Plastids evolved from free-living cyanobacteria through a process of primary endosymbiosis. The most widely accepted hypothesis derives three ancient lineages of primary plastids, i.e. those of glaucophytes, red algae and green plants, from a single cyanobacterial endosymbiosis. This hypothesis was originally predicated on the assumption that transformations of endosymbionts into organelles must be exceptionally rare because of the difficulty in establishing efficient protein trafficking between a host cell and incipient organelle. It turns out, however, that highly integrated endosymbiotic associations are more common than once thought. Among them is the amoeba Paulinella chromatophora, which harbours independently acquired cyanobacterial endosymbionts functioning as plastids. Sequencing of the Paulinella endosymbiont genome revealed an absence of essential genes for protein trafficking, suggesting their residence in the host nucleus and import of protein products back into the endosymbiont. To investigate this hypothesis, we searched the Paulinella endosymbiont genome for homologues of higher plant translocon proteins that form the import apparatus in two-membrane envelopes of primary plastids. We found homologues of Toc12, Tic21 and Tic32, but genes for other key translocon proteins (e.g. Omp85/Toc75 and Tic20) were missing. We propose that these missing genes were transferred to the Paulinella nucleus and their products are imported and integrated into the endosymbiont envelope membranes, thereby creating an effective protein import apparatus. We further suggest that other bacterial/cyanobacterial endosymbionts found in protists, plants and animals could have evolved efficient protein import systems independently and, therefore, reached the status of true cellular organelles.