Gating movements of colicin A and colicin Ia are different.

Both colicin A and colicin Ia belong to a subfamily of the bacterial colicins that act by forming a voltage-dependent channel in the inner membrane of target bacteria. Both colicin A and Ia open at positive and close at negative potential, but only colicin A exhibits distinctly biphasic turnoff kinetics, implying the existence of two open states. Previous work has shown that Colicin Ia gating is associated with the translocation of a region representing 4 of its alpha helices across the membrane. Also, if its C-terminal, channel-forming domain is detached from the other domains, its N-terminal alpha helix can now also cross the membrane, causing the conductance to drop by a factor of about 6. Colicin A gating also involves the translocation of an internal domain, but we find that its translocated domain is somewhat smaller than that of Ia. Furthermore, while its isolated C-terminal domain can also undergo a transition to a smaller conductance, the conductance change is only about 15%, and the transition does not involve the translocation of the N-terminal alpha helix. Trapping the N-terminus on the cis side prevents neither this small conductance transition nor the biphasic turn-off. So, while the gating of both channels involves large, currently inexplicable conformational changes, these motions are qualitatively different in the two proteins, which may be a reflection of the dissimilar kinetics of closing.

Protein translocation across planar bilayers by the colicin Ia channel-forming domain: where will it end?

Colicin Ia, a 626-residue bactericidal protein, consists of three domains, with the carboxy-terminal domain (C domain) responsible for channel formation. Whole colicin Ia or C domain added to a planar lipid bilayer membrane forms voltage-gated channels. We have shown previously that the channel formed by whole colicin Ia has four membrane-spanning segments and an approximately 68-residue segment translocated across the membrane. Various experimental interventions could cause a longer or shorter segment within the C domain to be translocated, making us wonder why translocation normally stops where it does, near the amino-terminal end of the C domain (approximately residue 450). We hypothesized that regions upstream from the C domain prevent its amino-terminal end from moving into and across the membrane. To test this idea, we prepared C domain with a ligand attached near its amino terminus, added it to one side of a planar bilayer to form channels, and then probed from the opposite side with a water-soluble protein that can specifically bind the ligand. The binding of the probe had a dramatic effect on channel gating, demonstrating that the ligand (and hence the amino-terminal end of the C domain) had moved across the membrane. Experiments with larger colicin Ia fragments showed that a region of more than 165 residues, upstream from the C domain, can also move across the membrane. All of the colicin Ia carboxy-terminal fragments that we examined form channels that pass from a state of relatively normal conductance to a low-conductance state; we interpret this passage as a transition from a channel with four membrane-spanning segments to one with only three.

Translocation of inserted foreign epitopes by a channel-forming protein.

Certain bacterial protein toxins are able to insert themselves into, and at least partially across, lipid bilayer membranes in the absence of any auxiliary proteins, by using unknown mechanisms to overcome the high energy barrier presented by the hydrophobic bilayer core. We have previously shown that one such toxin, colicin Ia, translocates a large, hydrophilic part of itself completely across a lipid bilayer in conjunction with the formation of an ion-conducting channel. To address the question of whether the colicin can translocate any arbitrary amino acid sequence, we have altered the translocated segment by inserting, singly, two different foreign epitopes. Colicins containing either epitope retain significant bactericidal activity and form channels of normal conductance in planar bilayers. Furthermore, antibodies added on the side of the bilayer opposite that to which the colicin was added interact specifically with the corresponding epitopes, producing an inhibition of channel closing. Thus, the inserted epitopes are translocated along with the rest of the segment, suggesting that a surprisingly small part of colicin Ia, located elsewhere in the molecule, acts as a nonspecific protein translocator.

Transmembrane insertion of the colicin Ia hydrophobic hairpin.

Colicin Ia is a bactericidal protein that forms voltage-dependent, ion-conducting channels, both in the inner membrane of target bacteria and in planar bilayer membranes. Its amino acid sequence is rich in charged residues, except for a hydrophobic segment of 40 residues near the carboxyl terminus. In the crystal structure of colicin Ia and related colicins, this segment forms an alpha-helical hairpin. The hydrophobic segment is thought to be involved in the initial association of the colicin with the membrane and in the formation of the channel, but various orientations of the hairpin with respect to the membrane have been proposed. To address this issue, we attached biotin to a residue at the tip of the hydrophobic hairpin, and then probed its location with the biotin-binding protein streptavidin, added to one side or the other of a planar bilayer. Streptavidin added to the same side as the colicin prevented channel opening. Prior addition of streptavidin to the opposite side protected channels from this effect, and also increased the rate of channel opening; it produced these effects even before the first opening of the channels. These results suggest a model of membrane association in which the colicin first binds with the hydrophobic hairpin parallel to the membrane; next the hairpin inserts in a transmembrane orientation; and finally the channel opens. We also used streptavidin binding to obtain a stable population of colicin molecules in the membrane, suitable for the quantitative study of voltage-dependent gating. The effective gating charge thus determined is pH-independent and relatively small, compared with previous results for wild-type colicin Ia.

Major transmembrane movement associated with colicin Ia channel gating.

Colicin Ia, a bacterial protein toxin of 626 amino acid residues, forms voltage-dependent channels in planar lipid bilayer membranes. We have exploited the high affinity binding of streptavidin to biotin to map the topology of the channel-forming domain (roughly 175 residues of the COOH-terminal end) with respect to the membrane. That is, we have determined, for the channel's open and closed states, which parts of this domain are exposed to the aqueous solutions on either side of the membrane and which are inserted into the bilayer. This was done by biotinylating cysteine residues introduced by site-directed mutagenesis, and monitoring by electrophysiological methods the effect of streptavidin addition on channel behavior. We have identified a region of at least 68 residues that flips back and forth across the membrane in association with channel opening and closing. This identification was based on our observations that for mutants biotinylated in this region, streptavidin added to the cis (colicin-containing) compartment interfered with channel opening, and trans streptavidin interfered with channel closing. (If biotin was linked to the colicin by a disulfide bond, the effects of streptavidin on channel closing could be reversed by detaching the streptavidin-biotin complex from the colicin, using a water-soluble reducing agent. This showed that the cysteine sulfur, not just the biotin, is exposed to the trans solution). The upstream and downstream segments flanking the translocated region move into and out of the bilayer during channel opening and closing, forming two transmembrane segments. Surprisingly, if any of several residues near the upstream end of the translocated region is held on the cis side by streptavidin, the colicin still forms voltage-dependent channels, indicating that a part of the protein that normally is fully translocated across the membrane can become the upstream transmembrane segment. Evidently, the identity of the upstream transmembrane segment is not crucial to channel formation, and several open channel structures can exist.

Charge selectivity of the designed uncharged peptide ion channel Ac-(LSSLLSL)3-CONH2.

Charge selectivity in ion channel proteins is not fully understood. We have studied charge selectivity in a simple model system without charged groups, in which an amphiphilic helical peptide, Ac-(Leu-Ser-Ser-Leu-Leu-Ser-Leu)3-CONH2, forms ion channels across an uncharged phospholipid membrane. We find these channels to conduct both K+ and Cl-, with a permeability ratio (based on reversal potentials) that depends on the direction of the KCl concentration gradient across the membrane. The channel shows high selectivity for K+ when [KCl] is lowered on the side of the membrane that is held at a positive potential (the putative C-terminal side), but only modest K+ selectivity when [KCl] is lowered on the opposite side (the putative N-terminal side). Neither a simple Nernst-Planck electrodiffusion model including screening of the helix dipole potential, nor a multi-ion, state transition model allowing simultaneous cation and anion occupancy of the channel can satisfactorily fit the current-voltage curves over the full range of experimental conditions. However, the C-side/N-side dilution asymmetry in reversal potentials can be simulated with either type of model.

A helical-dipole model describes the single-channel current rectification of an uncharged peptide ion channel.

We are designing simple peptide ion channels as model systems for the study of the physical principles controlling conduction through ion-channel proteins. Here we report on an uncharged peptide, Ac-(Leu-Ser-Ser-Leu-Leu-Ser-Leu)3-CONH2, designed to form an aggregate of parallel, amphiphilic, membrane-spanning alpha-helices around a central water-filled pore. This peptide in planar lipid bilayers forms ion channels that show single-channel current rectification in symmetric 1 M KCl. The current at a given holding membrane potential is larger than the current measured through the same channel when the potential is reversed. Based on our hypothesized gating mechanism, the larger currents flow from the peptide carboxyl terminus toward the amino terminus. We present an ionic electrodiffusion model based on the helical-dipole potential and the dielectric interfacial polarization energy, which with reasonable values for dipole magnitude and dielectric constants, accurately replicates the current-voltage data.

Separating figure from ground with a parallel network.

The differentiation of figure from ground plays an important role in the perceptual organization of visual stimuli. The rapidity with which we can discriminate the inside from the outside of a figure suggests that at least this step in the process may be performed in visual cortex by a large number of neurons in several different areas working together in parallel. We have attempted to simulate this collective computation by designing a network of simple processing units that receives two types of information: bottom-up input from the image containing the outlines of a figure, which may be incomplete, and a top-down attentional input that biases one part of the image to be the inside of the figure. No presegmentation of the image was assumed. Two methods for performing the computation were explored: gradient descent, which seeks locally optimal states, and simulated annealing, which attempts to find globally optimal states by introducing noise into the computation. For complete outlines, gradient descent was faster, but the range of input parameters leading to successful performance was very narrow. In contrast, simulated annealing was more robust: it worked over a wider range of attention parameters and a wider range of outlines, including incomplete ones. Our network model is too simplified to serve as a model of human performance, but it does demonstrate that one global property of outlines can be computed through local interactions in a parallel network. Some features of the model, such as the role of noise in escaping from nonglobal optima, may generalize to more realistic models.