The relationship between codon boundaries and multiple reading-frame preferences: coding organization of bacterial insertion sequences.

Theoretical considerations have shown that the five possible overlapping reading-frame configurations differ significantly in their coding flexibility and thus in their information content (Siegel and Fitch 1980; Smith and Waterman 1980). Contrary to expectation, the overlapping frame configuration allowing the greatest coding flexibility is rarely seen, whereas one of the most constraining is common. We point out here that this overlapping reading-frame paradox and an observed but unexplained preference in coding regions for a pyrimidine-purine at codon boundaries (Shepherd 1981; Jones and Kafatos 1982; Smith et al. 1983) are intimately linked. The codon boundary preference, which may be related to translation efficiency or accuracy, places constraints on the evolution of overlapping coding regions. These considerations may help identify actual coding regions in DNA sequences. We have analyzed five sequenced (enteric) bacterial insertion sequences for codon boundary incidences and reading-frame configurations and find that they are consistent with these proposed constraints.