Epstein-Barr Virus Entry^▿

(i) B cells.

Eight virus glycoproteins have been implicated in some way in EBV entry into either a B cell or an epithelial cell (Table 1). One of the most abundant of these in the virus envelope, gp350/220 (22), is responsible for attachment of the virus with high affinity (36) to the complement receptor type 2 (CR2) on B cells (7, 8, 40, 41, 60, 61). An EBV recombinant lacking gp350/220 can transform B cells with much-reduced efficiency (20), so CR2 is perhaps not the only portal by which the virus can access a B cell. It is, however, clearly the predominant one (Fig. 1). Antibodies to gp350/220 that block virus binding neutralize B-cell infection, and soluble forms of either CR2 or gp350/220 can do the same (35, 61).

Glycoprotein gp350/220 is a highly glycosylated single-pass membrane protein, which as a result of alternative splicing is made in two forms, with approximate masses of 350 and 220 kDa (1, 17). Residues 500 to 757, which include three repeats of a 21-amino-acid motif with amphipathic characteristics, are lost from the full-length 907-amino-acid protein as a result of the splice. It does, however, maintain the reading frame, and both forms of the protein retain the CR2 binding domain at the amino terminus (61). Solution of the structure of an unliganded form of gp350/220 has mapped this domain to a glycan-free surface (59) that includes a sequence with some similarity to the natural ligand of CR2, the C3dg fragment of complement (26, 39, 61). The extracellular domain of CR2 is composed of tandem repeats of structural modules 60 to 75 amino acids in length known as short consensus repeats (SCR), and the binding site for gp350/220 has been genetically mapped to a region composed of residues from the membrane distal SCR1 and SCR2 (30). Attachment initially positions the virus at approximately 50 nm from the cell surface (38), but the segmental flexibility that the repeated SCRs may give to CR2 (68) and the possibility of an exchange in binding from gp350 to gp220 may contribute to moving the virus closer to the cell membrane.

Binding of gp350/220 also triggers capping of CR2 and endocytosis of the virus (38, 60). CR2 can function as a signal transducer both independently and as part of a signal transduction complex that includes CD19 and CD35 (5). Cross-linking by gp350/220 activates NF-κB (53, 58) and induces interleukin-6 via a protein kinase C pathway (4, 62). None of these signaling events may be critical to penetration of the cell membrane, but they clearly have potential consequences for downstream events in infection.

(ii) Epithelial cells.

Attachment of the virus to an epithelial cell is a more-complicated affair (Fig. 2). CR2 is expressed at least at low levels on some epithelial cells in culture (6) in the absence of CD19 and CD35, and infection of an epithelial cell that has been engineered to express high levels of CR2 can be very efficient (3, 29). Notably, in this regard, EBV-mediated cross-linking of CR2 on epithelial cells, which lack the other components of the CR2 signaling complex found on B cells, stimulates relocalization and clustering of CR2 and the formin homolog overexpressed in spleen (FHOS/FHOD1). Formins are molecular scaffolds that nucleate actin, linking signal transduction to actin reorganization and gene transcription (10), and this may contribute to successful delivery of the virus to the nucleus of the cell. However, which, or even whether, epithelial cells normally express CR2 in vivo remains uncertain. Studies with monoclonal antibodies to CR2 were compromised by the finding that the antibody most commonly used cross-reacted with an unrelated epithelial cell protein (73).

At least three other possible attachment mechanisms have been proposed that involve neither gp350/220 nor CR2. The first was a demonstration that virus coated with immunoglobulin A specific to gp350/220 can bind productively to the polymeric immunoglobulin-A receptor (54). This may be particularly relevant to infection via the basolateral surface of an epithelial cell in an immune host, although, since in polarized cells virus was transported intact from the basolateral to the apical surface, it may be more relevant to trans epithelial transport than direct infection (9). The second was a demonstration that in the absence of CR2 a complex of two additional glycoproteins, gH and gL, can serve as epithelial ligands (34, 43). EBV gH, a type 1 membrane protein, is, like its homologs in other herpesviruses, misfolded and retained in the endoplasmic reticulum in the absence of the much smaller type 2 protein, gL (27, 72). No separate function in addition to facilitating folding and transport of gH has been ascribed to gL, and the two proteins, which associate noncovalently, are usually referred to as a unit, the gHgL complex. EBV derived from a B cell can bind well to a CR2-negative epithelial cell, but recombinant viruses that lack gHgL lose this ability (34, 43). A soluble form of gHgL made in baculovirus can bind specifically to epithelial cells, but not B cells, and its binding can be reduced by a monoclonal antibody specific for the gHgL complex (3). The same antibody can also reduce virus binding (34). These observations have been interpreted to mean that there is an epithelial cell receptor for gHgL that can serve in attachment. The identity of the molecule is not yet known, but operationally it has been referred to as gHgLR.

Finally, and most intriguing, on polarized epithelial cells, which presumably resemble more closely the environment that the virus encounters in vivo, an interaction between a multispan virus membrane protein encoded by the BMRF2 open reading frame and integrins has been demonstrated (63). Antibodies to integrins and to a BMRF2 fusion protein only partially block binding to polarized epithelial cells but have a significant impact on infection via the basolateral surface of the polarized monolayer. The RGD motif present in an exposed loop of BMRF2 has been shown to be a ligand for β1, α5, α3, and αv integrins (71). The protein is not required for cell-cell fusion (12, 31). However, whether the BMRF2 integrin interaction is primarily responsible for attachment or whether it is most relevant to postattachment events to which signaling may make an important contribution is not yet clear. There is apparently very little BMRF2 protein in the virion (22).

(i) B cells.

Fusion of the virus with a normal B cell (Fig. 1) requires endocytosis, which is triggered by the interaction between gp350/220 and CR2 and occurs in a low-pH compartment, although low pH itself is not required (32). The virus proteins that are both necessary and sufficient for efficient fusion are gHgL, gB, and gp42. A role for gH was first demonstrated when it was shown that virosomes made from proteins depleted of those that bound to an antibody to gH could attach to B cells but could not fuse (14). It was later confirmed by analysis of the phenotype of a recombinant gH-null virus (34). A role for EBV gB was not possible to evaluate genetically, since, unlike its homologs in other herpesviruses, it is essential for virus morphogenesis and egress (16). However, when cell-cell fusion assays were developed as models of virus-cell fusion, it became apparent that its contribution is also very important (12, 31).

The fourth and final protein involved in fusion of EBV with a B cell, gp42, has homologs only among the lymphocryptoviruses (48, 49). It is another type 2 membrane protein, and it associates noncovalently with gHgL. The original observation was that gp42 resembles a C-type lectin and that a soluble form made by replacing the signal sequence with one that is efficiently cleaved can bind to HLA class II (56). Within the three-part complex of gHgLgp42 in the virus, this interaction is now thought to provide the trigger for B-cell fusion. A monoclonal antibody that was initially mapped to gH (57) but ultimately proved to react with gp42 (28) inhibited interaction with HLA class II (27) and also blocked virus-cell fusion (33). Reciprocally, a monoclonal antibody to HLA class II that blocked interaction with gp42 also inhibited infection. The soluble form of gp42 initially used to document HLA class II binding could inhibit infection by competing with gp42 in the virus, and B cells that lacked HLA class II could be efficiently infected only if HLA class II expression was restored (27). A recombinant gp42-null virus could infect a B cell only if the cells and the bound virus were treated with an exogenous fusogen such as polyethylene glycol (65) or if the soluble form of gp42, which retained a gHgL binding domain at its amino terminus, was added in trans to reform three-part complexes (66).

Binding of gp42 to HLA class II shows some allelic specificity in that nonfunctional HLA-DQ alleles have been identified (13), but since all three alleles, HLA-DP, HLA-DQ, and HLA-DR, can be used, it is perhaps unlikely that this has a major impact in an outbred human population. A crystal structure of the ectodomain of gp42 liganded to HLA-DR1 has been solved (37). The protein is most closely related to natural killer cell receptors, such as the murine receptor LY49, which interact with major histocompatibility complex (MHC) class I molecules, but it does not use the canonical surface site which many C-type lectins use to bind to their ligands. Instead, this site forms a hydrophobic pocket that, it is suggested, may be used to accommodate another virus protein or another cell protein important for virus entry (37), perhaps after triggering by HLA class II. This is consistent with the observation that mutations in the hydrophobic pocket do not affect HLA class II binding but inhibit fusion (52). The contact region on HLA-DR1 is within the β1 domain to one side of the peptide binding groove. It was thought on theoretical grounds to be likely to interfere sterically with the binding of T-cell receptors to the peptide MHC complex and has been shown experimentally to reduce T-helper-cell recognition of B cells (45, 46, 56). Both peptide-loaded and immature MHC class II proteins that have not yet trafficked to the peptide-loading compartment can interact with gp42 (46), and, as discussed below, this has proven to have an interesting impact on virus tropism. Whether or not additional signaling events may be initiated as a result of a gp42-HLA class II interaction is not yet known, but since signaling via HLA class II molecules is well documented under other circumstances, this remains a reasonable possibility (67).

(ii) Epithelial cells.

Fusion with a normal epithelial cell, at least one that is nonpolarized, takes place at neutral pH and does not appear to require endocytosis (3, 32). Epithelial cells do not constitutively express HLA class II, and thus, not surprisingly, infection is not blocked by the monoclonal antibody that blocks gp42 and HLA class II interactions. A gp42-null virus can infect an epithelial cell as well as, or even better than, wild-type virus. However, not only is gp42 dispensable for epithelial cell infection, but its presence is actually inhibitory (66). Stoichiometric analysis of gHgL and gp42 in the virion indicated that wild-type EBV carries far more gHgL than gp42, implying that there are two kinds of gH complexes, one the three-part complex composed of gHgLgp42 and the other a two-part complex composed of only gHgL. Only three-part complexes can mediate B-cell infection, but conversion of all two-part complexes to three-part complexes by the addition of soluble gp42 in trans blocks epithelial cell infection. This was interpreted to mean that fusion with an epithelial cell is triggered by a direct interaction between gHgL and a novel epithelial cell surface molecule. Three monoclonal antibodies that interact with gH or gHgL and have no effect on B-cell infection can also inhibit infection of epithelial cells, whether virus is bound via gp350/220 to CR2 or via gHgL to gHgLR. One of them blocks gHgL binding to gHgLR, suggesting that the cell molecule that triggers fusion might be the same protein, gHgLR, which can serve as an attachment receptor in the absence of CR2. Consistent with this is the observation that soluble gp42 also blocks binding of the virus to gHgLR. Use of gHgL for attachment as well as fusion does, however, seem to compromise infection. The virus can bind efficiently to gHgLR on a CR2-negative epithelial cell, but infection levels are low, either because fusion becomes less efficient or perhaps because engagement of gHgLR does not induce a downstream event required for completion of transport of the virus genome to the nucleus (3).

Beyond this, epithelial cell penetration also requires higher levels of gB than does B-cell penetration (42), presumably reflecting a need for more gB in fusion (31). However, the mechanics of fusion itself remain obscure, as they do for all herpesviruses. Mutations in an amino-terminal region of gH that is predicted to form a coiled coil significantly reduced fusion in a cell-cell fusion assay (44), and insertion or point mutations in a region close to the transmembrane domain can either differentially affect B-cell and epithelial cell fusion or abrogate fusion with either cell type (69, 70). However, some epithelial cell fusion can be mediated in the absence of gHgL by a gB construct with a mutation in the cytoplasmic tail. This has been interpreted to mean that gB may be the critical fusion protein of EBV. Certainly the crystal structure of the homolog of gB in herpes simplex virus shows structural homology with the vesicular stomatitis virus G protein, which is the virus fusogen (15). Both form trimers characterized by an alpha-helical coiled-coil core and extended β-hairpins that are characteristic of class I and class II fusion proteins, respectively. At this point, however, perhaps the most that can be definitively said is that virus fusion with a B cell requires a trigger that is transmitted via gp42 to gHgL and gB, whereas fusion with an epithelial cell requires a trigger that is transmitted directly to gHgL. Whether there is a temporal cascade of events culminating in fusion mediated by gB or whether gHgL and gB coordinately contribute to the process remains to be determined.

gp42 as a switch of tropism.

The observation that two-part gHgL and three-part gHgLgp42 complexes are mutually exclusive for epithelial and B-cell entry suggested that the relative amounts of each complex in the virion might impact virus tropism. Analyses of viruses made in B cells and epithelial cells confirmed this. In a B cell, some three-part gHgLgp42 complexes prove to interact with immature MHC class II in the endoplasmic reticulum and are targeted to the HLA class II trafficking pathway, where they are vulnerable to degradation. This does not happen in an HLA class II-negative epithelial cell. Epithelial cell viruses are, as a result, rich in complexes containing gp42 and are as much as two orders of magnitude more infectious for a B cell than viruses produced by a B cell. B-cell viruses, lower in three-part complexes but rich in two-part complexes, are better able to infect an epithelial cell, although the phenotype is not nearly as striking. This suggested a model (Fig. 3) of persistent infection in which virus reactivating from a latently infected memory B cell as it terminally differentiates into a plasmablast (23) is equipped to infect an epithelial cell, whereas virus amplified in an epithelial cell is strongly B cell tropic (2).

EBV is an orally transmitted virus, and the switch in tropism effected by levels of gp42, which could impact transmission, raised the question of the cellular origins of viruses that are shed in saliva. Viruses made in a B cell, low in gp42, have been shown to be able to bind to gHgLR on an epithelial cell almost as well as they can bind CR2 on a B cell. In contrast, virus made in an epithelial cell, high in gp42, binds gHgLR very poorly (3). A comparison of the differential abilities of viruses from saliva and viruses from transformed B cells obtained from a series of EBV-positive donors then facilitated the determination that the majority of viruses shed in saliva are derived from HLA class II-negative cells (21). For the most-efficient transmission, this either means that the virus directly accesses a B cell in the oral mucosa or that transmission to an epithelial cell for amplification before accessing a B cell is influenced by other events. Two have been proposed, and both implicate the differential use of gp350/220.

gp350/220 as an impediment to epithelial infection.

Antibodies to gp350/220, which inhibit infection of a B cell dependent on CR2 for entry, can enhance infection of an epithelial cell, which is not (64). The effect is not mediated by Fc receptor binding but is further increased by antibody cross-linking, which may patch gp350/220 in the virus envelope. Saliva from EBV-seropositive individuals has similar effects that can be reversed by depletion of antibodies. The interpretation of these results was that, if gp350/220 is not absolutely required, then its presence may impede access of other essential proteins to the cell surface. Antibodies to gp350/220, an abundant and immunogenic protein, can overcome this in the saliva of an immune host to facilitate reinfection of an epithelial cell or to facilitate transmission through the epithelium in a new host.

There are long-standing observations that epithelial infection is more efficient if epithelial cells are cocultivated with B cells that are making the virus (19). A novel approach to investigation of the phenomenon demonstrated that binding EBV to a B cell before adding both to epithelial cells significantly enhanced infection of the epithelial cells, even in the absence of internalization and infection of the B cell (51). The mechanism of this B-cell transfer is uncertain but intriguing. The authors also suggested that gp350/220 is inhibitory to direct epithelial cell infection and proposed that the gp350/220 interaction with CR2 on the B cells unmasks envelope components necessary for epithelial cell infection. This parallels the hypothesis for enhanced infection in the presence of antibodies to gp350/220 in saliva. Neither study addressed what the envelope components necessary for efficient epithelial cell infection might be, but one likely candidate would be gB, the protein that is needed in higher amounts for epithelial than for B cell fusion.