Attachment proteins are responsible for the initial interaction between two membranes. In virus-cell fusion, virus-receptor interactions are mediated by carbohydrate moieties or cell adhesion domains on proteins or other molecules in the plasma membrane (Bomsel and Alfsen, 2003; Dimitrov, 2004). In many cases, multiple proteins participate in a single virus-cell attachment event, producing a complex interaction that occurs in a limited time frame. Several putative cell-cell attachment proteins share these characteristics. The best-characterized candidates are the four immunoglobulin (Ig)-superfamily members involved in Drosophila myoblast fusion: Sns, Hbs, Duf and Rst (Taylor, 2002). Each of these proteins contains several Ig-like domains, which are well-defined cell-cell adhesion domains. Null mutations in the sns or rst and duf gene, or overexpression of hbs, leads to defects in myoblast fusion at the attachment stage. Cells expressing Duf aggregate in vitro with cells expressing Sns and Hbs, and it is likely that these proteins function cooperatively in attachment in vivo. In Chlamydomonas, fus1 mutants are unable to attach to the mating process of the opposite mating type and fusion is never observed (Misamore et al., 2003). Fus1 is a single-pass transmembrane protein sequence that has similarity within five Ig-like repeats to bacterial invasins and intimins, which mediate the adhesion step that precedes bacterial invasion of host cells. The Caenorhabditis elegans sperm protein SPE-9 contains ten epidermal growth factor (EGF) repeats in its extracellular domain and is likely to act as an attachment factor in gamete fusion (Singson et al., 1998). All of these proteins contain cell adhesion domains, and optimal function of the myoblast fusion candidates requires the participation of several proteins. The similarities between cell-cell and viral-cell attachment proteins probably stem from their common function and might provide predictive criteria with which to evaluate candidates as for sperm-egg attachment proteins.
Fusion proteins directly mediate the mixing of two membrane bilayers. In virus-cell fusion, a fusion protein typically contains a single transmembrane domain and a fusion peptide – a sequence of 10-30 residues that form an amphiphilic domain at the N-terminus (class I) or within the protein (class II) that is crucial for fusion (Chernomordik and Kozlov, 2003; Jahn et al., 2003). Putative cell-cell fusion proteins are diverse in structure and few match the portrait of a canonical viral fusion protein. Whether this is because an alternative fusion mechanism is used for each case of cell-cell fusion or because these particular proteins do not represent the fusion protein, per se, is not yet clear. A strong candidate for a `true' fusion protein is the C. elegans protein EFF-1, which is necessary (Mohler et al., 2002) and sufficient (Shemer et al., 2004) for most epithelial cell fusion events in the developing worm. EFF-1 is unique in cell-cell fusion systems because it contains a putative fusion peptide within the protein and is thus similar to a class II viral fusagen (Shemer and Podbilewicz, 2003). In Saccharomyces cerevisiae, an absence of the pentaspan protein Prm1p on both mating types yields closely apposed (separated by 8 nm), but unfused, membranes in more than half of the mating pairs (Heiman and Walter, 2000). In Drosophila myoblast fusion, the only proteins currently shown to have a direct role in fusion are cytoplasmic (Loner, Myoblast city and Rols/Ants) and are probably important for cytoskeletal rearrangement that occurs following attachment and prior to membrane fusion (Taylor, 2003).