The biogenesis and maintenance of the entire endomembrane system is dependent upon membrane fusion proteins. Mounting evidence indicates that the integral membrane GTPase Atlastin is a membrane fusion protein involved in the homotypic fusion of the endoplasmic reticulum (ER) membrane suggesting a role in the biogenesis and maintenance of ER structure. I helped show that recombinant Drosophila atlastin is able to promote the fusion of synthetic membranes in vitro and that this fusion is dependent upon atlastin GTPase activity. The structure-function experiments presented here assist in elucidating domains required in the mechanism of atlastin mediated membrane fusion. ER homotypic fusion is dependent upon the self-association of Atlastin subunits in adjacent membranes to bring the bilayers into close molecular contact. Atlastin dimerization occurs in the presence of GTPγS but not GDP and stable dimerization is dependent upon a juxtamembrane middle domain three-helix bundle (3HB). The atlastin GTPase domain and 3HB form a potent soluble domain inhibitor of atlastin homotypic fusion, while the GTPase domain alone shows little inhibition. Designed GTPase domain mutations show that GTP binding and atlastin dimerization is insufficient to support fusion without GTP hydrolysis. Additionally, domain analysis of atlastin reveals that the C-terminal cytoplasmic domain of atlastin is absolutely required for membrane fusion, possibly through a protein-lipid interaction of an amphipathic alpha-helix. Genetic lesions in the human Atlastin-1 gene, SPG3A, result in a form of autosomal dominant hereditary spastic paraplegia (HSP). A better understanding of Atlastin function should lend significant insight into normal ER biogenesis and maintenance, as well as the pathology of human disease.