1. Introduction

Despite the diversity in the organisms and cell types that utilize cell fusion in their normal physiology and pathology, the fusion reactions share common features. All the fusion processes can be divided into steps: aggregation of the membranes, lipid bilayers immediate contact, rearrangement of outer lipids resulting in the formation of a stalk, stalk expansion yielding the hemifusion diaphragm, fusion pore formation, and pore expansion ^[1][7]. Facilitating these sequential events are a broad array of fusogens. Due to the large difference of the fusing membranes, some fusogens mediate fusion by presenting on only one of the fusing membranes, termed unilateral mechanism, while others present on both membranes, termed bilateral mechanism. In addition, fusion events are classified into homotypic or heterotypic; in a homotypic fusion, it is mediated by the interaction of the same type of protein (e.g., Influenza HA2), and in heterotypic fusion, it is mediated by different proteins (e.g., N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) complex and myomaker-myomerger). To initiate membrane fusion, the fusogens have evolved to contain specific regions, such as hydrophobic motifs and residues, that aid in membrane curvature, lipid rearrangement and mixing, and pore formation.

The first fusogens identified were the viral fusogens encoded by enveloped viruses, which play a vital role in merging membranes to mediate viral entry into target cells. A detailed description on the structure of the different classes of enveloped viral fusogens and the mechanism by which they mediate virus–cell fusion are explained in other reviews ^[2][3][4][8–10]. Briefly, enveloped viral fusogens are divided into three types based on structural discrepancies: (1) class I viral fusogens prominently contain α-helixes, with their fusion peptides at or near the N-terminus; (2) class II viral fusogens differ from class I fusogens by the primary presence of β-sheets structure with internal fusion peptides formed as loops at the tips of β-strands; and (3) class III viral fusogens are comprised of both α-helices and β-sheets along with an internal hydrophobic fusion loop for lipid interaction ^[2][4][8,10]. Enveloped viral fusogens mediate heterotypic interaction (viral and cell membrane) and are present unilaterally.

Another set of well-studied fusogens are the cell–cell fusion proteins, including syncytin, EFF-1, and HAP-2 that primarily mediate developmental fusion. The syncytin is an endogenous retroviral envelope protein required for cytotrophoblast fusion during placenta development ^[5][11]. The N- and C-terminal heptad repeats region of syncytin shared 44% and 62% sequence identity with the corresponding N-peptide and C-peptide of HIV gp160, respectively, indicating the similar structural profiling between the syncytin and the class I viral fusogens ^[6][12]. Besides analogous structure, syncytins present unilaterally, just like class I viral fusogens, and induce fusion of placental cytotrophoblast in vivo ^[5][11]. On the other hand, the somatic fusogen EFF-1 and the sexual gamete fusogen HAP2 resemble class II viral fusogens, containing three β-sheet-rich domains ^{[7][8][9][10]}[13–16]. EFF-1 mediate homotypic or heterotypic bilateral membrane fusion, while the HAP2 mediates fusion bilaterally or unilaterally depending on the organism ^{[11][12][13][14][15]}[17–21]. Given that the three cell–cell fusogens possess structural similarities with viral fusogens, we deduce that these fusogens might have descended from a common ancestor and evolved into divergent fusogens acting to achieve distinct functional outcomes (heterotypic or homotypic fusion) with discrepant modalities of actions (unilateral or bilateral).

2. Intravesicular Fusogen: SNARE Family

The SNARE complex is comprised of three proteins—synaptosomal-associated protein-25 (SNAP-25), syntaxin, and synaptobrevin (vesicle-associated membrane protein, VAMP). The molecular weight of the SNARE protein superfamily members ranges from ~20 kDa to 30 kDa, with approximately 60 members present in yeasts and mammals ^[16][22]. Syntaxin and SNAP-25 are t-SNAREs contributed by the target membrane, while synaptobrevin is a v-SNARE from the vesicular membrane. All SNARE proteins contain a common SNARE motif and structurally divergent N- and C-termini (Figure 1A). Specifically, syntaxin is composed of an N-terminal domain, a conserved SNARE motif, and inserted into the membrane via a transmembrane domain (TMD) at the C-terminus. Synaptobrevin contains either a longin domain or a short unstructured peptide at N-terminus, followed by the SNARE motif and a transmembrane domain anchored on the membrane. SNAP-25 family proteins lack a transmembrane domain: two SNARE motifs are linked by a linker region and then inserted into the membrane by palmitoylation modification ^[17][18][19][23–25]. SNAREs appear to lie at the center of the membrane fusion mediating vesicle fusion within the endomembrane system and the vesicle exocytosis, which includes the ER, the Golgi, endosomes, and lysosomes ^[20][21][26,27]. SNAREs mediate fusion through the trans-SNARE complex that, “zipping” from the distal N-terminal region to the proximal C-terminal region, brings the two opposing membranes closer and eventually completing the fusion of the membranes ^[22][23][24][28–30].

2. ER-Shaping Protein: Atlastin

In addition to intracellular vesicle fusion mediated by SNAREs, atlastin (ATL) is an ER-shaping protein that belongs to the dynamin superfamily of GTPases primarily responsible for generating and maintaining the unique reticular morphology of the ER [59]. Knockdown or overexpression of dominant-negative form of atlastin results in the deformation of Golgi and ER morphology and disrupts tubular connection [60,61]. In addition, mutations of atlastin dramatically reduced neuronal ER tubules in dendrites of Caenorhabditis elegans sensory neuron, PVD. This speaks to the requirement of atlastin to mediate ER fusion and proper organization of the ER network [62]. Atlastin contains a cytosolic GTPase domain at the N-terminus, followed by a three-helix bundle (3HB, middle domain), two closely spaced transmembrane domains, and a cytosolic amphipathic helix at the C-terminus [63] (Figure 1B). Atlastins distribute on both ER membranes, mediating membranes fusion that beginning with dimerization triggered by GTP binding. When GTP is hydrolyzed by the dimer, the GTPase domain and the middle domain undergo conformational changes that bring the two membranes together.

3.  ER-Shaping Protein: Atlastin

In addition to intracellular vesicle fusion mediated by SNAREs, atlastin (ATL) is an ER-shaping protein that belongs to the dynamin superfamily of GTPases primarily responsible for generating and maintaining the unique reticular morphology of the ER ^[25][59]. Knockdown or overexpression of dominant-negative form of atlastin results in the deformation of Golgi and ER morphology and disrupts tubular connection ^[26][27][60,61]. In addition, mutations of atlastin dramatically reduced neuronal ER tubules in dendrites of Caenorhabditis elegans sensory neuron, PVD. This speaks to the requirement of atlastin to mediate ER fusion and proper organization of the ER network ^[28][62]. Atlastin contains a cytosolic GTPase domain at the N-terminus, followed by a three-helix bundle (3HB, middle domain), two closely spaced transmembrane domains, and a cytosolic amphipathic helix at the C-terminus ^[29][63] (Figure 1B). Atlastins distribute on both ER membranes, mediating membranes fusion that beginning with dimerization triggered by GTP binding. When GTP is hydrolyzed by the dimer, the GTPase domain and the middle domain undergo conformational changes that bring the two membranes together.