Eukaryotic cytochrome c oxidase (CcO) and homologous prokaryotic forms of Rhodobacter and Paraccocus differ in the EPR spectrum of heme a. It was noted that a histidine ligand of heme a (H102) is hydrogen bonded to serine in Rhodobacter (S44) and Paraccocus CcOs, in contrast to glycine in the bovine enzyme. Mutation of S44 to glycine shifts the heme a EPR signal from g(z) = 2.82 to 2.86, closer to bovine heme a at 3.03, without modifying other properties. Mutation to aspartate, however, results in an oppositely shifted and split heme a EPR signal of g(z) = 2.72/2.78, accompanied by lower activity and drastically inhibited intrinsic electron transfer from CuA to heme a. This intrinsic rate is biphasic; the proportion that is slow is pH dependent, as is the relative intensity of the two EPR signal components. At pH 8, the heme a EPR signal at 2.72 is most intense, and the electron transfer rate (CuA to heme a) is 10-130 s(-1), compared to wild-type at 90,000 s(-1). At pH 5.5, the signal at 2.78 is intensified, and a biphasic rate is observed, 50% fast (approximately wild type) and 50% slow (90 s(-1)). The data support the prediction that the hydrogen-bonding partner of the histidine ligand of heme a is one determinant of the EPR spectral difference between bovine and bacterial CcO. We further demonstrate that the heme a redox potential can be dramatically altered by a nearby carboxyl, whose protonation leads to a proton-coupled electron transfer process.