Fusion of phospholipid vesicles with planar phospholipid bilayer membranes. II. Incorporation of a vesicular membrane marker into the planar membrane

Fusion of multilamellar phospholipid vesicles with planar phospholipid bilayer membranes was monitored by the rate of appearance in the planar membrane of an intrinsic membrane protein present in the vesicle membranes. An essential requirement for fusion is an osmotic gradient across the planar membrane, with the cis side (the side containing the vesicles) hyperosmotic to the opposite (trans) side; for substantial fusion rates, divalent cation must also be present on the cis side. Thus, the low fusion rates obtained with 100 mM excess glucose in the cis compartment are enhanced orders of magnitude by the addition of 5-10 mM CaCl2 to the cis compartment. Conversely, the rapid fusion rates induced by 40 mM CaCl2 in the cis compartment are completely suppressed when the osmotic gradient (created by the 40 mM CaCl2) is abolished by addition of an equivalent amount of either CaCl2, NaCl, urea, or glucose to the trans compartment. We propose that fusion occurs by the osmotic swelling of vesicles in contact with the planar membrane, with subsequent rupture of the vesicular and planar membranes in the region of contact. Divalent cations catalyze this process by increasing the frequency and duration of vesicle-planar membrane contact. We argue that essentially this same osmotic mechanism drives biological fusion processes, such as exocytosis. Our fusion procedure provides a general method for incorporating and reconstituting transport proteins into planar phospholipid bilayer membranes.     

Analysis of displacement currents in lipid bilayer membranes

The time course of a charge transfer through solvent-containing lipid bilayer membranes was measured in the range of 5 microseconds and 100 s. It is shown that a displacement charge value is proportional to the area of a torus surrounding the lipid bilayer membrane, but not to the black membrane area.     

Parameters affecting the fusion of unilamellar phospholipid vesicles with planar bilayer membranes

It was previously shown (Cohen, F. S., J. Zimmerberg, and A. Finkelstein, 1980, J. Gen. Physiol., 75:251-270) that multilamellar phospholipid vesicles can fuse with decane-containing phospholipid bilayer membranes. An essential requirement for fusion was an osmotic gradient across the planar membrane, with the vesicle-containing (cis) side hyperosmotic with respect to the opposite (trans) side. We now report that unilamellar vesicles will fuse with "hydrocarbon-free" membranes subject to these same osmotic conditions. Thus the same conditions that apply to fusion of multilamellar vesicles with planar bilayer membranes also apply to fusion of unilamellar vesicles with these membranes, and hydrocarbon is not required for the fusion process. If the vesicles and/or planar membrane contain negatively charged lipids, divalent cation (approximately 15 mM Ca++) is required in the cis compartment (in addition to the osmotic gradient across the membrane) to obtain substantial fusion rates. On the other hand, vesicles made from uncharged lipids readily fuse with planar phosphatidylethanolamine planar membranes in the near absence of divalent cation with just an osmotic gradient. Vesicles fuse much more readily with phosphatidylethanolamine-containing than with phosphatidylcholine-containing planar membranes. Although hydrocarbon (decane) is not required in the planar membrane for fusion, it does affect the rate of fusion and causes the fusion process to be dependent on stirring in the cis compartment.     

The N-terminal half of the heavy chain of botulinum type A neurotoxin forms channels in planar phospholipid bilayers

The heavy chain of botulinum type A neurotoxin forms channels in planar phospholipid bilayer membranes. Channel activity is confined to the N-terminal half of this chain; the C-terminal half is inactive. Channel activity is stimulated by low pH (4.5-5.5) on the cis side (the side to which protein is added), neutral pH on the opposite (trans) side, and cis positive voltages. These findings are strikingly similar to those previously reported for analogous fragments of diphtheria and tetanus toxins.     

Electrical oscillation and fluctuation in phospholipid membranes. Phospholipids can form a channel without protein

Fluctuations and/or step-wise changes in membrane potential and electrical current were observed in bilayer membranes of dioleoylphosphatidylcholine (DOPC) in the absence of any channel protein. The DOPC membranes consisted of three types: black lipid membranes, pipette-clamp membranes and lipid membranes transferred to porous filter paper by conventional Langmuir-Blodgett techniques. This finding is significant since phospholipids are the main constituents of biomembranes. Lipid molecules with a cis double bond in their carbon skeleton are suggested to be important in the gating or excitation of biomembranes.     

Single streptomyces lividans K(+) channels: functional asymmetries and sidedness of proton activation.

Basic electrophysiological properties of the KcsA K(+) channel were examined in planar lipid bilayer membranes. The channel displays open-state rectification and weakly voltage-dependent gating. Tetraethylammonium blocking affinity depends on the side of the bilayer to which the blocker is added. Addition of Na(+) to the trans chamber causes block of open-channel current, while addition to the cis side has no effect. Most striking is the activation of KcsA by protons; channel activity is observed only when the trans bilayer chamber is at low pH. To ascertain which side of the channel faces which chamber, residues with structurally known locations were mapped to defined sides of the bilayer. Mutation of Y82, an external residue, results in changes in tetraethylammonium affinity exclusively from the cis side. Channels with cysteine residues substituted at externally exposed Y82 or internally exposed Q119 are functionally modified by methanethiosulfonate reagents from the cis or trans chambers, respectively. Block by charybdotoxin, known to bind to the channel's external mouth, is observed only when the toxin is added to the cis side of channels mutated to be toxin sensitive. These results demonstrate unambiguously that the protonation sites linked to gating are on the intracellular portion of the KcsA protein.     

Interaction of the Antimicrobial Peptide Polymyxin B1 with Both Membranes of E. coli: A Molecular Dynamics Study

Antimicrobial peptides are small, cationic proteins that can induce lysis of bacterial cells through interaction with their membranes. Different mechanisms for cell lysis have been proposed, but these models tend to neglect the role of the chemical composition of the membrane, which differs between bacterial species and can be heterogeneous even within a single cell. Moreover, the cell envelope of Gram-negative bacteria such as E. coli contains two membranes with differing compositions. To this end, we report the first molecular dynamics simulation study of the interaction of the antimicrobial peptide, polymyxin B1 with complex models of both the inner and outer membranes of E. coli. The results of >16 microseconds of simulation predict that polymyxin B1 is likely to interact with the membranes via distinct mechanisms. The lipopeptides aggregate in the lipopolysaccharide headgroup region of the outer membrane with limited tendency for insertion within the lipid A tails. In contrast, the lipopeptides readily insert into the inner membrane core, and the concomitant increased hydration may be responsible for bilayer destabilization and antimicrobial function. Given the urgent need to develop novel, potent antibiotics, the results presented here reveal key mechanistic details that may be exploited for future rational drug development.     

Video fluorescence microscopy studies of phospholipid vesicle fusion with a planar phospholipid membrane. Nature of membrane-membrane interactions and detection of release of contents

Video fluorescence microscopy was used to study adsorption and fusion of unilamellar phospholipid vesicles to solvent-free planar bilayer membranes. Large unilamellar vesicles (2-10 microns diam) were loaded with 200 mM of the membrane-impermeant fluorescent dye calcein. Vesicles were ejected from a pipette brought to within 10 microns of the planar membrane, thereby minimizing background fluorescence and diffusion times through the unstirred layer. Vesicle binding to the planar membrane reached a maximum at 20 mM calcium. The vesicles fused when they were osmotically swollen by dissipating a KCl gradient across the vesicular membrane with the channel-forming antibiotic nystatin or, alternatively, by making the cis compartment hyperosmotic. Osmotically induced ruptures appeared as bright flashes of light that lasted several video fields (each 1/60 s). Flashes of light, and therefore swelling, occurred only when channels were present in the vesicular membrane. The flashes were observed when nystatin was added to the cis compartment but not when added to the trans. This demonstrates that the vesicular and planar membranes remain individual bilayers in the region of contact, rather than melding into a single bilayer. Measurements of flash duration in the presence of cobalt (a quencher of calcein fluorescence) were used to determine the side of the planar membrane to which dye was released. In the presence of 20 mM calcium, 50% of the vesicle ruptures were found to result in fusion with the planar membrane. In 100 mM calcium, nearly 70% of the vesicle ruptures resulted in fusion. The methods of this study can be used to increase significantly the efficiency of reconstitution of channels into planar membranes by fusion techniques.     

Photochemically induced charge separation occurring in bacteriorhodopsin. Detection by time-resolved dielectric loss.

Time-resolved dielectric loss (TRDL) measurements are reported for the photochemical excitation of bacteriorhodopsin (bR) in solid films of Halobacterium halobium purple membranes. These measurements provide an independent confirmation for the existence of an important component of charge separation in these membranes after photochemical excitation. The separation of charge is detected by the absorption of microwave energy by the multilayer films of purple membranes in a microwave cavity during flash photolysis experiments. The TRDL method has the advantage of being sensitive to charge separation occurring in both oriented and unoriented films of purple membranes. One disadvantage is that the water content of the samples must be minimized, however, there is some absorbed water present in our electrodeposited solid film samples. To the best of our knowledge, TRDL measurements have not been reported previously for photochemical charge separation in biological membranes. It is significant that an early decay component of TRDL in the 20-microseconds time domain corresponds to the relaxation of the negative charge displacement photocurrent in oriented samples of purple membranes. In addition, a component of charge separation persists during the first several hundred microseconds of the bR photocycle.     

Partitioning of Triton X-100, deoxycholate and C(10)EO(8) into bilayers composed of native and hydrogenated egg yolk sphingomyelin

We have used isothermal titration calorimetry (ITC) to study the thermodynamics of Triton X-100 (T(X-100)), deoxycholate and decyl octaethylene glycol (C(10)EO(8)) penetration into bilayers composed of native (ESM) and hydrogenated egg yolk sphingomyelin (DHSM). Light scattering measurements were used to study the point of saturation (R(e,sat)) and the onset of solubilization of membranes by the detergents. We found that DHSM bilayers at 25 degrees C were much more resistant to detergent partitioning (lower K) and gave higher reaction enthalpies (DeltaH) for all three detergents compared to the ESM bilayer system. Because DHSM lacks double bonds (Delta(4trans) and some cis bonds as well), attractive acyl chain interactions are favored in membranes of this lipid class. The high stability and cohesion of DHSM in membranes could be a crucial functional property of this lipid as it is enriched in eye lens membranes.     

Rotational mobility of an erythrocyte membrane integral protein band 3 in dimyristoylphosphatidylcholine reconstituted vesicles and effect of binding of cytoskeletal peripheral proteins

Band 3 protein was isolated from human erythrocyte membranes, purified, and reconstituted into a well-defined phospholipid bilayer matrix (dimyristoylphosphatidylcholine). The preparation yielded uniform single-bilayered vesicles of the diameter 40--80 nm. The rotational motion of band 3 was studied by saturation transfer electron spin resonance (ESR) spectroscopy of covalently attached maleimide spin-labels. The rotational mobility changed in response to the host lipid phase transition. The rotational correlation time was in a range from 73 (37 degrees C) to 94 microseconds (26 degrees C) in the fluid phase and from 240 (15 degrees C) to 420 microseconds (5 degrees C) in the solid phase. The motion was analyzed based on the anisotropic rotation of band 3 in the reconstituted vesicles. To obtain information on the rotational diffusion constant around the axis parallel to the membrane normal, we made an attempt to measure the angle between the spin-label magnetic axis and the membrane normal. The result gave 3.9 x 10(4) s-1 at 37 degrees C as a rough estimate for the diffusion constant. This is compatible to anisotropic rotation of a cylinder of radius 3.3 nm in a two-dimensional matrix with inner viscosity 2 P and inner thickness 4 nm. The cytoskeletal peripheral proteins caused a definite increase in the rotational correlation time (from 73 to 180 microseconds at 37 degrees C, for example). The restriction of the rotational mobility was shown to be due to the ankyrin-linked interaction between band 3 and spectrin-actin-band 4.1 proteins in the reconstituted membranes.     

Dynamics in a protein-lipid complex: nuclear magnetic resonance measurements on the headgroup of cardiolipin when bound to cytochrome c.

Deuterium and phosphorus nuclear magnetic resonance (NMR) has been used to investigate the dynamics of slow motional processes induced in bilayer cardiolipin upon binding with cytochrome c. 31P NMR line shapes suggest that protein binding induces less restricted, isotropic-like motions in the lipid phosphates within the ms time scale of this measurement. However, these motions impart rapid transverse relaxation to methylene deuterons adjacent to the phosphate in the lipid headgroup and so did not feature strongly in the NMR line shapes recorded from these nuclei by using the quadrupolar echo. Nonetheless, motional characteristics of the headgroup deuterons were accessible to a dynamic NMR approach using the Carr-Purcell-Meiboom-Gill multiple-pulse experiment. Compared to the well-studied case of deuterons in fatty acyl chains of bilayer phosphatidylcholine, the motions determining the 2H spin transverse relaxation in the headgroup of bilayer cardiolipin were much faster, having a lower limit in the 5-10 kHz range. On binding with cytochrome c, the T2 effecting motions in the cardiolipin headgroup became faster still, with rates comparable to the residual quadrupolar coupling frequency of the headgroup deuterons (approximately 25 kHz) and so coincided with the time scale for recording the quadrupolar echo (approximately 40 microseconds). It is concluded that the headgroup of cardiolipin does not exclusively report localized dynamic information but is particularly sensitive to collective motions occurring throughout the bilayer molecules. Although the rates of collective modes of motion may be dependent on the lipid type in pure lipid bilayers, these low-frequency fluctuations appear to occupy a similar dynamic range in a variety of lipid-protein systems, including the natural membranes.     

Diffusion limitation in the block by symmetric tetraalkylammonium ions of anthrax toxin channels in planar phospholipid bilayer membranes

Current flow through the channel formed in planar phospholipid bilayer membranes by the PA65 fragment of anthrax toxin is blocked, in a voltage-dependent manner, by tetraalkylammonium ions (at micromolar concentrations), which bind to a blocking site within the channel lumen. We have presented evidence that diffusion plays a significant role in the kinetics of blocking by tetrabutylammonium ion (Bu4N+) from the cis (toxin-containing) side of the membrane (Blaustein, R. O., E. J. A. Lea, and A. Finkelstein. 1990. J. Gen. Physiol. 96:921-942); in this paper we examine the implications and consequences of diffusion control for binding kinetics. As expected for a diffusion-affected reaction, both the entry rate constant (kcis1) of Bu4N+ from the cis solution to the blocking site and the exit rate constant (kcis-1) of Bu4N+ from the blocking site to the cis solution are reduced if the viscosity of that medium is increased by the addition of dextran. In conformity with both thermodynamics and kinetic arguments, however, the voltage-dependent equilibrium binding constant, Keq (= kcis-1/kcis1), is not altered by the dextran-induced viscosity increase of the cis solution. The entry rate constants (kcis1) for tetrapentylammonium (Pe4N+), tetrahexylammonium (Hx4N+), and tetraheptylammonium (Hp4N+) are also diffusion controlled, and all of them, including that for Bu4N+, attain a voltage-independent plateau value at large positive cis voltages consistent with diffusion limitation. Although the plateau value of kcis1 for Hx4N+ is only a factor of 3 less than that for Bu4N+, the plateau value for Hp4N+ is a factor of 35 less. This precipitous fall in value indicates, from diffusion-limitation theory, that the diameter of the channel entrance facing the cis solution is not much larger than the diameter of Hp4N+, i.e., approximately 12 A.     

Anthrax edema factor, voltage-dependent binding to the protective antigen ion channel and comparison to LF binding

Anthrax toxin complex consists of three different molecules, the binding component protective antigen (PA, 83 kDa), and the enzymatic components lethal factor (LF, 90 kDa) and edema factor (EF, 89 kDa). The 63-kDa N-terminal part of PA, PA(63), forms a heptameric channel that inserts at low pH in endosomal membranes and that is necessary to translocate EF and LF in the cytosol of the target cells. EF is an intracellular active enzyme, which is a calmodulin-dependent adenylate cyclase (89 kDa) that causes a dramatic increase of intracellular cAMP level. Here, the binding of full-length EF on heptameric PA(63) channels was studied in experiments with artificial lipid bilayer membranes. Full-length EF blocks the PA(63) channels in a dose, temperature, voltage, and ionic strength-dependent way with half-saturation constants in the nanomolar concentration range. EF only blocked the PA(63) channels when PA(63) and EF were added to the same side of the membrane, the cis side. Decreasing ionic strength and increasing transmembrane voltage at the cis side of the membranes resulted in a strong decrease of the half-saturation constant for EF binding. This result suggests that ion-ion interactions are involved in EF binding to the PA heptamer. Increasing temperature resulted in increasing half-saturation constants for EF binding to the PA(63) channels. The binding characteristics of EF to the PA(63) channels are compared with those of LF binding. The comparison exhibits similarities but also remarkable differences between the bindings of both toxins to the PA(63) channel.     

Effect of cholesterol on the structure and dynamic properties of unsaturated phospholipid bilayers

Molecular dynamics (MD) computer simulations of five different hydrated unsaturated phosphatidylcholine lipid bilayers built up by 18:0/18:1(n-9)cis PC, 18:0/18:2(n-6)cis PC, 18:0/18:3(n-3)cis PC, 18:0/20:4(n-6)cis PC, and 18:0/22:6(n-3)cis PC molecules with 40 mol% cholesterol, and the same five pure phosphatidylcholine bilayers have been performed at 303 K. The simulation box of a lipid bilayer contained 96 phosphatidylcholines, 64 cholesterols, and 3840 water molecules (48 phosphatidylcholine molecules and 32 cholesterols per layer and 24 water molecules per phospholipid or cholesterol in each case). The lateral self-diffusion coefficients of the lipids in these systems and mass density profiles with respect to the bilayer normal have been analyzed. It has been found that the lateral diffusion coefficients of phosphatidylcholine molecules increase with increasing number of double bonds in one of the lipid chains, both in pure bilayers and in bilayers with cholesterol. It has been found as well that the lateral diffusion coefficient of phosphatidylcholine molecules of a lipid bilayer with 40 mol% cholesterol is smaller than that for the corresponding pure phosphatidylcholine bilayer.     

Membrane models of E. coli containing cyclic moieties in the aliphatic lipid chain

Nearly all molecular dynamics simulations of bacterial membranes simplify the lipid bilayer by composing it of only one or two lipids. Previous attempts of developing a model E. coli membrane have used only 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) and (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol) POPG lipids. However, an important constituent of bacterial membranes are lipids containing a cyclopropane ring within the acyl chain. We have developed a complex membrane that more accurately reflects the diverse population of lipids within E. coli cytoplasmic membranes, including lipids with a cyclic moiety. Differences between the deuterium order profile of cyclic lipids and monounsaturated lipids are observed. Furthermore, the inclusion of the cyclopropane ring decreases the surface density of the bilayer and produces a more rigid membrane as compared to POPE/POPG membranes. Additionally, the diverse acyl chain length creates a thinner bilayer which matches the hydrophobic thickness of E. coli transmembrane proteins better than the POPE/POPG bilayer. We believe that the complex lipid bilayer more accurately describes a bacterial membrane and suggest the use of it in molecular dynamic simulations rather than simple POPE/POPG membranes.     

Stochastic model for electric field-induced membrane pores. Electroporation

Electric impulses (1-20 kV cm-1, 1-5 microseconds) cause transient structural changes in biological membranes and lipid bilayers, leading to apparently reversible pore formation ( electroporation ) with cross-membrane material flow and, if two membranes are in contact, to irreversible membrane fusion ( electrofusion ). The fundamental process operative in electroporation and electrofusion is treated in terms of a periodic lipid block model, a block being a nearest-neighbour pair of lipid molecules in either of two states: (i) the polar head group in the bilayer plane or (ii) facing the centre of a pore (or defect site). The number of blocks in the pore wall is the stochastic variable of the model describing pore size and stability. The Helmholtz free energy function characterizing the transition probabilities of the various pore states contains the surface energies of the pore wall and the planar bilayer and, if an electric field is present, also a dielectric polarization term (dominated by the polarization of the water layer adjacent to the pore wall). Assuming a Poisson process the average number of blocks in a pore wall is given by the solution of a non-linear differential equation. At subcritical electric fields the average pore size is stationary and very small. At supercritical field strengths the pore radius increases and, reaching a critical pore size, the membrane ruptures (dielectric breakdown). If, however, the electric field is switched off, before the critical pore radius is reached, the pore apparently completely reseals to the closed bilayer configuration (reversible electroporation ).     

Proton concentration jumps and generation of transmembrane pH-gradients by photolysis of 4-formyl-6-methoxy-3-nitrophenoxyacetic acid

Proton concentration gradients across membranes are important for many biological energy transducing processes. The kinetics of proton dependent processes can be studied by pH-jump methods in which protons are photochemically released. In the following we describe the synthesis and the properties of photolabile 4-formyl-6-methoxy-3-nitrophenoxyacetic acid, a 'caged proton'. The synthesis is based on vanillin, which is alkylated with chloroacetic acid to give a carboxylic acid (pK = 2.72). In a second step a nitro group ortho to the formyl group is introduced. Photochemical proton release occurs by a reaction mechanism analogous to the well known photochemical formation of 2-nitrosobenzoic acid from 2-nitrobenzaldehyde. The pK values of the photoproduct are 0.75 and 2.76, respectively, thus allowing the use of the compound in a wide pH-range. The quantum yield is 0.18, lower than in the case of the 2-nitrobenzaldehyde/2-nitrosobenzoic acid system (phi = 0.5). The release of the proton in a flash photolysis experiment occurs within less than 1 microseconds. The spectrum of photolabile compound has absorption maxima at 263 nm and 345 nm, respectively. Its permeability across a lipid bilayer membrane is very low (permeability coefficient Pd approximately equal to 10(-9) cm.s-1 at pH 8) so that transmembrane proton concentration gradients can be generated.     

Exchange diffusion of dopamine induced in planar lipid bilayer membranes by the ionophore X537A

The ionophore X537A causes a large increase in the [(14)C]dopamine (a catecholamine) permeability of planar bilayer membranes. Dopamine transport increases linearly with the ionophore concentration. At relatively high concentrations in the presence of dopamine, the ionophore omdices a conductance which is nearly ideally selective for the dopamine cation. However, the total dopamine flux as determined in tracer experiments is not affected by an electric field and is over 10(5) times larger than predicted from the estimated dopamine conductance. Increasing the dopamine concentration on the side containing radioactive dopamine (the cis side) saturates the dopamine transport. This saturation is relieved by trans addition of nonradioactive dopamine, tyramine, H(+), or K(+). With unequal concentrations of dopamine cis and trans (49 and 12.5 mM), the unidirectional dopamine fluxes are equal. Increasing H(+) cis and trans decreases dopamine transport. It is concluded that at physiological pH, the X537A-induced transport of dopamine occurs via an electrically silent exchange diffusion of dopamine cation with another cation (e.g., dopamine(+), H(+), or K(+)). X537A induces a Ca(++)-independent release of catecholamines from sympathetic nerves by interfering with intracellular storage within storage vesicles (R.W. Holz. 1975. Biochim. Biophys. Acta. 375:138-152). It is suggested that X537A causes an exchange of intravesicular catecholamine with a cytoplasmic cation (perhaps K(+) or H(+)) across the storage vesicle membrane.     

The transverse distribution of phospholipids in the membranes of Golgi subfractions of rat hepatocytes.

The transverse distribution of phospholipids in the membranes of subfractions of the Golgi complex was investigated by using phospholipase C and 2,4,6-trinitrobenzenesulphonic acid as probes. In trans-enriched Golgi membranes, 26% of the phosphatidylethanolamine is available for reaction with trinitrobenzenesulphonate or for hydrolysis by phospholipase C, and 72% of the phosphatidylcholine is hydrolysed by phospholipase C. In cis-enriched Golgi membranes, 45% of the phosphatidylethanolamine is available for reaction with trinitrobenzenesulphonate and for hydrolysis by phospholipase C, and 95% of the phosphatidylcholine is hydrolysed by phospholipase C. Under the conditions used with either probe the contents of the Golgi vesicles labelled with either [3H]palmitic acid or [14C]leucine were retained. Galactosyltransferase activity of the membrane vesicles was partially inhibited by the experimental procedures used to investigate the transverse distribution of phospholipids. However, the residual activity was latent, suggesting that the vesicles remained closed. Trinitrobenzenesulphonic acid caused no detectable morphological change in either Golgi fraction. Phospholipase C treatment caused morphological changes, including fusion of vesicles and the appearance of 'signet-ring' profiles in some vesicles; however, the vesicles remained closed and the bilayer was retained. It appears, therefore, that neither probe causes major disruption of the Golgi vesicles nor gains access to the inner surface of the membrane bilayer. These observations suggest that phospholipids have a transverse asymmetry in Golgi membranes, that this distribution differs in trans and cis membranes, and that the phospholipid structure of Golgi membranes is inconsistent with a simple flow of membrane bilayer from endoplasmic reticulum to Golgi membranes to plasma membrane.