We report that long-chain poly-L-glutamine forms cation-selective channels when incorporated into artificial planar lipid bilayer membranes. The channel was permeable to alkali cations and H(+) ions and virtually impermeable to anions; the selectivity sequence based on the single-channel conductance was H(+) >> Cs(+) > K(+) > Na(+). The cation channel was characterized by long-lived open states (often lasting for several minutes to tens of minutes) interrupted by brief closings. The appearance of the channel depended critically on the length of polyglutamine chains; ion channels were observed with 40-residue stretches, whereas no significant conductance changes were detected with 29-residue tracts. The channel-forming threshold length of poly-L-glutamine was thus between 29 and 40 residues. A molecular mechanics calculation suggests a mu-helix (. Biophys. J. 69:1130-1141) as a candidate molecular structure of the channel. The channel-forming nature of long-chain poly-L-glutamine may provide a clue to the elucidation of the pathogenetic mechanism of the polyglutamine diseases, a group of inherited neurodegenerative disorders including Huntington's disease.

Electrostatic calculation of the gramicidin channel is performed on the basis of a three-dielectric model in which the peptide backbone of the channel is added as a third dielectric region to the conventional two-dielectric channel model (whose pore radius is often referred to as the effective pore radius reff). A basic principle for calculating electrostatic fields in three-dielectric models is introduced. It is shown that the gramicidin channel has no unique value of reff. The reff with respect to the "self-image energy" (i.e., the image energy in the presence of a single ion) is 2.6-2.7 A, slightly depending upon the position of the ion (the least-square value over the whole length of the pore is 2.6 A). In contrast, the reff with respect to the electric potential due to an ion (and hence the reff with respect to the interaction energy between two ions) is dependent upon the distance s of separation; it ranges from 2.6 to greater than 5 A, increasing with an increase in s. However, for the purpose of rough estimation, the reff with respect to the self-image energy can also be used in calculating the electric potential and the interaction energy, because the error introduced by this approximation is an overestimation of the order of 30% at most. It is also shown that the apparent dielectric constant for the interaction between two charges depends markedly upon the positions of the charges. In the course of this study, the dielectric constant and polarizability of the peptide backbone in the beta-sheet structure is estimated to be 10 and 8.22 A3.

A search was made in terms of molecular mechanics calculation for tubular, or pore-forming, single-stranded helices of poly-L-amino acids. Such a helix was found in the vicinity of (phi, psi, omega) = 81 degrees, 98 degrees, 170 degrees) in the conformational space. It was the 6.2(20) helix of right-handedness. The 6.2(20) helix, here named the "mu helix," had a cylindrical pore along its helical axis. The diameter of the pore was 6.6 A on the basis of the atom centers of carbonyl carbons and amino nitrogens. The left-handed mu helix was less stable than the right-handed counterpart. The conformation energy of the mu helix, expressed relative to that of the alpha helix of the same polypeptide, depended to a great extent on amino acid composition. It varied over a range of a few kilocalories per mol per residue above and below the conformation energy of the alpha helix of the same polypeptide. This marked diversity in the relative conformation energy of the mu helix can be ascribed primarily to the difference in the relative position of alpha-carbons between the mu and the alpha helices. The conformational entropy made only a small contribution, if any, to the relative stability of the mu helix. There was a hydrogen-bonded network of side chains in the mu helices of poly-L-glutamine and poly-L-asparagine.

No abstract

Basic nuclear magnetic resonance (NMR) features of 23Na ions bound to the gramicidin channel (packaged into lecithin liposomes) were studied. The first binding constant K1 of Na+ was not significantly dependent on channel models employed. With the two-identical-site model (Model I), K1 was 13.7 (+/- 1.4) molal-1 (in the activity basis) at 25 degrees C; when the binding of a third ion was included (Model II), it was 13.0 (+/- 2.0) molal-1. The second binding constant K2 was model dependent; it was 1.6 (+/- 0.2) and 3-4 molal-1 for Models I and II, respectively. The rate constants, k-1 and k-2, of Na+ for exit from singly and doubly loaded channels, respectively, were 8 X 10(5) s-1 less than or equal to k-1 less than or equal to 3 X 10(6) s-1 and 8 X 10(5) s-1 less than or equal to k-2 less than or equal to 1.0 X 10(7) s-1 at 25 degrees C; the lower bound represents a rough approximation of k-1. The ratio k-2/k-1 was greater than one and did not greatly exceed 20. From the competition experiment, K1 of T1+ was 5.7 (+/- 0.6) X 10(2) molal-1. The longitudinal relaxation time T1 of bound 23Na in the state of single occupancy (T 1B sing) was virtually independent of models, 0.56 (+/- 0.03) and 0.55 (+/- 0.04) ms at 25 degrees C for Models I and II, respectively. For the state of double occupancy, T1 of bound 23Na (T 1B doub) was model dependent: 0.27 (+/- 0.01) and 0.4-0.6 ms for Models I and II. The correlation time tau c of bound 23Na was 2.2 (+/- 0.2) ns at 25 degrees C for single occupancy; tau c for double occupancy was not significantly different from this value. The estimated tau c was found to involve no appreciable contribution of the exchange of 23Na between the channel and the bulk solution. Thé quadrupole coupling constant chi was 1.0 (+/- 0.1) MHz for 23Na in single occupancy; chi for double occupancy was 0.9-1.4 MHz, depending on models. A lower bound of the average quadrupole coupling constant chi alpha was 0.13-0.26 MHz at 25 degrees C for 23Na in single occupancy; this value represents a rough approximation of chi alpha at this temperature. An argument based on the estimated chi alpha and the known conformation of the gramicidin channel suggests that the binding site is a small domain near the channel end. Within the framework of Model I, Tb was faster than Tljn; this inequality was attributed to an increased chi in the presence ofa second cation, which was not explained in terms of electrostatic interactions between bound cations, implying a conformation change upon binding of cations.