In transformed tobacco (Nicotiana tabacum) plastids, we flank the marker genes with recombinase target sites to facilitate their posttransformation excision. The P1 phage loxP sites are identical 34-bp direct repeats, whereas the phiC31 phage attB/attP sites are 54- and 215-bp sequences with partial homology within the 54-bp region. Deletions in the plastid genome are known to occur by recombination between directly repeated sequences. Our objective was to test whether or not the marker genes may be lost by homologous recombination via the directly repeated target sites in the absence of site-specific recombinases. The sequence between the target sites was the bar(au) gene that causes a golden-yellow (aurea) leaf color, so that the loss of the bar(au) gene can be readily detected by the appearance of green sectors. We report here that transplastomes carrying the bar(au) gene marker between recombinase target sites are relatively stable because no green sectors were detected in approximately 36,000 seedlings (Nt-pSS33 lines) carrying attB/attP-flanked bar(au) gene and in approximately 38,000 seedlings (Nt-pSS42 lines) carrying loxP-flanked bar(au) gene. Exceptions were six uniformly green plants in the Nt-pSS42-7A progeny. Sequencing the region of plastid DNA that may derive from the vector indicated that the bar(au) gene in the six green plants was lost by gene conversion using wild-type plastid DNA as template rather than by deletion via directly repeated loxP sites. Thus, the recombinase target sites incorporated in the plastid genome for marker gene excisions are too short to mediate the loss of marker genes by homologous recombination at a measurable frequency.