Currently there is a great interest in using scanning probe microscopy to study living cells. However, in most cases the contact the probe makes with the soft surface of the cell deforms or damages it. Here we report a scanning ion conductance microscope specially developed for imaging living cells. A key feature of the instrument is its scanning algorithm, which maintains the working distance between the probe and the sample such that they do not make direct physical contact with each other. Numerical simulation of the probe/sample interaction, which closely matches the experimental observations, provides the optimum working distance. The microscope scans highly convoluted surface structures without damaging them and reveals the true topography of cell surfaces. The images resemble those produced by scanning electron microscopy, with the significant difference that the cells remain viable and active. The instrument can monitor small-scale dynamics of cell surfaces as well as whole-cell movement.

The change in conductance of a small electrolyte-filled capillary owing to the passage of sub-micrometre-sized particles has long been used for particle counting and sizing. A commercial device for such measurements, the Coulter counter, is able to detect particles of sizes down to several tenths of a micrometre. Nuclepore technology (in which pores are etched particle tracks) has extended the lower limit of size detection to 60-nm particles by using a capillary of diameter 0.45 micron (ref. 4). Here we show that natural channel-forming peptides incorporated into a bilayer lipid membrane can be used to detect the passage of single molecules with gyration radii as small as 5-15 A. From our experiments with alamethicin pores we infer both the average number and the diffusion coefficients of poly(ethylene glycol) molecules in the pore. Our approach provides a means of observing the statistics and mechanics of flexible polymers moving within the confines of precisely defined single-molecule structures.

Membrane-bound proteinaceous nanoscale pores allow us to simultaneously observe the thermodynamic and kinetic properties of differently sized polymers within their confines. We determine the dynamic partitioning of poly(ethylene glycol) (PEG) into the pore formed by Staphylococcus aureus α-toxin and evaluate the free energy of polymer confinement by measuring polymer-induced changes to the pore's ionic conductance. The free energy deduced from the partition coefficient has a sharper dependence on polymer length (or weight) than scaling theory predicts. Moreover, the polymer-induced conductance fluctuations show a striking nonmonotonic dependence on the polymer molecular weight. The movement of polymer inside the pore is characterized by a diffusion coefficient that is orders of magnitude smaller than that for polymer in the bulk aqueous solution, which suggests that PEG has an attractive interaction with the pore. Using an ad-hoc approach, we show that a simple molecular weight-dependent modification of the polymer's diffusion coefficient accounts for these results, but only qualitatively. Given that PEG associates with hydrophobic regions in proteins, we also conclude that, contrary to the conventional view of ion channels, the aqueous cavity of the α-toxin pore's interior is, to some extent, hydrophobic.

With few exceptions, membrane lipids are usually regarded as a kind of filler or passive solvent for membrane proteins. Yet, cells exquisitely control membrane composition. Many phospholipids found in plasma membrane bilayers favor packing into inverted hexagonal bulk phases. It was suggested that the strain of forcing such lipids into a bilayer may affect membrane protein function, such as the operation of transmembrane channels. To investigate this, we have inserted the peptide alamethicin into bilayer membranes composed of lipids of empirically determined inverted hexagonal phase "spontaneous radii" Ro, which will have expectably different degrees of strain when forced into bilayer form. We observe a correlation between measured Ro and the relative probabilities of different conductance states. States of higher conductance are more probable in dioleoylphosphatidylethanolamine, the lipid of highest curvature, 1/Ro, than in dioleoylphosphatidylcholine, the lipid of lowest curvature.

Alamethicin, a 20-amino acid peptide, has been studied for a number of years as a model for voltage-gated channels. Recently both the x-ray structure of alamethicin in crystal and an NMR solution structure have been published (Fox and Richards, 1982. Bannerjee et al., 1983). Both structures show that the amino end of the molecule forms a stable alpha-helix nine or 10 residues in length and that the COOH-terminal ends exhibits a variable hydrogen bonding pattern. We have used synthetic analogues of alamethicin to test various hypotheses of its mode of action. As a result of these studies we propose a channel structure in which the COOH-terminal residues bond together as a beta-barrel, leaving the alpha- helices free to rotate under the influence of the electric field and gate the channel. Though the number of monomers per channel varies with experimental conditions, the gating charge per monomer stays close to that expected from an alpha-helical gate. We can also alter the sign of the voltage which turns on a channel by varying the charge on the alamethicin analogue. Channels are always slightly cation-selective even though formed by monomers with negative, positive, or zero formal charge. Channels are less stable in low ionic strength solutions than high. Finally, alamethicin conductance parameters vary systematically with changes in membrane thickness. We show how these results and others in the literature can be explained by a fairly detailed structural model. The model can be easily generalized to a form more suited to high molecular weight single-peptide-chain proteins.

Channel access resistance has been measured to estimate the characteristic size of a single ion channel. We compare channel conductance in the presence of nonpenetrating water-soluble polymers with that obtained for polymer-free electrolyte solution. The contribution of the access resistance to the total alamethicin channel resistance is approximately 10% for first three open channel levels. The open alamethicin channel radii inferred for these first three levels from the access resistance are 6.3, 10.3, and 11.4 A. The dependence of channel conductance on polymer molecular weight also allows evaluation of the channel dimensions from polymer exclusion. Despite varying conductance, it was shown that steric radii of the alamethicin channel at different conductance levels remain approximately unchanged. These results support a model of the alamethicin channel as an array of closely packed parallel pores of nearly uniform diameter.

Although the dynamics of cell membranes and associated structures is vital for cell function, little is known due to lack of suitable methods. We found, using scanning ion conductance microscopy, that microvilli, membrane projections supported by internal actin bundles, undergo a life cycle: fast height-dependent growth, relatively short steady state, and slow height-independent retraction. The microvilli can aggregate into relatively stable structures where the steady state is extended. We suggest that the intrinsic dynamics of microvilli, combined with their ability to make stable structures, allows them to act as elementary ''building blocks'' for the assembly of specialized structures on the cell surface.

We manipulate lipid bilayer surface charge and gauge its influence on gramicidin A channel conductance by two strategies: titration of the lipid charge through bulk solution pH and dilution of a charged lipid by neutral. Using diphytanoyl phosphatidylserine (PS) bilayers with CsCl aqueous solutions, we show that the effects of lipid charge titration on channel conductance are masked 1) by conductance saturation with Cs+ ions in the neutral pH range and 2) by increased proton concentration when the bathing solution pH is less than 3. A smeared charge model permits us to separate different contributions to the channel conductance and to introduce a new method for "bilayer pKa" determination. We use the Gouy-Chapman expression for the charged surface potential to obtain equilibria of protons and cations with lipid charges. To calculate cation concentration at the channel mouth, we compare different models for the ion distribution, exact and linearized forms of the planar Poisson-Boltzmann equation, as well as the construction of a "Gibbs dividing surface" between salt bath and charged membrane. All approximations yield the intrinsic pKain of PS lipid in 0.1 M CsCl to be in the range 2.5-3.0. By diluting PS surface charge at a fixed pH with admixed neutral diphytanoyl phosphatidylcholine (PC), we obtain a conductance decrease in magnitude greater than expected from the electrostatic model. This observation is in accord with the different conductance saturation values for PS and PC lipids reported earlier (, Biochim. Biophys. Acta. 552:369-378) and verified in the present work for solvent-free membranes. In addition to electrostatic effects of surface charge, gramicidin A channel conductance is also influenced by lipid-dependent structural factors.