Circulatory systems of vertebrate and invertebrate metazoans are very different. Large vessels of invertebrates are constituted of spaces and lacunae located between the basement membranes of endodermal and mesodermal epithelia, and they lack an endothelial lining. Myoepithelial differentation of the coelomic cells covering hemal spaces is a frequent event, and myoepithelial cells often form microvessels in some large invertebrates. There is no phylogenetic theory about the origin of the endothelial cells in vertebrates. We herein propose that endothelial cells originated from a type of specialized blood cells, called amoebocytes, that adhere to the vascular basement membrane. The transition between amoebocytes and endothelium involved the acquisition of an epithelial phenotype. We suggest that immunological cooperation was the earliest function of these protoendothelial cells. Furthermore, their ability to transiently recover the migratory, invasive phenotype of amoebocytes (i.e., the angiogenic phenotype) allowed for vascular growth from the original visceral areas to the well-developed somatic areas of vertebrates (especially the tail, head, and neural tube). We also hypothesize that pericytes and smooth muscle cells derived from myoepithelial cells detached from the coelomic lining. As the origin of blood cells in invertebrates is probably coelomic, our hypothesis relates the origin of all the elements of the circulatory system with the coelomic wall. We have collected from the literature a number of comparative and developmental data supporting our hypothesis, for example the localization of the vascular endothelial growth factor receptor-2 ortholog in hemocytes of Drosophila or the fact that circulating progenitors can differentiate into endothelial cells even in adult vertebrates.