This structure has been used to model the MeV H-stalk, and mutagenesis experiments in both systems have reached similar conclusions, localizing the F activation site to a few amino acids surrounding the kink (87,118). Functional evidence for the requirement of a conformational change of the HN-and H-stalks for F-trimer  Fig. 2C for MeV H. In this HN-head atomic structure (PDB codes 4MZA and 4XJR) (153), residues interacting with sialic acid are indicated in orange. Center, HeV G (PDB code 2X9M) (155) residues contacting ephrin-B2 are indicated in purple. Right, H-protein contact sites for SLAM (blue residues) and nectin-4 (yellow residues) are indicated on the MeV H-head structure (PDB code 2ZB5) (156). Green residues bind both SLAM and nectin-4. B, side views of the tetrameric stalks represented by four cylinders. The stalks are comprised of a dimer of dimers (green and blue cylinders, with length proportional to number of aa). Arrows represent flexible linkers that connect to the globular head domains. The white stars indicate a kink in the parallel four-helix bundle organization of the stalks. Blue wavy line, cytoplasmic tail, shown for only one subunit. Gray box, plasma membrane. Center, thick red lines connecting yellow dots (Cys residues) indicate disulfide bonds that stabilize the HeV G-dimers or the tetramer. Right, thick red lines indicate disulfide bonds that stabilize the MeV H-dimer. JBC REVIEWS: Receptor-mediated cell entry of paramyxoviruses activation was obtained by cross-linking specific positions by engineered disulfide bonds, which inhibits function; removal of certain bonds by reduction restores function (87,118). Similarly, there is evidence that the henipavirus G-stalks modulate conformational changes (82).
On the membrane-distal side of the stalk, the interactions among the four subunits of the HN-, G-, and H-proteins have interesting differences (Fig. 4B). Certain HN-protein dimers are stabilized by one disulfide bond at the top of the stalk (102,119,124,125), whereas others, such as HPIV3 HN, lack this stabilization (Fig. 4B, left) (126, 127). On the other hand, both henipavirus G-proteins and the morbillivirus H-proteins contain two disulfide bonds at the top of the stalk for dimer stabilization (Fig. 4B, center and right). The G-protein has a third disulfide bond in the stalk that serves to stabilize the tetramer (Fig. 4B, center, C146) (128).
The mechanisms of signal transmission through the attachment protein also have interesting variations, as discussed below.
Membrane fusion by HN-proteins-Structural analyses of the NDV HN-stalk revealed extensive interactions with HN-heads: HN-heads are positioned along the stalk in a "four heads down" state likely to interfere with F-trimer interactions (117). Based on this structure, Jardetzky and Lamb (89) proposed a stalkexposure/induced-fit model of fusion triggering by HN-proteins. In this model, HN-proteins would be initially folded in the "four heads down" conformation, which would prevent premature triggering of the F-trimer during intracellular transport.
Structural analysis of another HN tetramer, namely the PIV5 ectodomain complexed with a glycan receptor, revealed another unanticipated conformation, named "two heads up/two heads down" (119). Evidence was presented suggesting that this conformation, which allows interaction of the F-protein from one side of the stalk, represents an intermediate before HN reaches a final "four heads up" conformation after receptor binding to both dimers.
This would expose the F-interacting segment at the top of the HN-stalk, allowing HN to contact F and trigger fusion via a putative induced fit mechanism (118). In this model, the stalk is JBC REVIEWS: Receptor-mediated cell entry of paramyxoviruses a "provocateur" of fusion; consistent with this model, headless HN-stalks can trigger fusion in the absence of receptors, albeit at low efficiency (83). Notably, the rearrangement of the HNheads is facilitated by the flexibility of the head-stalk linker segment (Fig. 4B) (129). Porotto and co-workers (20) suggest that, in addition to inducing rearrangement of the HN-heads about the stalks, receptors also promote clustering of HN/F complexes and that this additional step is required for HN to transmit the F-triggering signal.
Membrane fusion by G-proteins-Analyses of the NiV fusion process indicate that interactions of the F-trimer with both the G-head and the G-stalk occur prior to receptor binding (82,130,131). Therefore, Aguilar and colleagues (84) have proposed a "bidentate" model for G-protein fusion triggering. According to this model, the initial conformation of the globular heads atop the stalks prevents F-activation by the G-stalk (128). Similar to HN, expression of just the G-protein stalk is sufficient to trigger fusion (85).
In the bidentate model, receptor binding causes the G-protein head domain to undergo two sequential conformational changes, leading to the exposure of the membrane-distal end of the G-protein stalk that acts to trigger the F-protein conformational change (84). In both the HN and the G activation models, the top of the attachment protein stalk is a functional and structural linker between the head domain and the rest of the stalk, serving to propagate the F-triggering signal by undergoing conformational changes.
The central segment, which includes the kink at the junction of the 11-mer and 7-mer hydrophobic repeat regions, triggers F-trimer refolding (88,120,132). The hydrophobic hinge maintains the H-protein in an autorepressed state prior to receptor binding (134). The head connectors govern proper H-protein tetramerization (133). According to the model, membranebound receptors pull on the heads (Fig. 6B) (86). This causes refolding of the hinge region, followed by transduction of the signal down to the central segment of the stalk, which conformational change triggers F-trimer refolding and membrane fusion (Fig. 6C).
In this model, H initially folds into a conformation that precludes F-activation (122). This is analogous to the "four heads down" conformation observed for HN-proteins, although the initial H-conformation does not prohibit H-F interactions (135), allowing formation of H-F complexes without F-triggering (136). Analogously to HN-and G-proteins, expression of just the H-protein stalk is sufficient to trigger fusion (137).
Three variations on one cell entry mechanism-In summary, the cell entry mechanisms of all paramyxoviruses share one core process: upon receptor binding, attachment protein tetramers trigger F-trimer refolding and membrane fusion. The provocateur, bidentate, and safety catch models have many similarities, including movement of the heads upon receptor binding and involvement of the stalk in signal transmission.
The models also have interesting differences. The provocateur model accounts for structural evidence suggesting that the HN-protein heads physically hinder the interactions of their stalk with F-trimers; no such evidence was presented for the H-proteins. The safety catch model accounts for formation of H-tetramer and F-trimer complexes during transport to the cell surface; in contrast, HN-F complexes form only at the cell surface. Finally, the bidentate model combines elements of the other two models and proposes a sequence of conformational changes by which the G-heads elicit F-trimer refolding.
The dissociation constants from the cognate receptors also differ, ranging from subnanomolar for G-protein, to 20 -200 nM for H-protein, to micromolar for HN-proteins. Thus, viruses with HN-proteins must contact many receptors before their attachment is stabilized, whereas only a few receptor contacts are required to stabilize the attachment of viruses with G-proteins. The geometry of receptor binding also differs, with receptors pulling the G-and HN-heads from their central funnel but the H-heads from one side. Despite these differences, all input signals are converted to the same output, F-trimer refolding and membrane fusion.

Conclusions and future directions
The paramyxoviruses of deepest human disease concern include those that have evolved away from binding cells through sialic acid and have gained the ability to cross the epithelial barrier by binding specific proteins. Morbilliviruses have adapted to a complex ecological niche defined by two receptors. First, these viruses use SLAM to infect immune cells. They rapidly and massively reproduce therein and can reach all of the organs of the host within these cells. Morbilliviruses then target certain epithelia through a protein expressed on their basolateral side, nectin-4, which is expressed most abundantly in the tracheal epithelium. Additional virus amplification at a site facilitating aerosolization contributes to extremely efficient contagion.
In another paradigm for the study of viral disease, henipaviruses, including NiV and HeV, use different members of the ephrin receptor protein family to spread within their hosts. Ephrins are ubiquitous membrane proteins, and their high conservation among mammals facilitates henipavirus zoonosis and allows for broad host tropism. Ephrin-B2, the receptor shared by all henipaviruses isolated to date, is expressed in the respiratory epithelium (39), vascular endothelium (41), and neurons (138). This broad receptor distribution accounts for the multisystemic vasculitis and encephalitis observed during both HeV and NiV infections (139).
Fueled both by the identification of proven infectious agents and by computational analyses of sequences present in cells of different species, the Paramyxoviridae family more than doubled its members within a few years (3,4,109,140). New members include viruses that encode HN-proteins that interact with sialic acid and others that encode G-proteins that may bind protein receptors. Two of the newly classified taxa (narmoviruses and pararubuloviruses) are predicted to have protein JBC REVIEWS: Receptor-mediated cell entry of paramyxoviruses Figure 6. Morbillivirus fusion triggering mechanism. One H-tetramer and one F-trimer are shown. Only one monomer of each H-head dimer is shown for clarity. The MeV H-stalk was modeled based on the NDV stalk structure (PDB code 3T1E) (117). A green cylinder is used to represent the head-proximal region of the stalk, which acts as a spacer. Blue lines, the flexible linkers, which, together with the connectors (purple lines), link the stalk to the head domains. ϩ, location of the hydrophobic hinge; yellow circles, Cys residues. Blue star, kink in the stalk centered on a module (red helices) that is critical for F-triggering. The MeV F-trimer head domain crystal structure is presented with the three monomers indicated by different shades of gray (PDB code 5YXW) (154). Residues critical for receiving the triggering signal from H are shaded yellow and define a groove made by the interface of two adjacent F-monomers (132). A, the glycoprotein complex prior to receptor interaction. B, receptors bind and pull on the H-heads, leading to a conformational change centered on the hydrophobic hinge of the linker (circled black plus sign). This signal is transmitted down the stalk to induce a conformational change (blue star with black border) that triggers F-protein refolding. C, post-fusion, the F-trimer refolds, fusing the viral membrane with the plasma membrane. The post-fusion form of F is represented by the HPIV3 F post-fusion structure (PDB code 1ZTM) (151). receptors due to the absence of the neuraminic acid binding motif (4). Study of the receptor interactions and of the critical post-entry processes, such as innate immune control, will be crucial to understand the tissue and species tropism of these emerging paramyxoviruses.
For example, CedV is a henipavirus that is apathogenic in animals known to be susceptible to Nipah and Hendra disease. This can be attributed mainly to its inability to control the innate immune response (109,110,141,142), but probably also to its different receptor interactions (100,111). Thus, CedV, the only nonbiosafety level 4 henipavirus currently available, affords us with a model system to study the interplay between receptor interactions and innate immunity control and determine how these processes contribute to tropism and virulence of henipaviruses.
Finally, research on the newly identified paramyxoviruses is required to anticipate their zoonotic potential and to generate effective vaccines and antiviral compounds. Because fusion requires refolding of the F-protein, paramyxovirus antivirals can be developed to block this conformational shift. Indeed, in small-animal models, inhibitors developed to capture F in its prefusion state have been effective in preventing MeV entry and spread (143,144). In an alternative approach, small-molecule inhibitors of the viral polymerase can protect from lethal morbillivirus infection (145). Importantly, recent analyses of paramyxovirus tropism in natural hosts have identified both entry and postentry steps that result in virus attenuation. This knowledge will facilitate engineering of effective vaccines for currently emerging zoonotic viruses, as well as for pathogens with receptor specificities predestining them to cause zoonotic disease.