Expert Reviews in Molecular Medicine: http://www.expertreviews.org/
Accession information: (01)00345-3h.htm (shortcode: txt001dsl); 6 August 2001
Reprint/PDF version How to cite this article

Mechanisms of cell entry by influenza virus

Karen J. Cross, Laura M. Burleigh and David A. Steinhauer

A wide range of viruses, including many human and animal pathogens representing various taxonomic groups, contain genomes that are enclosed in lipid envelopes. These envelopes are generally acquired in the final stages of assembly, as viruses bud from regions of the membrane of the infected cell at which virally encoded membrane proteins have accumulated. The viruses procure their membranes during this process and mature particles ‘pinch off’ from the cellular membranes. Under most circumstances, initiation of another round of infection is dependent on two critical functions supplied by the envelope proteins. The virus must bind to cell-surface receptors of a new host cell, and fusion of the viral and cellular membranes must occur to transfer the viral genome into the cell. Enveloped viruses have evolved a variety of mechanisms to execute these two basic functions. Owing to their relative simplicity, studies of binding and fusion using enveloped viruses and their components have contributed significantly to the overall understanding of receptor–ligand interactions and membrane fusion processes – fundamental activities involved in a plethora of biological functions.

Expert Reviews in Molecular Medicine © Cambridge University Press ISSN 1462-3994

Author contact details

Influenza A viruses belong to the Orthomyxoviridae family of enveloped viruses. They contain negative-stranded RNA genomes with eight RNA segments that encode ten viral proteins. Initiation of infection by influenza requires binding of the virus to host cell-surface receptors, endocytosis, and fusion of the viral and endosomal membranes. The fusion process allows release of the viral genome into the cytoplasm, whereupon it migrates to the nucleus – the site of viral transcription and replication. For influenza, the protein responsible for both receptor binding and membrane fusion is haemagglutinin (HA), which, for most strains, is the most abundant glycoprotein on the surface of the virion. In addition to its binding and fusion functions, it is the primary target for neutralising antibodies. Figure 1 (fig001dsl) shows an electron micrograph of budded influenza virus particles, in which the densely packed HA and neuraminidase (NA) spikes are clearly visible protruding externally from the viral membrane.

Several X-ray crystal structures have been determined for HA, including the structures of the three principal molecular conformations that the protein assumes during the virus life cycle. These are the precursor structure, the structure of cleaved native HA present on the surface of infectious virions, and the structure of HA following the conformational changes involved in membrane fusion. HA structures have also been determined in complexes with receptor analogues and bound to neutralising antibodies. The knowledge of such structures and the wealth of biochemical and functional data have made influenza HA one of the most important models for structure–function studies of viral glycoproteins.

This article considers the two primary events at the initial stage of influenza infection – receptor binding and membrane fusion – and particularly the relevance of the HA structure for these processes. A recent review by Whittaker in Expert Reviews in Molecular Medicine (Ref. 1) broadly covers the entire life cycle of influenza from infection, through nuclear import, to virus release, and provides a solid foundation for this contribution. Another review on the role of HA in virus entry has also been published recently, and this provides a comprehensive background on the HA structural information that has been accumulated to date (Ref. 2).

Structure of HA
The human influenza virus A/Aichi/2/68 is the prototype of the H3 antigenic subtype viruses responsible for the pandemic of Hong Kong ’flu in 1968, and antigenically related viruses of this subtype continue to circulate in the human population to this day. The HA of this virus is the source of most of the high-resolution structural information for this glycoprotein and forms the focus of this review.

HA synthesis and expression
The mature polypeptide is a type 1 membrane glycoprotein of 550 amino acids with an amino (N)-terminal signal sequence, a transmembrane domain near the carboxy (C)-terminus and a short cytoplasmic tail. It is initially synthesised as a 76 kDa polypeptide using host cell transcription and translation machinery. Insertion of the protein into the membrane of the endoplasmic reticulum (ER), signal peptide cleavage and core glycosylation all take place co-translationally. Aichi HA has seven N-linked glycans attached to asparagine side-chains of the growing polypeptide. To fold properly, HA must be glycosylated and acquire its six intramolecular disulphide bonds. Several of the Aichi HA glycosylation sites are located near the N-terminus of the polypeptide and it has been shown that these sites direct the protein into the calnexin/calreticulin chaperone system to assist in its folding and oligomerisation prior to transport to the Golgi complex (Ref. 3). In the Golgi complex, further oligosaccharide trimming occurs, as well as association of the HA trimers with sphingolipid-cholesterol lipid microdomains referred to as detergent-insoluble glycolipid-enriched domains (DIGS) or lipid rafts (Ref. 4). Both the transmembrane domain (Ref. 5) and the palmitate moieties attached to the three cysteine residues in the cytoplasmic domain (Refs 6, 7) have been implicated in the affinity of HA for DIGS. HA, as well as the other viral membrane proteins NA and M2, are preferentially transported to the apical surface of polarised epithelial cells, where virus assembly and budding occur.

HA protein structure
The mature HA forms homotrimers of 220 kDa. Each monomer is initially present as a single polypeptide precursor (HA₀) that is subsequently cleaved into two subunits: HA₁ and HA₂. These subunits are linked by a single disulphide bond between residue 14 of HA₁ and residue 137 of HA₂. It is the cleaved form of HA that is present on the surface of infectious viruses (discussed later). Treatment of virus particles with the protease bromelain cleaves HA close to the viral membrane and solubilises the majority of the HA trimer intact (Ref. 8). The X-ray crystal structure of this bromelain-released HA ectodomain, referred to as BHA, was first reported in 1981 (Ref. 9). The HA trimer projects approximately 135Å from the viral membrane and has two distinct regions (Fig. 2; fig002dsl). The base of the cylindrical trimer is formed by a stem-like region that contains residues from both the HA₁ and HA₂ subunits, and its dominant feature is a core of HA₂ helical domains that form the ‘spine’ of the protein. This ‘spine’ is composed of 50-residue a-helices (one from each monomer) which associate as a triple-stranded coiled coil. Conformational changes in this coiled coil are required for membrane fusion (see below). A hydrophobic domain situated at the N-terminus of each HA₂ subunit is also critical for membrane fusion and, as such, is often referred to as the fusion peptide (see below). In the BHA structure, the fusion peptides are buried in the interior of the trimer within the membrane-proximal regions of the long helices where the coiled coil opens up (Fig. 2; fig002dsl). At the membrane-distal end of this fibrous stem of the molecule there is a large globular head formed entirely by residues from HA₁. The most notable feature of the head regions of each monomer is an eight-stranded antiparallel b-sheet structure. These membrane-distal domains contain the receptor-binding site and principal recognition sites for neutralising antibodies.

Receptor binding
Structure of the receptor-binding site
Evidence that influenza virus binds to oligosaccharides containing 5-N-acetyl neuraminic acid (sialic acid) attached to host cell-surface glycoproteins and glycolipids was accumulating as early as the 1940s and 1950s, when it was shown that influenza preparations were capable of agglutinating erythrocytes (Ref. 10) and that this activity could be reversed or inhibited by specific antiviral sera, by an enzymatic activity associated with the virus (provided by the NA), or by exogenously added bacterial NA (reviewed in Ref. 11). Elucidation of the structure of the HA ectodomain alone (Ref. 9), and subsequently in complexes with sialic-acid-containing receptor analogues (Refs 12, 13, 14, 15) provided considerable insight into receptor binding by HA. A shallow pocket at the membrane-distal tip of the globular HA domain of each monomer was proposed as a receptor-binding site when the structure of the Aichi HA was first determined (Ref. 9). It was noted that amino acids in this area were highly conserved, whereas surrounding residues had been more affected by antigenic drift as viruses evolved from year to year as a result of human immunity. Further evidence for this site as the binding pocket was obtained by comparing the amino acid sequences of HAs from viral mutants that preferentially recognised terminal sialic acids attached to galactose via a-(2,3) linkages with the sequence of HA from Aichi virus, which favours a-(2,6) linkages (Ref. 16). The mutants contained glutamine rather than leucine at HA₁ residue 226, which is situated at the left side of the proposed binding site. The membrane-distal pocket of conserved residues was subsequently confirmed as the binding site by determining the X-ray crystal structure of several different BHA complexes with sialic-acid-containing receptor analogues (Refs 12, 13, 14, 15).

Irrespective of which receptor analogue is crystallised with HA, the crystallography data indicate that the sialic acid portion of the molecule is superimposable among the structures. A schematic diagram of sialic acid in the HA binding pocket is displayed in Figure 3 (fig003dsl). The pyranose ring of sialic acid faces the base of the site. The axial carboxylate, the acetamido nitrogen and the 8- and 9-hydroxyl groups of the glycerol side-chain orient into the pocket and form hydrogen bonds with HA residues, whereas the 4-hydroxyl points out of the site into solution (Refs 13, 17). The base of the binding site is formed by HA residues Tyr98 and Trp153, which are part of a conserved hydrogen-bonded network of amino acids that also includes His183 and Tyr195. These residues might contribute to binding function in two ways: their interactions with one another might help support the architecture of the binding pocket, and three of these four residues directly contact sialic acid. Tyr98 and His183 both form hydrogen bonds with the glycerol moiety, and Trp153 is in van der Waals contact with both the glycerol C7 carbon and the acetamido methyl group. If the HA is viewed from a lateral perspective, residues 134–138 constitute the front of the binding site, with the hydroxyl group of Ser136 and main-chain atoms of residues 135 and 137 forming hydrogen bonds with sialic acid. The left side of the pocket is composed of residues 224–228, and the side-chains of Glu190 and Leu194 protrude down from a small a-helix at the membrane-distal ridge of the site to contact the receptor.

Mutagenesis of the binding site
The contributions made by specific conserved HA residues to binding have been assessed by mutagenesis (Ref. 18). Experiments analysing the binding of expressed HAs to human erythrocytes support the concept that the network of hydrogen bonds linking residues 98, 183, 195 and 153 might be instrumental for maintaining the structural integrity of the site. HA mutants containing a Trp to Ala substitution at position 153 (Ref. 18) or a His to Ala substitution at position 183 (D.A. Steinhauer et al., unpublished) are not expressed at the cell surface, possibly because of a defect in protein folding. All HAs with residues Tyr98, His183, Tyr195 and Trp153 substituted individually for Phe are transported to the plasma membrane but, in each case, the sialic-acid-binding activity is low. Among other mutants analysed, it is notable that binding is decreased by substitutions at Ser136, and that a Leu to Ala change at residue 194 almost completely abrogates binding.

Several of the mutant HAs have been successfully incorporated into infectious influenza viruses using reverse genetics (Ref. 18). Notable among these is the Tyr to Phe98 (Y98F) mutant, as this displays very limited binding activity (5%) relative to wild-type HA and agglutinates erythrocytes very poorly. Interestingly, however, this mutant is capable of replicating to levels comparable with wild-type virus in both Madin–Darby canine kidney (MDCK) cells and embyonated chicken eggs – the two most common substrates for the growth of influenza viruses. It is possible that in these hosts high densities of cell-surface sialic acid compensate for the poor HA binding activity of the mutant viruses. To examine this, wild-type and Y98F mutant viruses were propagated on a mutant MDCK cell line that contains reduced levels of cell-surface sialic acid (Ref. 19). In this case, whereas wild-type virus yields were reduced only marginally with respect to the virus titres obtained on normal MDCK cells, the yields observed for the Y98F mutant were several logs lower. These experiments support the concept that virus attachment to host cells involves multiple HA–receptor interactions that act cooperatively, and that productive binding requires an appropriate balance between binding affinity and receptor density. Cooperativity in binding was suggested by the observation that the affinity of HA for monovalent sialic acid derivatives is relatively low, with a dissociation constant of 2–3 mM, in contrast with the tight binding of virus particles to cells (Ref. 20). Approaches for the design of inhibitors of influenza binding to host cell-surface receptors therefore involve either multivalent sialic acid analogues or monovalent sialosides with features that allow them to bind HA with higher affinity (Ref. 14).

Receptor-binding specificity
As mentioned above, the HAs of different virus strains can have distinct binding properties depending on the type of linkage by which sialic acid is attached to the penultimate galactose of the oligosaccharide receptor. In general, human strains bind preferentially to a-(2,6)-linked sialic acid whereas avian strains favour a-(2,3)-linked sialic acid (Ref. 2). The structural basis for receptor specificity has not yet been adequately resolved. The residues most frequently implicated in this phenomenon are HA₁ 226 and 228, although other residues can be involved. Strains favouring a-(2,3) linkages usually have glutamine at position 226 and glycine at 228, and those that preferential bind to a-(2,6)-linked sialic acid generally have leucine and serine at these positions, but these tendencies are by no means absolute. Although the position of sialic acid in the pocket is nearly superimposable regardless of its linkage type, the conformation of the sugar chain attached to it varies. Crystallography studies on HA complexes with pentasaccharides (Ref. 12) have shown that a-(2,6)-linked polysaccharides adopt a folded conformation and protrude from the right side of the pocket proximal to Leu194. Polysaccharides containing an a-(2,3) linkage were observed as an extended conformation that exits the binding site from the left side.

The linkage type of cell-surface carbohydrates at the principal sites of virus replication might be an important factor in determining species specificities of influenza viruses. In this regard it is notable that linkage-specific lectins have been used to show that a-(2,6)-linked, but not a-(2,3)-linked, sialyloligosaccharides are detectable on the surface of epithelial cells of the human trachea (Ref. 21). Similar experiments have been carried out on tissue samples from the intestine of ducks and the trachea of pigs (Ref. 22), which are the main sites of virus replication during natural infection of these species. These experiments demonstrated that the duck intestine contains predominantly a-(2,3)-linked oligosaccharides, whereas pig trachea contains carbohydrates with both a-(2,3)- and a-(2,6)-linked sialic acid. Whereas human influenza viruses generally do not replicate well in avian species, and vice versa (Refs 23, 24), pigs are susceptible to respiratory infections by both avian and human strains (Refs 25, 26). Indeed, it has been proposed that pigs might sometimes play a role in initiating influenza pandemics by serving as a ‘mixing vessel’, in which antigenically novel avian–human reassortant viruses might be generated following coinfection (Ref. 27).

Influenza virus neutralisation
Not only is the membrane-distal region of HA the location of the receptor-binding site, it is also the area to which most anti-HA neutralising antibodies bind. This was identified by selecting viral mutants that grow in the presence of neutralising monoclonal antibodies (mAbs) and mapping the location on the HA structure of the changes responsible for escape (Refs 2, 28). This approach identified mutations located throughout the surface of the membrane-distal head domains. Electron microscopy studies of such mAbs complexed with wild-type BHA confirmed that the mAbs bind to locations on the structure at which the changes leading to antibody resistance were selected (Ref. 29). X-ray crystallography studies of HA complexed with the Fab fragment of the neutralising antibody HC19 demonstrated that the antibody footprint covers residues at the right side of the binding site and that, upon binding, several residues of the antibody extend further into the pocket relative to their positions in the structure of the uncomplexed Fab fragment. HA complexes with another mAb, HC45, showed it to bind to a region at the base of the head domains, more distal from the receptor-binding site (Ref. 30). However, the bound mAb is oriented such that regions of the molecule distal from the footprint are in a position to interfere with HA–receptor interactions. These results, together with the observation that the membrane-distal domains around the receptor-binding site are most affected by antigenic drift, argue for receptor-binding interference as an important mechanism for antibody neutralisation of influenza viruses.

Membrane fusion
Binding of influenza virus to the cell-surface receptors is followed by endocytosis through clathrin-coated pits and vesicles. Fusion of the viral and cellular membrane takes place in endosomal compartments, and this ultimately allows for the transfer of viral nucleocapsids into the cytoplasm. The nucleocapsids subsequently migrate to the nucleus, where viral RNA transcription and replication take place. These initial stages of virus replication are dependent on the acidification of endosomes by cellular proton pumps. The reduction in endosomal pH triggers the conformational changes in HA that lead to membrane fusion (Refs 31, 32, 33). Furthermore, the interior of the virus particles is also acidified as a result of the proton-channel activity of the small viral transmembrane protein M2. This acidification causes the viral nucleocapsids to dissociate from the viral matrix protein, M1, and is important for their subsequent import into the nucleus. The role of M2 in virus entry and the import of viral gene segments into the nucleus is not addressed here (for recent reviews on these subjects see Refs 1 and 34).

HA structural changes required for fusion
The HA₂ subunit is the principal mediator of fusion, and cleavage of HA₀ into HA₁ and HA₂ is a prerequisite for the activation of fusion potential and virus infectivity (Refs 35, 36). The HA cleavage site, the enzymes involved, and the role of cleavage in virus pathogenicity have recently been reviewed (Ref. 37). HA precursor cleavage is required for fusion activity for two reasons. First, the conformational changes that accompany cleavage generate a structure that can then be triggered in the endosome to undergo the subsequent structural changes required for fusion. Second, the newly formed N-terminus of HA₂ generated upon cleavage plays a major role in the fusion process. This HA₂ N-terminus is a highly conserved hydrophobic domain and is commonly referred to as the fusion peptide, because its insertion into the target membrane is a critical phase of the fusion process.

In the native (cleaved) HA present on the surface of infectious virions, the long helix of each HA₂ subunit is linked by an extended chain to a shorter helix, which in turn is attached to the fusion peptide via two antiparallel b-strands (Fig. 4; fig004dsl). The hydrophobic fusion peptides present at the N-terminus of each of the three HA₂ monomers are found buried in the interior of the HA trimer, approximately 100Å from the top of the globular membrane-distal domain and 35Å from the viral membrane. At the pH of fusion, which is generally between pH 5.0 and 6.0 depending on the virus strain, the fusion peptides become extruded and the membrane-distal domains are detrimerised. Initially, this presented problems for attempts at crystallising BHA at low pH, as these molecules had disordered head domains and formed aggregated protein micelles owing to the association of the extruded fusion peptides (Ref. 28). These constraints on the crystallisation of low-pH HA were overcome by generating soluble fragments of these aggregates by proteolysis. Treatment of low-pH BHA aggregates with either protease Lys C or trypsin solubilises the globular HA₁ domains as monomers (HA₁ residues 28–328) as a result of cleavage at a lysine residue at HA₁ position 27. The remaining aggregates can then be solubilised by digestion with thermolysin, which removes the fusion peptide by cleavage at position 23 and then 37 of HA₂. The soluble trimeric product that results from such treatments contains HA₂ residues 38–175 disulphide-linked to HA₁ residues 1–27 in the low-pH conformation, and this fragment is referred to as TBHA₂. (A linear representation of HA fragments that have been crystallised is shown in Figure 5; fig005dsl.)

The X-ray crystal structure of TBHA₂ (Ref. 38) and the structure of the membrane-distal HA₁ domains liberated by Lys C digestion complexed with the Fab fragment of a mAb (Ref. 39) reveal that three major refolding events occur as a result of acidification. The globular membrane-distal domains of HA₁ (pink in Fig. 4; fig004dsl) that become detrimerised retain their secondary structure and are tethered to the trimeric TBHA₂ fragment via residues 28–43 of HA₁ (Ref. 39). The location of the monomeric head domains during fusion is unknown because of the disordered structure of these connecting residues. In HA₂ only 30 amino acids (residues 76–105 of HA₂) maintain the same structure in both native BHA and TBHA₂ (yellow in Fig. 4; fig004dsl). The two major acid-induced conformational changes in the HA₂ subunit appear to be more significant for the fusion process. One involves the refolding of an extended chain formed by residues 55–75 into an a-helix (light blue in Fig. 4; fig004dsl). These residues now become part of an extension of the coiled coil, which in the TBHA₂ structure also includes the residues that form the small helix of the native HA hairpin loop (red in Fig. 4; fig004dsl). As a result, the fusion peptide is relocated by 100Å from its buried position in the trimer to the membrane-distal tip of the newly formed coiled coil. The other structural change involves residues 106–112 in the membrane-proximal region of the native HA coiled coil (green in Fig. 4; fig004dsl), which unfold to form a loop. This helix-to-loop transition reorients residues at the C-terminal end of the a-helix by about 180° and these residues pack as an antiparallel helix against the long coiled coil. The reorientation of these helical residues also causes the inversion of residues C-terminal to this domain.

Escherichia coli-expressed HA₂ proteins
The significance of these changes for fusion has become more apparent from studies on HA₂ molecules expressed using E. coli. When HA₂ residues 38–175 (termed EBHA₂; see Fig. 5; fig005dsl) are expressed at neutral pH they fold directly into a soluble trimer that has a structure indistinguishable from that of TBHA₂ (Ref. 40). This observation reinforced the concept that native HA exists as a metastable molecule, which when triggered to induce fusion folds into a lower energy; fig005dsl), have provided further insights into HA-mediated fusion (Ref. 41). The EHA₂ molecule contains all HA₂ ectodomain residues except those of the fusion peptide. Compared with TBHA₂ and EBHA₂ it contains 15 additional amino acids at its N-terminus and is ten residues longer at the C-terminus. The main feature of this structure that is not apparent from that of TBHA₂ is that, as a result of the inversion of the HA stalk caused by the refolding of HA₂ residues 106–112, both the N- and C-termini are clearly at the same end of the molecule. In EHA₂, the residues C-terminal to the last region of ordered structure in TBHA₂ (dark pink in Fig. 4; fig004dsl) can be traced as a non-helical chain that extends antiparallel to the N-terminal coiled coil helices to the tip of the molecule. The N-terminal helix starts at HA₂ residue 38 and is preceded by amino acids Ala35, Ala36 and Asp37, which are conserved in all influenza A and B viruses. These conserved residues form a cap-like structure that terminates the triple-stranded coiled coil (Ref. 41). They cross laterally relative to the direction of the coiled coil to interact with the adjacent helix, and also form contacts with one another. It is possible that these interactions serve to stabilise this region of the structure.

The structure of EHA₂ suggests a model in which the relocation of the fusion peptide to the top of the coiled coil and the inversion of the HA stalk relative to the viral membrane bring both the N- and C-terminal membrane-associating domains to the same end of this rod-shaped molecule. This provides a mechanism for placing the two membranes in close proximity with one another, and this is a feature that is shared by several other proteins involved in fusion events. Recent X-ray crystallography and nuclear magnetic resonance (NMR) studies show that several other viral and cellular SNARE (for soluble NSF attachment protein receptor) fusion proteins contain ectodomains that, during fusion, exist as rod-like a-helical structures with coiled coil domains. In each example, the two membrane-associating domains are present at the same end of the molecule (reviewed in Ref. 42).

Characterisation of the fusion peptide
It is clear that the composition and/or structure of the fusion peptide, and possibly the transmembrane domain, can be critical for fusion function of influenza and other viruses. Fusion peptides are short, relatively hydrophobic sequences that are well conserved within, but not between, virus families. Although fusion peptide regions of the glycoproteins of unrelated viruses have little if any sequence conservation they do share certain features. As would be expected of domains that interact with membranes, they contain several large hydrophobic residues. They also contain several glycine residues interspersed throughout their sequences, the significance of which remains poorly understood.

The fusion peptide of the HA of the A/Aichi/2/68 virus has the sequence GLFGAIAGFIENGWEGMIDGWYG, with the amino acids in bold being completely conserved in all naturally occurring influenza A virus subtypes (Ref. 43). Mutagenesis studies have addressed the fusion properties of HAs with changes in this domain (Refs 44, 45, 46, 47, 48). Much of this work has concentrated on substitutions at the N-terminal residue of the fusion peptide, and most experiments indicate that glycine is the preferred amino acid at this position. In experiments using heterokaryon formation by HA-expressing cells as an assay for fusion (fusion of HA-expressing cells with one another following incubation at low pH), only an alanine substitution for the N-terminal glycine was found to result in effective function. It is interesting to note that a leucine was observed at position 1 of the fusion peptide when H7N7 subtype viruses were grown in the presence of thermolysin as the activating protease (Ref. 49). This suggests that residues other than glycine can be accommodated at this position when a suitable level of selective pressure is applied. At present it is not known what level of fusion is required for virus infectivity, but this is a question that is currently being addressed using reverse genetics (K.J. Cross et al., unpublished). Mutagenesis studies on other glycine residues show that alanine and glutamic acid can functionally replace glycine at position 4, but an alanine substitution for glycine at position 8 causes inhibition of fusion activity (Refs 44, 47). Despite the extremely conserved nature of the fusion peptide domain in isolates of influenza, it is known from laboratory experiments that, under appropriate selective pressures, mutant viruses can arise containing changes in this region (Refs 49, 50, 51, 52). These have been observed at several different positions, but in most examples such changes are of a conservative nature.

The structure that fusion peptides adopt when associated with membranes is one of the central issues that remains to be resolved. Hydrophobic photolabelling studies show that the N-terminal 22 residues of HA₂ are responsible for interaction with target membranes (Ref. 53), and these and other studies suggest that portions of this segment might adopt a helical structure for insertion into the membrane (Refs 54, 55, 56). It is often hypothesised that hydrophobicity and helix formation are necessary for interaction and destabilisation of membranes (Ref. 57). However, for the HA fusion peptide the experimental evidence indicates the presence of both a-helix and b-structure (Ref. 58). It has been suggested that the presence of oblique oriented or tilted helical peptides with an asymmetric gradient of hydrophobicity could result in membrane destabilisation by perturbing the regular packing of lipids. Several viral fusogenic proteins, lipases, apolipoproteins, proteins of neurodegenerative diseases, signal sequences, toxins and fusion proteins of spermatozoids have been found to contain peptides that can be modelled as such; however, the shortcoming of this proposal is the inability to prove that these peptides act as helical structures (Ref. 59). Recently, an NMR structure was published for a 20-residue fusion peptide analogue of HA indicating that, in micelles of dodecylphosphocholine, residues 2–18 formed an amphipathic a-helix (Ref. 60). It is not yet clear if this is a true reflection of the physiological structure, as the sequence of this peptide contains several glutamic acid substitutions compared with the wild-type sequence. Although there are clear obstacles involved in obtaining high-resolution structural information on this region of HA, this will be crucial for a comprehensive understanding of membrane fusion.

The composition of the membrane-anchor domain near the C-terminus of HA₂ can also have relevance for fusion. An HA mutant in which the transmembrane domain was replaced by a glycosylphosphatidylinositol (GPI) lipid anchor was found to cause only hemifusion: it was capable of mediating the fusion between the outer leaflets of donor and target membranes but not the transfer of aqueous contents (Refs 61, 62). Although recent studies using more-sensitive electrophysiological techniques suggest that the GPI-HA mutant can form small aqueous channels, it appears that these do not efficiently expand into fully functional fusion pores (Ref. 63), suggesting a role for the transmembrane domain in the fusion process. There are also examples of single-residue substitutions in the transmembrane domains that lead to inhibition of fusion function (Refs 64, 65). Despite this, there is evidence for a degree of permissiveness regarding the composition of transmembrane domains and fusion function. For example, chimaeric HAs containing the transmembrane domains of other viral and cellular proteins are capable of supporting fusion activity (Refs 62, 66).

Theoretical models for membrane fusion
Currently, one of the principal challenges in the field of membrane fusion concerns the events that occur after the two membranes are brought into close proximity with one another. For HA-mediated fusion, there does not seem to be a clear consensus regarding the number of trimers necessary to initiate fusion pore formation. Measurement of HA surface density suggests the involvement of more than one HA trimer (Ref. 67) and the rate of fusion is influenced by both HA surface density and target lipid composition (Ref. 68). On the basis of studies on fusion kinetics, it was proposed that the concerted action of three or four HA trimers is necessary (Ref. 69). In another study, monitoring of the fusion kinetics between individual pairs of HA-expressing cells and erythrocytes suggested that the fusion pore is composed of six HA trimers (Ref. 70). Yet another study argued on theoretical grounds that at least eight or more HA trimers need to be aggregated at the fusion site, but that only two or three of these need to undergo the conformational change necessary for fusion (Ref. 71).

Numerous models regarding the mechanisms involved in HA-mediated membrane fusion have been proposed, and virtually all take into account the acid-induced conformational changes described above. However, there is a great deal of diversity among these models, and views range from those in which the initial fusion pore is thought to be lipidic to those in which it is entirely proteinaceous.

The ‘stalk-pore hypothesis’ (Refs 72, 73) is illustrated in Figure 6 (fig006dsl). According to this model, the conformational change involving the extension of the coiled coil drives the initial interaction of the fusion peptide with the target membrane. It is suggested that HA trimers then cluster and a cooperative interaction occurs among the fusion peptides. Refolding of the membrane-proximal region of the long a-helices of HA₂ then leads to bending of the membranes. Since HA remains firmly attached to both target membrane through its fusion peptide and viral membrane through its transmembrane domain any such concerted bending action would drive the membrane bilayers into close proximity, leading to a hemifusion intermediate. Hemifusion occurs when the outer lipid bilayer leaflets of the two opposing membranes are pulled together, forming a highly bent stalk. The outer leaflets fuse and the distal leaflets are pulled towards each other, forming a dimple. The intermediate hemifusion diaphragm continues to expand until the accumulated tension destabilises the structure, rupturing it to form a lipid-lined fusion pore. This initially flickers, then undergoes transient and reversible opening and closing, and finally dilates irreversibly to allow for mixing of the aqueous contents.

On the basis of experimental evidence for fusion pore conductance before the onset of lipid mixing (Refs 74, 75), an alternative model was proposed in which fusion occurs through the concerted action of multiple HA trimers that form an oligomeric ring-like structure. This spans both target and viral bilayers and results in the production of an entirely proteinaceous bridge. The pore that forms would initially be lined by proteins and would only later be invaded by lipids when the trimers have dissociated. In a variation of the proteinaceous pore model (Ref. 76), it has been speculated that insertion of fusion peptides into the target membrane might initially cause the formation of small stable pores that allow for the leakage of small molecules. HA trimer interactions and association would then lead to larger pore formation, and pore expansion would follow via the hemifusion pathway described above through to complete fusion.

Another model has been hypothesised in which the fusion peptides insert into the viral membrane rather than the target membrane (Ref. 77). The refolding of the HA then causes the inserted peptides to bend the membrane around the trimer. The initial deformation of the membrane is followed by the assembly of a ring-like cluster of HA trimers that surround a dimple of viral lipids. This lipidic dimple then grows towards the target membrane, where the bending stresses at the dimple top aid in the fusion process.

Yet another hypothesis involves the aggregation of at least eight HAs at the fusion site with their fusion peptides buried in the viral membrane (Ref. 78). Of these, only two or three undergo the conformational change and extract their fusion peptides to the target membrane, and, as a result, a hydrophobic defect is created in the outer monolayer of the virus. As lipid flow is restricted as a result of aggregation of the trimers, the defect is remedied by recruitment of lipids from the outer monolayer of the target membrane, forming a stalk that connects the two bilayers, leading to fusion.

Concluding remarks
Although studies on influenza HA have been at the forefront of research into cell entry by enveloped viruses, many questions involving HA binding and membrane fusion remain unresolved. The molecular basis for receptor-binding specificity and the role this plays in viral tropism and host range is not as clear-cut as is often portrayed. The relationship between binding affinity and receptor availability requires further study. This understanding will be important for the design of antiviral compounds that block binding, and will be particularly relevant for the use of HA mutants as providers of fusion function in gene therapy vectors containing other molecules that confer target-cell specificity. The mechanism of virus neutralisation by antibodies and its relationship with receptor-binding properties will remain a concern regarding vaccine design. The interplay between characteristics of the HA and the NA viral surface proteins, and the influence of this on the emergence of variants with the capacity to cross species barriers and replicate efficiently in novel hosts, is not yet fully appreciated. In addition, it is possible that this HA–NA interplay might eventually become a factor in the emergence of mutant influenza viruses that develop resistance to antiviral compounds that target HA or NA.

Many details regarding the mechanism of HA-mediated membrane fusion at the molecular and biophysical level remain elusive. In this respect, studies comparing HA fusion with other viral and cellular fusion systems might ultimately aid understanding. Although the issue is the same for each system – getting opposing membranes to merge – there is a great deal of diversity regarding the specifics of how fusion is mediated. Among the viral proteins, there are clear distinctions with respect to several properties: (1) the requirement for cleavage of a precursor protein; (2) the structural rearrangements that occur and the reversibility of such rearrangements; (3) the location of the fusion peptides within the protein (internally or at the terminus of a polypeptide); (4) the presence of separate viral binding and fusion proteins; (5) the requirement for cellular co-receptors; and (6) the cellular location of fusion and the pH at which it takes place (see for example Refs 79 and 80). Studies on cellular vesicle fusion involving the SNARE protein complexes have also intensified in recent years, and these will continue to provide an abundance of novel structural and functional information that will aid in the understanding of membrane fusion in general (Refs 81, 82, 83, 84, 85).

Acknowledgements and funding
We thank John Skehel [National Institute for Medical Research (NIMR), London, UK] and Steve Wharton (NIMR) for critical comments on the manuscript, and Lesley Calder (NIMR) for providing Figure 1. The authors’ work is funded by the Medical Research Council, UK.

References
1 Whittaker, G.R. (2001) Intracellular trafficking of influenza virus: clinical implications for molecular medicine. Exp Rev Mol Med, 8 February, http://www.expertreviews.org/01002447h.htm

2 Skehel, J.J. and Wiley, D.C. (2000) Receptor binding and membrane fusion in virus entry: the influenza hemagglutinin. Annu Rev Biochem 69, 531-569, PubMed

3 Molinari, M. and Helenius, A. (2000) Chaperone selection during glycoprotein translocation into the endoplasmic reticulum. Science 288, 331-333, PubMed

4 Keller, P. and Simons, K. (1998) Cholesterol is required for surface transport of influenza virus hemagglutinin. J Cell Biol 140, 1357-1367, PubMed

5 Scheiffele, P., Roth, M.G. and Simons, K. (1997) Interaction of influenza virus haemagglutinin with sphingolipid-cholesterol membrane domains via its transmembrane domain. Embo J 16, 5501-5508, PubMed

6 Melkonian, K.A. et al. (1999) Role of lipid modifications in targeting proteins to detergent- resistant membrane rafts. Many raft proteins are acylated, while few are prenylated. J Biol Chem 274, 3910-3917, PubMed

7 Zhang, J, Pekosz, A. and Lamb, R.A. (2000) Influenza virus assembly and lipid raft microdomains: a role for the cytoplasmic tails of the spike glycoproteins. J Virol 74, 4634-4644, PubMed

8 Brand, C.M. and Skehel, J.J. (1972) Crystalline antigen from the influenza virus envelope. Nat New Biol 238, 145-147, PubMed

9 Wilson, I.A., Skehel, J.J. and Wiley, D.C. (1981) Structure of the haemagglutinin membrane glycoprotein of influenza virus at 3 A resolution. Nature 289, 366-373, PubMed

10 Hirst, G.K. (1941) The agglutination of red cells by allantoic fluid of chick embryos infected with influenza virus. Science 94, 22-23

11 Gottschalk, A. (1959) The chemistry of virus receptors. In The Viruses (Burnet, F.M. and Stanley, W.M., eds), pp. 51-61, Academic Press, New York, USA

12 Eisen, M.B. et al. (1997) Binding of the influenza A virus to cell-surface receptors: structures of five hemagglutinin-sialyloligosaccharide complexes determined by X-ray crystallography. Virology 232, 19-31, PubMed

13 Sauter, N.K. et al. (1992) Binding of influenza virus hemagglutinin to analogs of its cell-surface receptor, sialic acid: analysis by proton nuclear magnetic resonance spectroscopy and X-ray crystallography. Biochemistry 31, 9609-9621, PubMed

14 Watowich, S.J., Skehel, J.J. and Wiley, D.C. (1994) Crystal structures of influenza virus hemagglutinin in complex with high-affinity receptor analogs. Structure 2, 719-731, PubMed

15 Weis, W. et al. (1988) Structure of the influenza virus haemagglutinin complexed with its receptor, sialic acid. Nature 333, 426-431, PubMed

16 Rogers, G.N. et al. (1983) Single amino acid substitutions in influenza haemagglutinin change receptor binding specificity. Nature 304, 76-78, PubMed

17 Hanson, J.E. et al. (1992) Proton nuclear magnetic resonance studies of the binding of sialosides to intact influenza virus. Virology 189, 525-533, PubMed

18 Martin, J. et al. (1998) Studies of the binding properties of influenza hemagglutinin receptor- site mutants. Virology 241, 101-111, PubMed

19 Brandli, A.W. et al. (1988) A polarized epithelial cell mutant deficient in translocation of UDP- galactose into the Golgi complex. J Biol Chem 263, 16283-16290, PubMed

20 Sauter, N.K. et al. (1989) Hemagglutinins from two influenza virus variants bind to sialic acid derivatives with millimolar dissociation constants: a 500-MHz proton nuclear magnetic resonance study. Biochemistry 28, 8388-8396, PubMed

21 Baum, L.G. and Paulson, J.C. (1990) Sialyloligosaccharides of the respiratory epithelium in the selection of human influenza virus receptor specificity. Acta Histochem Suppl 40, 35-38, PubMed

22 Ito, T. et al. (1998) Molecular basis for the generation in pigs of influenza A viruses with pandemic potential. J Virol 72, 7367-7373, PubMed

23 Beare, A.S. and Webster, R.G. (1991) Replication of avian influenza viruses in humans. Arch Virol 119, 37-42, PubMed

24 Hinshaw, V.S. et al. (1983) Altered tissue tropism of human-avian reassortant influenza viruses. Virology 128, 260-263, PubMed

25 Hinshaw, V.S. et al. (1981) Replication of avian influenza A viruses in mammals. Infect Immun 34, 354-361, PubMed

26 Kida, H. et al. (1994) Potential for transmission of avian influenza viruses to pigs. J Gen Virol 75, 2183-2188, PubMed

27 Scholtissek, C. et al. (1985) The nucleoprotein as a possible major factor in determining host specificity of influenza H3N2 viruses. Virology 147, 287-294, PubMed

28 Wiley, D.C. and Skehel, J.J. (1987) The structure and function of the hemagglutinin membrane glycoprotein of influenza virus. Annu Rev Biochem 56, 365-394, PubMed

29 Wrigley, N.G. et al. (1983) Electron microscopy of influenza haemagglutinin-monoclonal antibody complexes. Virology 131, 308-314, PubMed

30 Fleury, D. et al. (1999) A complex of influenza hemagglutinin with a neutralizing antibody that binds outside the virus receptor binding site. Nat Struct Biol 6, 530-534, PubMed

31 Huang, R.T., Rott, R. and Klenk, H.D. (1981) Influenza viruses cause hemolysis and fusion of cells. Virology 110, 243-247, PubMed

32 Maeda, T. and Ohnishi, S. (1980) Activation of influenza virus by acidic media causes hemolysis and fusion of erythrocytes. FEBS Lett 122, 283-287, PubMed

33 White, J., Matlin, K. and Helenius, A. (1981) Cell fusion by Semliki Forest, influenza, and vesicular stomatitis viruses. J Cell Biol 89, 674-679, PubMed

34 Whittaker, G.R., Kann, M. and Helenius, A. (2000) Viral entry into the nucleus. Annu Rev Cell Dev Biol 16, 627-651, PubMed

35 Klenk, H.D. et al. (1975) Activation of influenza A viruses by trypsin treatment. Virology 68, 426-439, PubMed

36 Lazarowitz, S.G. and Choppin, P.W. (1975) Enhancement of the infectivity of influenza A and B viruses by proteolytic cleavage of the hemagglutinin polypeptide. Virology 68, 440-454, PubMed

37 Steinhauer, D.A. (1999) Role of hemagglutinin cleavage for the pathogenicity of influenza virus. Virology 258, 1-20, PubMed

38 Bullough, P.A. et al. (1994) Structure of influenza haemagglutinin at the pH of membrane fusion. Nature 371, 37-43, PubMed

39 Bizebard, T. et al. (1995) Structure of influenza virus haemagglutinin complexed with a neutralizing antibody. Nature 376, 92-94, PubMed

40 Chen, J. et al. (1995) A soluble domain of the membrane-anchoring chain of influenza virus hemagglutinin (HA2) folds in Escherichia coli into the low-pH-induced conformation. Proc Natl Acad Sci U S A 92, 12205-12209, PubMed

41 Chen, J., Skehel, J.J. and Wiley, D.C. (1999) N- and C-terminal residues combine in the fusion-pH influenza hemagglutinin HA(2) subunit to form an N cap that terminates the triple- stranded coiled coil. Proc Natl Acad Sci U S A 96, 8967-8972, PubMed

42 Skehel, J.J. and Wiley, D.C. (1998) Coiled coils in both intracellular vesicle and viral membrane fusion. Cell 95, 871-874, PubMed

43 Nobusawa, E. et al. (1991) Comparison of complete amino acid sequences and receptor-binding properties among 13 serotypes of hemagglutinins of influenza A viruses. Virology 182, 475-485, PubMed

44 Gething, M.J. et al. (1986) Studies on the mechanism of membrane fusion: site-specific mutagenesis of the hemagglutinin of influenza virus. J Cell Biol 102, 11-23, PubMed

45 Nobusawa, E. et al. (1995) The role of acidic residues in the "fusion segment" of influenza A virus hemagglutinin in low-pH-dependent membrane fusion. Arch Virol 140, 865-875, PubMed

46 Qiao, H. et al. (1999) A specific point mutant at position 1 of the influenza hemagglutinin fusion peptide displays a hemifusion phenotype. Mol Biol Cell 10, 2759-2769, PubMed

47 Steinhauer, D.A. et al. (1995) Studies of the membrane fusion activities of fusion peptide mutants of influenza virus hemagglutinin. J Virol 69, 6643-6651, PubMed

48 Walker, J.A. and Kawaoka, Y. (1993) Importance of conserved amino acids at the cleavage site of the haemagglutinin of a virulent avian influenza A virus. J Gen Virol 74, 311-314, PubMed

49 Orlich, M. and Rott, R. (1994) Thermolysin activation mutants with changes in the fusogenic region of an influenza virus hemagglutinin. J Virol 68, 7537-7539, PubMed

50 Daniels, R.S. et al. (1985) Fusion mutants of the influenza virus hemagglutinin glycoprotein. Cell 40, 431-439, PubMed

51 Lin, Y.P. et al. (1997) Adaptation of egg-grown and transfectant influenza viruses for growth in mammalian cells: selection of hemagglutinin mutants with elevated pH of membrane fusion. Virology 233, 402-410, PubMed

52 Orlich, M., Linder, D. and Rott, R. (1995) Trypsin-resistant protease activation mutants of an influenza virus. J Gen Virol 76, 625-633, PubMed

53 Durrer, P. et al. (1996) H+-induced membrane insertion of influenza virus hemagglutinin involves the HA2 amino-terminal fusion peptide but not the coiled coil region. J Biol Chem 271, 13417-13421, PubMed

54 Harter, C. et al. (1989) Hydrophobic binding of the ectodomain of influenza hemagglutinin to membranes occurs through the "fusion peptide". J Biol Chem 264, 6459-6464, PubMed

55 Lear, J.D. and DeGrado, W.F. (1987) Membrane binding and conformational properties of peptides representing the NH2 terminus of influenza HA-2. J Biol Chem 262, 6500-6505, PubMed

56 Wharton, S.A. et al. (1988) Membrane fusion by peptide analogues of influenza virus haemagglutinin. J Gen Virol 69, 1847-1857, PubMed

57 Pecheur, E.I. et al. (1999) Peptides and membrane fusion: towards an understanding of the molecular mechanism of protein-induced fusion. J Membr Biol 167, 1-17, PubMed

58 Gray, C. et al. (1996) Effect of the N-terminal glycine on the secondary structure, orientation, and interaction of the influenza hemagglutinin fusion peptide with lipid bilayers. Biophys J 70, 2275-2286, PubMed

59 Brasseur, R. (2000) Tilted peptides: a motif for membrane destabilization (hypothesis). Mol Membr Biol 17, 31-40, PubMed

60 Dubovskii, P.V. et al. (2000) Structure of an analog of fusion peptide from hemagglutinin. Protein Sci 9, 786-798, PubMed

61 Kemble, G.W., Danieli, T. and White, J.M. (1994) Lipid-anchored influenza hemagglutinin promotes hemifusion, not complete fusion. Cell 76, 383-391, PubMed

62 Melikyan, G.B., White, J.M. and Cohen, F.S. (1995) GPI-anchored influenza hemagglutinin induces hemifusion to both red blood cell and planar bilayer membranes. J Cell Biol 131, 679-691, PubMed

63 Markosyan, R.M., Cohen, F.S. and Melikyan, G.B. (2000) The lipid-anchored ectodomain of influenza virus hemagglutinin (GPI-HA) is capable of inducing nonenlarging fusion pores. Mol Biol Cell 11, 1143-1152, PubMed

64 Cleverley, D.Z. and Lenard, J. (1998) The transmembrane domain in viral fusion: essential role for a conserved glycine residue in vesicular stomatitis virus G protein. Proc Natl Acad Sci U S A 95, 3425-3430, PubMed

65 Melikyan, G.B. et al. (1999) Amino acid sequence requirements of the transmembrane and cytoplasmic domains of influenza virus hemagglutinin for viable membrane fusion. Mol Biol Cell 10, 1821-1836, PubMed

66 Kozerski, C. et al. (2000) Modification of the cytoplasmic domain of influenza virus hemagglutinin affects enlargement of the fusion pore. J Virol 74, 7529-7537, PubMed

67 Ellens, H. et al. (1990) Fusion of influenza hemagglutinin-expressing fibroblasts with glycophorin-bearing liposomes: role of hemagglutinin surface density. Biochemistry 29, 9697-9707, PubMed

68 Clague, M.J., Schoch, C. and Blumenthal, R. (1991) Delay time for influenza virus hemagglutinin-induced membrane fusion depends on hemagglutinin surface density. J Virol 65, 2402-2407, PubMed

69 Danieli, T. et al. (1996) Membrane fusion mediated by the influenza virus hemagglutinin requires the concerted action of at least three hemagglutinin trimers. J Cell Biol 133, 559-569, PubMed

70 Blumenthal, R. et al. (1996) Dilation of the influenza hemagglutinin fusion pore revealed by the kinetics of individual cell-cell fusion events. J Cell Biol 135, 63-71, PubMed

71 Bentz, J. (2000) Minimal aggregate size and minimal fusion unit for the first fusion pore of influenza hemagglutinin-mediated membrane fusion. Biophys J 78, 227-245, PubMed

72 Hernandez, L.D. et al. (1996) Virus-cell and cell-cell fusion. Annu Rev Cell Dev Biol 12, 627-661, PubMed

73 Jahn, R. and Sudhof, T.C. (1999) Membrane fusion and exocytosis. Annu Rev Biochem 68, 863-911, PubMed

74 Lindau, M. and Almers, W. (1995) Structure and function of fusion pores in exocytosis and ectoplasmic membrane fusion. Curr Opin Cell Biol 7, 509-517, PubMed

75 Tse, F.W., Iwata, A. and Almers, W. (1993) Membrane flux through the pore formed by a fusogenic viral envelope protein during cell fusion. J Cell Biol 121, 543-552, PubMed

76 Bonnafous, P. and Stegmann, T. (2000) Membrane perturbation and fusion pore formation in influenza hemagglutinin-mediated membrane fusion. A new model for fusion. J Biol Chem 275, 6160-6166, PubMed

77 Kozlov, M.M. and Chernomordik, L.V. (1998) A mechanism of protein-mediated fusion: coupling between refolding of the influenza hemagglutinin and lipid rearrangements. Biophys J 75, 1384-1396, PubMed

78 Bentz, J. (2000) Membrane fusion mediated by coiled coils: a hypothesis. Biophys J 78, 886-900, PubMed

79 Gaudin, Y. et al. (1999) Rabies virus-induced membrane fusion. Mol Membr Biol 16, 21-31, PubMed

80 Lamb, R.A. (1993) Paramyxovirus fusion: a hypothesis for changes. Virology 197, 1-11, PubMed

81 Lin, R.C. and Scheller, R.H. (2000) Mechanisms of synaptic vesicle exocytosis. Annu Rev Cell Dev Biol 16, 19-49, PubMed

82 McNew, J.A. et al. (2000) Compartmental specificity of cellular membrane fusion encoded in SNARE proteins. Nature 407, 153-159, PubMed

83 Pelham, H.R. and Rothman, J.E. (2000) The debate about transport in the Golgi–two sides of the same coin? Cell 102, 713-719, PubMed

84 Scales, S.J., Bock, J.B. and Scheller, R.H. (2000) The specifics of membrane fusion. Nature 407, 144-146, PubMed

85 Sutton, R.B. et al. (1998) Crystal structure of a SNARE complex involved in synaptic exocytosis at 2.4 A resolution. Nature 395, 347-353, PubMed

| home | search | glossary | links | sitemap | contact |

Expert Reviews in Molecular Medicine © Cambridge University Press ISSN 1462-3994 (Disclaimer and copyright)
Editorial Office: Centre for Applied Research in Educational Technologies (CARET), 1st Floor, 16 Mill Lane, Cambridge, CB2 1SB, UK. Tel: +44 (0)1223 765 375; Fax: +44(0)1223 765 505; E-mail: ermm@caret.cam.ac.uk