Although Kch of Escherichia coli is thought to be a K(+) channel by sequence homology, there is little evidence that it actually conducts K(+) ions in vitro or in vivo. We isolated gain-of-function (GOF) Kch mutations that render bacteria specifically sensitive to K(+) ions. Millimolar added K(+), but not Na(+) or sorbitol, blocks the initiation or continuation of mutant growth in liquid media. The mutations are mapped at the RCK (or KTN) domain, which is considered to be the cytoplasmic sensor controlling the gate. Additional mutations directed to the K(+)-filter sequence rescue the GOF mutant. The apparent K(+)-specific conduction through the 'loose-cannon' mutant channel suggests that the wild-type Kch channel also conducts, albeit in a regulated manner. Changing the internal ATG does not erase the GOF toxicity, but removes kch's short second product, suggesting that it is not required for channel function in vivo. The mutant phenotypes are better explained by a perturbation of membrane potential instead of internal K(+) concentration. Possible implications on the normal function of Kch are discussed.