Measurements with different chemically modified gramicidins in lipid bilayer membranes were used to discriminate between the dimneric rLD helix proposed by Urry and the dimeric parallel or antiparallel helices proposed by Veatch and Blout. Evidence for the 7rLD helix was obtained on the basis of the different actions of a negatively charged 0-pyromellitylgramicidin and a negatively charged N-pyromellityldesformylgramicidin on lipid bilayer membranes. O-Pyromellitylgramicidin forms ionic channels in lipid membranes when it is applied to both sides of the membrane. In contrast to unmodified gramicidin, O-pyromellitylgramicidin is inactive when it is applied only to one side of the membrane. N-Pyromellityldesformylgramicidin does not form ionic channels in lipid bilayer membranes whether it is applied to one or both sides of the membrane. These results support the view that the gramicidin channel is formed by two 7rLD helices. Dimer formation by head-to-head association of two 7rL D helices needs six intermolecular hydrogen bonds, which are located at the formyl end of the molecule and which occur deep within the lipid membrane. In the head-to-head associated 7rL,D helix the absence of the formyl group leads to an inactivation of the e tide, whereas in a parallel or antiparallel double-stranded Eefix the absence of the formyl group should have only minor effects.Gramicidin A is a linear pentadecapeptide consisting of 15 hydrophobic alternating D,L-amina acids. It induces, in natural membranes and artificial lipid membranes, a high permeability for small monovalent cations. There is strong evidence that the peptide forms pore-like channels (1-12). For the structure of the channel two proposals exist. One possibility is a head-to-head association (formyl end to forrnyl end) of two helical monomers, where the helix of the monomer is a 7rLD-helix (7, 13). The helix in the dimeric form is stabilized by 22 intramolecular and 6 intermolecular hydrogen bonds. Urry et al. (7) proposed four types of 7rL,D helices, where the xr6 Dhelix seems to be the most probable. With the length of that dimer (30 A) and the diameter of the hydrophilic hole (4 A), the best geometrical dimensions are obtained to explain the results gotten in the membrane experiments. The other possible lXL,D helices either are too short (r8,D and lrLD helices) to form a transmembrane channel or the pore size is too small (1.8 A diameter for -rD L helix).An alternate model, a double-stranded helix, in which both peptide chains are coiled. around a common axis, has been proposed by Veatch et al. (14,15). In contrast to the 7rL,D helices, these helices have no intramolecular hydrogen bonds but 20-28 intermolecular hydrogen bonds (14). A common feature of both models is the existence of a narrow pore along the axis of the dimer that is lined with oxygen atoms of the peptide carbonyls. The essential difference between both models is that the lrLD helices dimerize when the formyl group is intact so that the intermolecular hydrogen bonds can form a bridg...