Chloroplast glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is composed of two different subunits, GAPA and GAPB, which are encoded in the nucleus by two related genes of eubacterial origin. In the present work the genes encoding chloroplast GAPA and GAPB from pea have been cloned and sequenced. The gene for GAPB is split by eight introns. Two introns interrupt the region encoding the transit peptide and six are found within the region encoding the mature subunit, four of which are in identical or similar positions relative to genes for cytosolic GAPDH of eukaryotic organisms. As opposed to this, the gene encoding pea GAPA has only two introns in the region encoding the mature subunit. These findings strongly support the "intron early" hypothesis and suggest that the low number of introns 'in the gene for chloroplast GAPA is due to differential loss of introns during the streamlining period of the chloroplast genome following the GAPB/GAPA separation. We deduce from this that eubacteria and chloroplasts contained GT-AG introns until relatively recently and that the duplication event leading to the genes encoding GAPB and GAPA and their respective transit peptides occurred in the chloroplast progenitor prior to the successive transfer and functional reintegration of these genes into the nuclear environment. These conclusions imply that GAPA/GAPB transit peptides are of eubacterial origin.The endosymbiotic hypothesis, suggesting that chloroplasts may be descendants of engulfed prokaryotes related to present-day cyanobacteria, dates back to an idea of Schimper in 1883 (1). It has been explicitly developed during the first two decades of this century, has later been extended to mitochondria, and has received fresh impetus in more modem times by the elaborated discussions of Margulis and other evolutionary thinkers (see refs. 2 and 3 also for historical aspects). With the advent of macromolecular sequencing techniques, evidence in favor of the symbiotic origin of chloroplasts rapidly accumulated, mainly in the form of sequence data demonstrating a relatively close relationship between chloroplast-specific genes and genes of (cyano)bacteria (refs. 4-8; for reviews see refs. 9 and 10). A recent highlight has been the determination of the complete nucleotide sequences of the chloroplast genomes from three evolutionarily distant land plants, tobacco, rice, and liverwort (for a review see ref. 11).While the genetic content of the chloroplast genome has been completely elucidated for several species, comparatively little is known about the structure and evolution of most nuclear genes encoding chloroplast components. There are hundreds of such genes (12) coding for enzymes of various metabolic pathways and for proteins of thylakoid and envelope membranes. Most of these genes must have existed in the eubacterial chloroplast progenitor, and during the course of plant cell evolution they either were transferred into the nucleus or were replaced by nuclear gene functions following duplications or modifications ofpree...