The gene for the Thermotoga maritima Trk potassium transporter component TrkA was originally thought to be a frameshift mutation and not to encode a functional protein. However, expression from this gene yielded a complex consisting of two distinct proteins designated TM1088A and -B. Genetic complementation of Escherichia coli mutants unable to transport potassium suggests that TM1088A/B is part of a functional Trk potassium transporter complex with the membrane protein TM1089. The protein structure for TM1088A shows a characteristic Rossmann fold indicating an NAD+ binding site and has structural similarity to potassium channel-related proteins. Ligand binding studies indicated that ATP, ADP, and AMP stabilized TM1088A to a much greater degree than NADH and NAD, consistent with the crystal structure of TM1088A, which contains a bound AMP natural ligand at the characteristic GXGXXG nucleotide binding site. Mutation of single and all glycines at this nucleotide binding site eliminated in vitro protein stabilization by the ligand, yet these mutated proteins could still functionally complement the E. coli potassium uptake mutants. We predict that this new two-subunit class of TrkA proteins is present in a number of organisms. A further subclass of the predicted two-subunit TrkA proteins lack an identifiable membrane-spanning subunit of the Trk K+ transporter. This class, as exemplified by Mycobacterium tuberculosis, did not complement E. coli potassium transport with the native E. coli TrkH; thus, it may require a novel TrkH-like protein for activity or provide an alternate function in vivo.