Members of the voltage-gated-like ion channel superfamily have a conserved pore structure. Transmembrane helices that line the pore (M2 or S6) are thought to gate it at the cytoplasmic end by bending at a hinge glycine residue. Proline residues favor bending of alpha-helices, and substitution of proline for this glycine (G219) dramatically stabilizes the open state of a bacterial Na(+) channel NaChBac. Here we have probed S6 pore-lining residues of NaChBac by proline mutagenesis. Five of 15 proline-substitution mutants yielded depolarization-activated Na(+) channels, but only G219P channels have strongly negatively shifted voltage dependence of activation, demonstrating specificity for bending at G219 for depolarization-activated gating. Remarkably, three proline-substitution mutations on the same face of S6 as G219 yielded channels that activated upon hyperpolarization and inactivated very slowly. Studies of L226P showed that hyperpolarization to -147 mV gives half-maximal activation, 123 mV more negative than WT. Analysis of combination mutations and studies of block by the local anesthetic etidocaine favored the conclusion that hyperpolarization-activated gating results from opening of the cytoplasmic gate formed by S6 helices. Substitution of multiple amino acids for L226 indicated that hyperpolarization-activated gating was correlated with a high propensity for bending, whereas depolarization-activated gating was favored by a low propensity for bending. Our results further define the dominant role of bending of S6 in determining not only the voltage dependence but also the polarity of voltage-dependent gating. Native hyperpolarization-activated gating of hyperpolarization- and cyclic nucleotide-gated (HCN) channels in animals and KAT channels in plants may involve bending at analogous S6 amino acid residues.