Second derivative absorption spectra are reported for the aa3-cytochrome c oxidase from bovine cardiac mitochondria, the aq-600 ubiquinol oxidase from Bacillus subtilis, the ba3-cytochrome c oxidase from Thermus thermophilis, and the aco-cytochrome c oxidase from Bacillus YN-2000. Together these enzymes provide a range of cofactor combinations that allow us to unequivocally identify the origin of the 450-nm absorption band of the terminal oxidases as the 6-coordinate low-spin heme, cytochrome a. The spectrum of the aco-cytochrome c oxidase further establishes that the split Soret band of cytochrome a, with features at 443 and 450 nm, is common to all forms of the enzyme containing ferrocytochrome a and does not depend on ligand occupancy at the other heme cofactor as previously suggested. To test the universality of this Soret band splitting for 6-coordinate lowspin heme A systems, we have reconstituted purified heme A with the apo forms of the heme binding proteins, hemopexin, histidine-proline-rich glycoprotein and the H64VIV68H double mutant of human myoglobin. All 3 proteins bound the heme A as a (bis)histidine complex, as judged by optical and resonance Raman spectroscopy.In the ferroheme A forms, none of these proteins displayed evidence of Soret band splitting. Heme A-(bis)imidazole in aqueous detergent solution likewise failed to display Soret band splitting. When the cyanide-inhibited mixedvalence form of the bovine enzyme was partially denatured by chemical or thermal means, the split Soret transition of cytochrome a collapsed into a single band at 443 nm. Taken together these data suggest that the observation of Soret splitting, including a feature at 450 nm, results from specific protein-cofactor interactions that are unique to the cytochrome a-binding pocket of the terminal oxidases. The conservation of this unique binding pocket among evolutionarily distant species may reflect some mechanistic significance for this structure.