Group I introns are mobile RNA enzymes (ribozymes) that encode conserved primary and secondary structures required for autocatalysis. The group I intron that interrupts the tRNA-Leu gene in cyanobacteria and plastids is remarkable because it is the oldest known intervening sequence and may have been present in the common ancestor of the cyanobacteria (i.e., 2.7-3.5 billion years old). This intron entered the eukaryotic domain through primary plastid endosymbiosis. We reconstructed the phylogeny of the tRNA-Leu intron and tested the in vitro self-splicing ability of a diverse collection of these ribozymes to address the relationship between intron stability and autocatalysis. Our results suggest that the present-day intron distribution in plastids is best explained by strict vertical transmission, with no intron losses in land plants or a subset of the Stramenopiles (xanthophyceae/phaeophyceae) and frequent loss among green algae, as well as in the red algae and their secondary plastid derivatives (except the xanthophyceae/phaeophyceae lineage). Interestingly, all tested land plant introns could not self-splice in vitro and presumably have become dependent on a host factor to facilitate in vivo excision. The host dependence likely evolved once in the common ancestor of land plants. In all other plastid lineages, these ribozymes could either self-splice or complete only the first step of autocatalysis.