We discuss the suggestion that differences in the nucleotide composition between plastid and nuclear genomes may provide a selective advantage in the transposition of genes from plastid to nucleus. We show that in the adenine, thymine (AT)–rich genome of Borrelia burgdorferi several genes have an AT–content lower than the average for the genome as a whole. However, genes whose plant homologues have moved from plastid to nucleus are no less AT–rich than genes whose plant homologues have remained in the plastid, indicating that both classes of gene are able to support a high AT–content. We describe the anomalous organization of dinoflagellate plastid genes. These are located on small circles of 2–3 kbp, in contrast to the usual plastid genome organization of a single large circle of 100–200 kbp. Most circles contain a single gene. Some circles contain two genes and some contain none. Dinoflagellate plastids have retained far fewer genes than other plastids. We discuss a similarity between the dinoflagellate minicircles and the bacterial integron system.