GENERAL DISCUSSION Prof. J. F. Nagle (Carnegie-Mellon University, Pittsburgh, PA) addressed Dr Gersh- feld: These are interesting experimental results. Your characterization of T" as being a critical temperature seems a most plausible suggestion, but one that I find difficult to reconcile with your characterization of the state at T" as being a surface' bilayer. As you have anticipated in your publications, a surface bilayer seems so counterintuitive that perhaps other interpretations should be entertained. I would like tentatively to suggest a different model. In this model for temperatures slightly lower than T" there is a monolayer with occasional fluctuating infolding to form bilayers underneath part of the monolayer. As T" is approached the infolding becomes critically unstable, so that there is one (or more) bilayer beneath all the monolayer. Above T* the accumulated subsurface bilayers peel of€ the monolayer owing to their own weight.This infolding model predicts a surface coverage at T* greater than the equivalent of three monolayers ( i e . one monolayer and one bilayer) over a very narrow temperature range; owing to the difficulty of precise temperature control this higher surface coverage could easily have been missed in your surface coverage measurements that reported a maximum surface coverage near two. Surface potential measurements might be a better way to distinguish between the two interpretations. A single symmetric bilayer would have no net potential across it, whereas a monolayer, or a monolayer with one or more bilayers beneath it, would have typical monolayer potentials of ca.500mV. Therefore, as a function of temperature the surface potential would change by ca. 500 mV according to your surface bilayer interpretation. In contrast, the surface potential would change much less according to the infolding monolayer model sketched above. Dr N. L. Gershfeld (National Institutes of Health, Bethesda, MD) replied: The model you propose has some appealing features because it may provide a reasonable model for films with densities between one and two monolayers at temperatures above and below T". However, at T* the model is inconsistent with a number of our observations. ( a ) Within the experimental error of the radioisotope measurement (*2%) the film density does not fluctuate after equilibrium has been attained, but remains at the bilayer density.Even after stirring the dispersion vigorously, the bilayer density is recovered after equilibrium is re-established.' ( b ) Resistance to the evaporation of water occurs only when the dispersion is at T" (k0.1 K); the monolayer gives little or no resistance to evaporation.2 In your model the properties of the monolayer would dominate over the entire temperature range, and we would not have observed any evaporation resistance at T*. We view the surface bilayer to be a continuous bilayer structure with water on both surfaces. Since the film density falls fairly steeply to values between one and two monolayers at temperatures above and below T", the surface potential may very well approach the monolayer value, particularly if the temperature is not carefully controlled at T".We have estimated that the temperature should be controlled to at least kO.001 K to observe pure bilayer properties at T".2 1 K. Tajima and N. L. Gershfeld, Biophys. J., 1985, 47, 203. 2 L. Ginsberg and N. L. Gershfeld, Biophys. J., 1985, 47, 211. Dr D. Marsh (Max-Planck-Institut Gottingen, West Germany) turned to Dr Gersh- feld. I would like to ask you to comment on the extremely high aqueous solubilities which you observe for DMPG. At the approximate position of the Krafft point in fig. 3 of your paper, the critical micelle concentration (c.m.c.) is ca. mol dme3. Interpola- tion of the logarithmic chain-length dependence of the c.m.c.s measured for diacyl 6364 0 General Discussion I I I I , i+-J m phosphatidylcholines' suggests that the c.m.c.for dimyristoyl phosphatidylcholine is ca. mol dm-'. Dr King and I have measured the c.m.c.s of a short-chain, spin-labelled phosphatidylcholine and phosphatidylglycerol using electron spin resonance spectros- copy. The salt dependence of the c.m.c.s at pH 7 is given in fig. 1. In the absence of added salt, (at a counter ion concentration of 10-4moldm-3) the c.m.c. of phos- phatidylglycerol is ca. 10 times greater than that of phosphatidylcholine. Hence I would expect the c.m.c. of DMPG to be of the order of ca. lo-* mol dm-3. 1. R. J. M. Tausk, J. Karmiggelt, C. Oudshoorn and J. Th. G. Overbeek Biophys. Chem., 1974,1, 175; 184; R. Smith and C. Tanford, J. Mol. Biol., 1972, 67, 75. Dr N .L. Gershfeld (National Institutes of Health, Bethesda, MD) said: Dr Marsh's estimate of the c.m.c. for DMPG is based on the assumption that the value of 10 for the ratio of c.m.c.s that he measured for the spin-labelled compounds was applicable to the dimyristoyl analogues. Solubility vs. temperature data for DMPC suggest that it is not (unpublished results). The solubility data indicate that the Krafft temperatures, where the solubility of the lipid increases markedly with temperature, are approximately the same for DMPG and DMPC; however, the solubility, and the c.m.c., for DMPC at this temperature is ca. 10-7moldm-3. Thus, a factor of ca. 1000 rather than 10 might be anticipated for the c.m.c. ratios of the dimyristoyl analogues. Moreover, the extrapolation used to obtain the c.m.c.of DMPC may not be valid. Indeed, the extrapolation underestimates the c.m.c. of DPPC obtained by Smith and Tanford by at least an order of magnitude. Given these two corrections to Dr Marsh's estimate of the c.m.c. of DMPG, our value for the solubility of DMPG appears to be reasonable.General Discussior 65 Dr D. P. Siege1 (Proctor and Gamble, Cincinnatti, OH) posed two questions: ( a ) I wonder if the existence of the ‘jelly’ phase may be due in part to the strong electrostatic repulsions between bilayers in these nearly salt-free systems. Such interac- tions could make the dispersion quite viscous because either the Debye length may be significant compared to the average inter-bilayer separation in the dispersion, or because the electrostatic repulsion is so strong (i.e.the surface electrostatic potential in the system is so large) that the repulsive forces between bilayer aggregates is significant even when they are many Debye lengths apart. If this were true, the effective viscosity of the system would increase with increasing lipid concentration, as you observed, because of the corresponding decrease in average inter-bilayer separation in the disper- sion. Have you observed this viscous phase in the presence of added electrolytes? ( b ) One of the most interesting aspects of your work is the implication that the spontaneous production of unilamellar vesicles at temperatures around T* is a higher- order phase transition in the La phase. Higher-order phase transitions are accompanied by discontinuities in the corresponding derivatives of the free energy with respect to temperature and pressure ( e.g.discontinuities in the heat capacity and isothermal compressibility accompany second-order phase transitions). Is there any experimental evidence for discontinuities in these derivatives in the L, phase at T*? Dr N. L. Gershfeld (National Institutes of Health, Bethesda, MD) replied: ( a ) Preliminary studies indicate that the ‘jelly’ phase persists in mol dm-3 NaCl solutions, although the viscosity of this phase appears to decrease at this ionic strength. At still higher NaCl concentrations, e.g. 0.1 mol d ~ l l - ~ , the ‘jelly’ phase does not seem to form. ( b ) Heat-capacity measurements in the region of T* have not been reported, but we have begun to examine this temperature interval very carefully with a high-precision adiabatic calorimeter.However, we anticipate that the effects are likely to be very small because the transformation is between two liquid-crystalline bilayer states, the jelly and the unilamellar vesicle. Dr M. N. Jones (University of Manchester) remarked: We have found that sonicated vesicles of DPPC or DPPC+PI are disrupted by glass surfaces in the form of either solid or porous glass beads. This disruption catalysed by a glass surface results in the formation of a more turbid lipid suspension, the precise structure of which has yet to be established. I note that you suggest that glass beads facilitate the formation of vesicles from crystals of DMPG. Would you care to suggest a mechanism for the process and comment on the size of the vesicles formed in the presence of the glass beads? Dr N.L. Gershfeld (National Institutes of Health, Bethesda, MD) replied: In the absence of glass beads the diameter of the vesicles formed at T* is of the order of 0.5 pm, while in the presence of beads the vesicles are ca. 5-50 pm. The former estimate is based on preliminary light-scattering measurements. Given the paucity of data regarding this effect I hesitate to suggest any mechanisms for the process, except to note that the glass beads appear to increase the size of the vesicles in both of our systems. Prof. D. G. Hall (Unilever Research Port Sunlight Laboratory) said: My comments are concerned with fig. 1 of the paper: me vs T. (1) I would expect a change in slope at T,, but there are too few data points for this to be apparent, do you agree? (2) The break at T* appears to be quite sharp.A break of this kind suggests the occurrence of a first-order phase transition. Indeed it is hard to see how a higher-order transition can be responsible. Since there is no evidence of a first-order transition of T* in the bulk phases could the effect be due to a first-order transition in the interface? If so should there not be a sharp change in the surface excess r? Has such a change been observed? If the transition is blurred the methods developed in ref. ( 1 ) may be useful. 1. D. G. Hall, J. Chem. SOC., Furuduy Trans. 2, 1972, 68, 668.66 General Discussion Dr N. L. Gershfeld (National Institutes of Health, Bethesda, MD) replied: The major focus of this paper was to establish that at T", where the surface pressure is a maximum, the phospholipid dispersion forms a suspension of unilamellar vesicles.We were primarily concerned with identifying T* for DMPG, and therefore obtained only the minimum number of data sufficient to establish this temperature. The points you have raised about the relationship between the (rIe, T ) data of fig. 1 and lipid transitions have been addressed in detail with other phospholipids. Thus, for dispersions of dimyristoylphosphatidylcholine (DMPC), whose values of T", T, and AH, are similar to those for DMPG, we have observed a change in slope at T,, and were able to evaluate the latent heat of the transition from the slopes at this temperature [ref.(7) of the paper]. The Il, vs. T curve for DMPC shows a sharp break at T* similar to the one exhibited in our fig. 1. However, independent measurements of the surface concentration using radiotracers shows that the film density is a continuous function of the temperature with a maximum at T* [ref. (7) and (8) of the paper]. Since a first-order transition requires that the surface concentration be discontinuous at the transition temperature, the transition at T* is assumed to be of higher order. For DMPG, T* coincides with the temperature where the 'jelly' is transformed to the unilamellar vesicles (fig. 4 of the paper). This bulk phase transformation occurs over several degrees, and we have therefore concluded that it is a higher-order transition. We assume that an equivalent phenomenon occurs in the surface film at T*.Mr F. A. M. Leermakers, Dr J. M. H. M. Scheutjens and Prof. J. Lyklema (Agricultural University of Wageningen, The Netherlands) said: Data on the free lipid concentrations in equilibrium with a membrane are very useful for thermodynamical analyses. Our statistical-thermodynamical theory predicts the equilibrium volume fraction of the lipids for given head-solvent, solvent-tail and tail- head interaction parameters.' It appears that these concentrations are very close to the critical micelle concentration (c.m.c.). The interaction parameters used are enthalpies per kT, Le. inversely proportional to the absolute temperature. Hence, from our theory the temperature dependence of the critical volume fraction can also be obtained. Fig.2 shows the result for a series of non-ionic lecithin-like molecules. The curves look very similar to the experimental data given in fig. 3 of the paper by Gershfeld et al. which are replotted in our fig. 2. Only at high temperatures do significant deviations between theory and experiment occur. As sug- gested in the paper, these deviations might be due to the formation of micelles which can not be separated by centrifugation. Fig. 2 also shows that the lines for different tail lengths are parallel with respect to each other, both above and below T,. The vertical distance between the curves is ca. 1.2, i.e. 0.3 per added tail segment, which correlates well with known c.m.c. data for small ionic lipids.* Furthermore, the molecular-weight dependence of the discontinuities agrees with experimental T, data.Does Dr Gershfeld have experimental indications that for different tail lengths the lines are indeed parallel to each other and shifted as predicted above? 1 F. A. M. Leermakers, J. M. H. M. Scheutjens and J. Lyklema, Biophys. Chem., 1983, 18, 353. 2 Solution Behavior ofSurfactants, ed. K. L. Mittal (Plenum Press, New York, 1982), vol. 1 and 2. Dr N. L. Gershfeld (National Institutes of Health, Bethesda, MD) replied: The only T us. solubility data presently completed are for DMPG. However, the slope of the solubility curve depends on the latent heat of the gel-liquid-crystal transition. Since this heat varies with chain length, we would not expect the solubility vs. temperature curves for the homologues of DMPG to be parallel. Dr G.Cevc (Essen University, West Germany) said: Dr Parsegian told us that the addition of tetradecane to phosphatidylethanolamine in the inverted hexagonal phaseGeneral Discussion 67 10 30 50 T / "C Fig. 2. Dependence of the volume fraction 4* in equilibrium with a membrane on temperature. The theoretical curves apply to membranes composed of nonionic lecithin-like molecules with variable tail length x. The experimental curve is replotted from fig. 3 of the paper by Gershfeld et al. The interaction parameters at 300 K are: xAB = 1.2, xAS = 1.6, xBS = 0 for tail-head, tail- solvent and head-solvent, respectively. The trans-gauche energy is (275/300)kT. does not change the radius of the water cylinders appreciably, whereas partitioning of this alkane into the phosphatidylcholine hydrocarbon matrix increases the swelling of the latter lipid in such phase substantially.However, the samples of phosphatidy- lethanolamine that he was comparing contained 5 YO and those of phosphatidylcholine 20% of the alkane. Would the conclusion be different if the tetradecane concentration was in both cases the same? Secondly I would like to ask where Dr Parsegian envisages that the tetradecane molecules sit in the lipid hydrocarbon region? Does he think that different localization of the alkane molecules in the lipid hydrocarbon cores might provide a partial explanation for the large differences in the behaviour of phosphatidylethanolamine and phos- phatidylcholine upon the tetradecane-induced perturbation? Dr V.A. Parsegian (National Institutes of Health, Bethesda, MD) replied: (1) We have in fact looked over a range of alkane concentrations. The two concentrations used in the text were above saturating values in their respective systems. It may be that extra alkane is simply not incorporated into a regular lattice but sits aside in a separate pool invisible to low-angle X-ray diffraction. We expect further therefore no difference in our conclusions if both preparations be at 20% tetradecane. (2) We suggest favoured accumulation of deuterated alkane in the directions of d,,, . Alkanes can go elsewhere, but not to the same concentration. Certainly there can be differences in alkane distribu- tion in PC and PE systems, since each of these species is expected to have different spontaneous curvatures.Prof. D. A. Haydon (Cambridge University) said: I note that n-tetradecane was used to test for the existence of stress in the alkyl chain region of the H phase. In lipid68 General Discussion bilayers the penetration of n-alkanes into the hydrocarbon chain region declines as the size of the n-alkane increases’ and n-tetradecane lies in the chain-length range in which this effect is becoming important. Thus n-tetradecane is not the best choice of hydrocar- bon. Have the authors investigated the effects of shorter n-alkanes-on the H phase? 1 R Fettiplace, D. M. Andrews and D. A. Haydon, J. Membr. Biof., 1971, 5, 277. Dr V. A. Parsegian (National Institutes of Health, Bethesda, MD) answered Prof. Haydon as follows: This is a good question! Ideally one would like to use an alkane small enough to penetrate all hydrocarbon regions, but small substances are too volatile to be practical.Dodecane acts much as tetradecane on the phase structure, but we have used no alkanes smaller than dodecane. It might help to remember that we are looking at alkanes in HII lattices not in bilayers. Difficulties of tetradecane entry need not be the same here. The neutron- diffraction observation that alkane is indeed entering the lattice together with the null result of penetrant alkane on the osmotic stress as a function of lattice spacing should suffice to support our present conclusion. Having said this, it would not surprise us if slightly different behaviour were indeed seen with different medium-length alkanes ( n = 10-16, say).For very small alkanes we expect the entropy of distribution to be relatively important; for longer-chain alkanes packing constraints on their incorporation should be more important. Dr D. Marsh ( Max-Planck-Institut Gottingen, West Germany) addressed Dr Parsegian. First I have a comment regarding acyl chain stress in the HII phase. We have measured the dimensional changes at the L,-HII phase transition for two saturated phosphatidylethanolamines of differing chain-length.’ A geometrical calculation of the water dependence of the hexagonal chain dimension, d,,, , for diarachinoyl phosphatidylethanolamine, based on these dimensional measurements, is given in fig. 3. At the water content corresponding to the limiting hydration of the HII phase at a temperature just above the L,-HII transition, the value of d,,, is equal to the lipid thickness in the fluid lamellar phase immediately below the transition.This also holds approximately true for the hexagonal phase transition in didodecyl phosphatidy- lethanolamine. Thus the L,-HII transition takes place without any increase in hydrocar- bon chain extension. This result seems to be in full agreement with your experiments on the effects of added alkane and may constitute a guiding principle for hydrocarbon chain packing in the HI, phase. My question relates to your interpretation of the relative hydration properties of PE and PC at very low water activity. From your fig. 3 (and cf. also the accompanying fig. 3) it is clear that the simple ‘water cylinder’ structure of the HI, phase cannot hold at these very low water contents. What is known about the detailed molecular structure of the HII phase under these conditions, and is it possible that the osmotic stress properties are dominated by the structural rearrangement which must take place in the water/ headgroup/chains at these low water compositions, and not by the headgroup hydration properties per se ? 1 J.M. Seddon, G. Cevc, R. D. Kaye and D. Marsh, Biochemistry, 1984, 23, 2634. Dr V. A. Parsegian (National Institutes of Health, Bethesda, MD) replied: Yes, a water cylinder picture is certainly too idealized at low water contents. That is why we referred to the polar aqueous region there as a ‘mash’. Our suspicion is that the hydration in this regime depends only weakly on the constraint that the polar groups are attached to acyl chains.We hope to test this suspicion by measuring the hydration of polar group preparations. We will then see whether there are large PE/ PC differences.General Discussion - 0 69 I 1 1 1 I I lipids per perimeter water content, (1 - c) Fig. 3. Variation of the maximum length of a lipid molecule, d,,,, as a function of fractional water content, (1 - c) w/w in the HI* pbase, calculated using purely geometrical considerations and the values of area per molecule = 49 A2 and partial specific volume 6, = 1.043 cm3 g-' measured for diarachinoyl phosphatidylethanolamine in the HI, phase at maximum hydration.' The horizon- tal dashed line corresponds to the lipid thickness in the fluid lamellar La phase immediately below the L,-HII transition and the full line to that in the crystalline lamellar L, phase.' Dr J.M. Seddon (University of Southampton) asked Dr. Parsegian. (1) What values of area per molecule (at the water/lipid interface) did you measure in the HII phase, and are these values lower than those measured in La? (2) Do you consider it likely that a tighter headgroup packing in the HII phase of PE would permit more extensive and/or stronger hydrogen bonding between the headgroups than in the L, phase? (3) Could this mean that the apparent similarity between the pressure curves for DOPE- DOPC-tetradecane and DOPE at low hydration is actually spurious, in the former case the principal contribution to the work coming from the dehydration of the relatively polar headgroups, whereas in the latter case the work arising partly from dehydrction, and partly from a breaking of hydrogen bonds upon the structural rearrangement of the PE headgroups that must occur below some minimum radius of the water cylinders? Dr V.A. Parsegian (National Institutes of Health, Bethesda, MD) replied to p r Seddon's three questionsoas follows: ( 1 ) The HII areas are as folloys: DOPE, 53.5 A3 (full hydration) to 23.9 A2; DOPES;;S% tetradecane, 55.3 to 24.3 A2; DOPE: DOPC 3 : 1 + 20% tetradecane, 71.3 to 30.9 A2. Only the last can be easily compar$d with La, where for DOPE: DOPC 3: 1 (no tetradecane) areas ran from 66 to 61.4 A. The HII values are lower than for L, only after some dehydration. (2) Smaller areas should certainly allow tighter head-group packing.The continuous change in area will make it impossible to maintain a particular head-group packing over much of a hydration70 General Discussion range. Specific arrangements of hydrogen bonding are beyond us now. (3) Perhaps the similarity is spurious, but we think not. A similar convergence is seen with egg PC and egg PE lamellae at high osmotic stress [ref. (32) of our paper]. Dr G. J. T. Tiddy (Unilever Research Port Sunlight Laboratory) asked Dr Parsegian. How critical is the position of the lipidlwater boundary in the calculation of ro and r? What is the change in K , if the headgroups are included in the aqueous region? Dr V. A. Parsegian (National Institutes of Health, Bethesda, MD) replied: Not To see this, recall that we originally got K , from critical; there is an increase in K , to bring it even closer to the planar bilayer values.-_- rl ro If we use another measure of radius p = r(1 + q) and use the same definition P m P1) lim - PI+PO 1 1 -_- P1 Po where and we have --- rl ro 1.e. the estimate of K , increases by a factor (1 + q)3. molecule, we obtain (1 + q ) from Now if we add a volume v, per polar group to the volume of water v, per phospholipid 2, + v , v p 2 P--- - v r 2 - (1 + T I 2 VW and a correction factor to K , , [ 1 + ( u p / v,,,)]~’~. From our measured density of DOPE polar groups we have 215.2 0, = x 0.66 = 236 A3. 0.602 x loz4 From the data for DOPE HI* phase at zero stress v, = 600 A3, to give a maximal correction of 1.64 or 64%. So 4.6kT becomes 7.6kT, in better accord with the (7-17)kT expected from planar membranes [ref.(26) of our paper]. Prof. B. de Kruijff (Utrecht, The Netherlands) said: I wish to make one comment and pose one question to Dr Parsegian.General Discussion 71 (1) First, I would like to comment that I, with you, argue that it is essential to obtain detailed insight into the phase behaviour of lipids under equilibrium conditions. However, since biological membranes are not at equilibrium, insight into (intermediate) non-equilibrium lipid structures in model membranes is likely to provide a greater understanding of the possible biological significance of lipid polymorphism. (2) Concerning your studies on the effect of alkanes on phase structure of hydrated PE and PC, I would like to ask whether you have any experimental data on the localisation of these molecules in either the bilayer or the hexagonal Hrr phase.This in view of a possible discrimination between an acyl chain disordering and ‘space-filling’ type of mechanism for HII phase formation by these molecules. Dr V. A. Parsegian (National Institutes of Health, Bethesda, MD) replied: (1) Of course, as long as you really know how to transfer to biological situations non-equilibrium information gathered in vitro. (2) Your excellent question is related to that put by Prof. Haydon. As analysed so far, the neutron diffraction studies mentioned in our paper show only that alkanes are entering the lattice in a way that corresponds to HII symmetry. We are now analysing the data to assign a more specific alkane location.We do tend to think of the alkane as filling a space where, lattice dimensions suggest, the phospholipid acyl chains cannot reach without a large expenditure of energy. (However, the idea of a volume that acyl chains cannot fill is an idealization that breaks down on a molecular scale.) There can indeed also be some disordering of acyl chains by diffusely distributed alkane. To go further into the matter, we are investigating a wider range of alkane sizes. Prof. G. Lindblom (University of Ume6, Sweden) said: I have a comment on Dr Parsegian’s paper. We have investigated the ternary phase diagram (for the temperature range 25-50 “C) for the system dioleyl-lecithin-dodecane-water.’ This system exhibits several liquid-crystalline phases, e.g.lamellar, cubic and hexagonal. The most interesting feature of the phase diagram is that there is a transition from the lamellar phase to a reversed hexagonal phase with increasing water content. the phase transition can be conveniently followed by 31P n.m.r. as shown in fig. 4. These results are in good agreement with Dr. Parsegian’s theory for the formation of HII phases in excess water. These findings also have important implications for some of the functions of the biological membrane. Thus for example we have found2 that the bacterium Acholeplasma laidlawii changes the membrane lipid composition in a remarkable way upon addition of dodecane to the growth medium. This is necessary in order to keep the bilayer membrane intact. 1 M. Sjolund, G.Lindblom, L. Rilfors and G. Arvidson, Biophysics, submitted for publication. 2 A. Wieslander, L. Rilfors and G. Lindblom, Biochemistry, in press. Dr V. A. Parsegian (National Institutes of Health, Bethesda, M D ) said to Prof. Fromherz. How can you be sure of a well defined chemical potential after the shattering disturbance of sonication? The question is especially worrying in view of the polydisper- sity shown in your plate 1. Your ‘discs’ presumably have different perimeter-to-area ratios with necessarily different resulting mid-disc cholate concentrations and ‘chemical potentials’ for each disc. Prof. P. Fromherz ( University of Ulm, West Germany) replied: We have to distinguish two processes of quite different time constant. Sonication creates fragments of finite edge tension.This is a non-equilibrium situation with respect to the state of aggregation of the bilayer. In a relatively slow process the system relaxes to the metastable state of closed vesicles. (In a further slow process the system would relax by fusion to extended bilayers, which would be the equilibrium situation again.) We assume that during the slow relaxation the distribution of cholate between edge, interior of bilayer and bulk water is in equilibrium, because the exchange of surfactants is a fast relaxing reaction.72 General Discussion Fig. 4. 101.3 MHz 31P n.m.r. spectra recorded at 25 "C from DOPC-'H,O-n-dodecane mixtures with a DOPC-n-dodecane molar ratio of 1 : 2 and with (a) 14% (w/w), (b) 44% and ( c ) 54% 2H20. Then in the non-equilibrium situations of the dispersion during relaxation from discs to vesicles the chemical potential of cholate is a well defined quantity as determined by the concentration of cholate in bulk water.Owing to the destruction of edge during the closure of discs to vesicles the chemical potential may change slightly during the slow relaxation of the dispersion, implying a slow shift of the distribution equilibrium. Dr L. Fisher ( CSIRO, Sydney, Australia) said: The system egg-lecithin-taurocheno- desoxycholate is known to form large rod-shaped micelles. Could the objects shown in your plate 2 be related to such micellar structures? Prof. P. Fromherz (University of Ulm, West Germany) answered: Considering the periodicity of the pattern in plate 1 we interpret the pattern as stacks of bilayer discs.The stacks are formed by the stain. Before staining these large aggregates do not exist, as indicated by dynamic light scattering. At high concentration of cholate stable rod-shaped micelles may exist. I do not know whether they appear as similar patterns in the electron microscope after negative stain. Prof. J. K. Thomas (University of Notre Dame, ID) said: Sometime ago' we investi- gated the effect of bile-acid surfactants on lecithin vesicles. We found that the bile acid tends to form domains in the vesicle and eventually leads to vesicle rupture and to the formation of mixed micelles. Would these effects in any way affect your model of edge-bile acid stabilization? 1 J. K. Thomas, D. A. N. Moms, F. Castellino and R. McNeil, Biochim.Biophys. Acta, 1980, 599, 380.General Discussion 7 3 Prof. P. Fromherz (University of Ulm, West Germany) replied: Titration of vesicles by cholate leads to a drop of the edge tension, even if no actual edges are present. The edge tension governs the probability of the formation of transient pores. If the edge tension is further lowered by addition of edge actant a pore may grow above a critical size such that rupture occurs. This process of rupture appears before the edge tension vanishes, depending on the elastic modulus and the osmotic pressure. Systematic experiments on the formation of transient pores and on burst in relation to the concentra- tion of emulsification (vanishing edge tension) and to the c.m.c. of various edge actants are in progress.Dr R. Schubert (Chimrgische Klinik, Tubingen, West Germany) said: On the basis of our investigations on detergent-lipid interactions, a few remarks should be made concerning vesicle formation via dialysis and the effect of detergent binding on vesicular membranes. If cholesterol or sphingomyelin is partially substituted for egg-yolk lecithin or if temperature is decreased but still above T, , dialysis of lipid-cholate mixed micelles results in larger vesicles. This effect may be due to changes (i.e. reduction) in the edge tension of the mixed disc micelles. On the one hand, however, we found an increase in vesicle size with increasing dialysis temperature, when C,EO, was used as detergent. Whether the concept of edge activity is also valid for such non-ionic detergents, therefore, remains to be clarified.On the other, the kinetic control of detergent dialysis by reduced cutoff of the dialysis membrane as well as a higher lipid concentration result in larger vesicles, both of which are independent of changes in the edge binding constant. These observations tend to indicate that, before disc-vesicle transition, the size of the largest discs that finally close to vesicles is determined by the frequency of disc fusion. Moreover, the concept of shape transformation from a flat bilayer to a closed vesicle does not adequately explain the higher amount of lipid in the outer vesicle leaflet, postulated on the basis of sterical considerations. Our studies on cholate-membrane interaction clearly demonstrate that the size of the fluctuating vesicle pores induced by membrane-bound detergent can only be small.Large hydrophilic molecules are only released from vesicles shortly after cholate addition by transient pores arising during a sudden membrane foldover.' The rapid closure of these larger pores suggests that the inner edges in membrane pores formed by bile salts are far less stable than the other edges in mixed disc micelles. 1 R. Schubert, K. Beyer, H. Wolburg and K-H. Schmidt, Biochemistry, 1986, 25, 5263. Prof. P. Fromherz (University of Ulm, West Germany) commented in reply: ( a ) The sensitivity to temperature of the structure of dispersions of detergents with headgroups of oligo-oxyethylene is well known. Thus a modulation of edge-activity by temperature is possible. ( b ) During dialysis two processes occur.( 1 ) A reduction of the cholate concentration with concomitant enhancement of edge tension. (2) A growth of the discs with finite edge tension by lateral fusion. In the case of fast dialysis the enhancement of edge-tension is fast as compared to growth. That is to say the discs become unstable at a relatively small size owing to the high edge tension. In the case of slow dialysis the discs have time to grow at a relatively low edge tension. There the instability occurs at a relatively large size. The model of edge activity leads thus to the first rationalization of the correlation slow dialysis/large vesicles and fast dialysis/ small vesicles. The relation has been drawn explicitly in the size us. concentration diagram in our first paper on the subject [fig.4 of our ref. (9)]. ( c ) The theory of disc-vesicle transition developed in our paper and in our ref. (9) treats the bilayer as a two-dimensional liquid with bending elasticity. A phenomenologi- cal model describing the energetics of bending by an elastic modulus cannot explain,74 General Discussion of course, the molecular mechanism of the process. In molecular terms the redistribution of lipid from the inner to the outer monolayer is a natural mechanism to lower the elastic energy. At the open edge this redistribution is a fast process. ( d ) Concerning the formation of pores in closed vesicles, the concept of edge activity apparently describes the instability of edges in closed vesicles with a finite edge tension as compared to the edges of mixed micelles with vanishing edge tension. Dr D.S. Dimitrov (Bulgarian Academy of Sciences, Sofia, Bulgaria) said: There are at least three important findings in Prof. Fromherz's work: (1) demonstration of open discs as intermediates of vesicle formation, (2) estimates of the edge energy and the curvature elasticity and (3) effects of cholate on the rate of vesicle formation. I would like to add that membrane viscosity can contribute to the kinetics of vesicle formation, especially in the cases where there are no activation barriers. This may be the case for formation of giant vesicles, where the curvature elasticity effects can be neglected. We have made a simple estimate for the rate of liposome formation assuming that the driving force is the edge tension and that there are no activation barriers.The forces which resist the edge tension are due to the membrane viscosity and curvature elasticity. Then an approximate balance of forces leads to the following expression for the time of liposome formation time = viscosity x radius/ (edge tension - curvature elasticity/ radius). Using values for the two-dimensional membrane viscosity, curvature elasticity, edge tension and liposome radius on the order of erg, lo-* dyn and 10 pm, respectively, we get times on the order of 100 s. It is seen that the curvature elasticity effect can be neglected for giant liposomes. For small vesicles it is important and determines a minimum radius of the order of curvature elasticity/line tension = cm. Unfortunately, there is no theory for the size distribution.It should be pointed out that for small vesicles the edge energy per disc, which eventually will transform to a liposome, is of the order of thermal energy kT. For giant liposomes, however, if we assume that the mechanism of liposome formation is that of disc-vesicle transformation and the disc edge has the same edge energy as for small vesicles, then the total edge energy per liposome is much larger than the thermal energy. In this case the size distribution should be very narrow which is apparently not the case. Therefore, the effective edge tension is probably very small. This leads to long times of liposome formation. The above formula indicates that the kinetics of liposome formation should depend on the mem- brane viscosity. This dependence may be responsible for the increased rate of liposome formation at high temperatures.In addition, liposomes do not form from lipids in solid state because their viscosity is very high. s.P., Prof. J. F. Holzwarth (Fritz-Haber Institut, Berlin, West Germany) said to Prof. Fromherz: In your paper you discuss the influence of amphiphiles on the stabilisation of bilayer systems with 'open' edges (i.e. the hydrophobic part in contact with water). You also mention briefly the paper by Batzri and Korn [ref. (31)] which gives the first description of the so-called injection method for preparing vesicles. In this method, which was further refined by Kremer et al.' as well as Eck and Holzwarth? an alcoholic solution of lipids is carefully injected into a tenfold volume of buffer solution over a period of several minutes.During the injection process the buffer solution is kept 5-10 "C above the phase transition temperature (T,) of the lipid and carefully stirred. Afterwards the mixture is dialysed against buffer at a temperature of 5-10 "C above T' for at least 8 h to remove the alcohol from the bilayer (vesicle) preparation. In this way very stable preparations of small unilamellar vesicles (SUV) are formed, the size of which varies between 20 and 100nm depending on the final concentration of lipids (typically 1-10 mmol drnp3). If one measures the turbidity of such preparations in thePlate 1. See caption to fig. 6. (To facep. 75)General Discussion I/] h Y d .- * 2 - v 0 c -E 0, 2 1 - 75 ----- - - ---- - 2 - -. -__ ' --- -: -.r\ ' '\ d \ \ I 1 1 I > l o o 90 F 80 70 4 60 D 5 50 40 30 20 10 10 20 30 40 50 60 70 80 90 100 sizelnm Fig. 6. Size distribution of unilamellar vesicles from 2.5 mmol dm-3 dipalmitoylphosphatidyl- choline (DPPC) lipids derived from negatively stained preparations in an electron microscope, shown in plate 1. Total number of vesicles = 121 1; maximum number = 80.3; size at maximum = 23.3 nm; direct width parameter = 1.33. temperature range of the phase transition one gets the results given in fig. 5 . My question is: Can you explain the hysteresis which is found in the presence of 10% ethanol in contrast to the dialysed vesicles by a stabilization of aggregates with edges caused by the alcohol? If such aggregates are dialysed they might lose the alcohol from the edges and stabilize themselves by forming unilamellar vesicles.The second question is: Can the size distribution, an example of which is given in fig. 6 and plate 1, be explained76 General Discussion by the amount of alcohol which is available to stabilize intermediates with edges? An indication of this is that higher concentrations of lipids in the alcoholic solution obtained by keeping the volume ratio buffer/alcohol constant results in larger vesicles. 1 J. M. Kremer, M. W. Eskev, C. Pathmamanocharan and C. Wiersema, Biochemistry, 1977, 16, 3932. 2 V. Eck and J. F. Holzwarth, in Surfactants in Solution, ed. K. L. Mittal and B. Lindman (Plenum, New York, 1984), vol. 3, pp. 2059-2079. Prof. P. Fromherz (University of Ulm, West Germany) replied: From the chemical structure of ethanol one would expect it to be an edge actant.’ However, I have no estimate of its edge activity as compared to the cholates.The size of vesicles is governed by the kinetics of the growth of the lipid aggregates and by the rate of change of the edge tension for a particular recipe of vesiculation. The data available on the method of ethanol injection233 and our own preliminary studies are too incomplete to provide a solid basis for an interpretation. 1 P. Fromherz, Chem. Phys. Lett., 1983,94, 259. 2 S. Batzri and E. D. Korn, Biochim. Biophys. Acta, 1973, 298, 1015. 3 J. M. Kremer, M. N. Eskev, C. Pathmamanochoran and C. Wiersma, Biochemistry, 1977, 16, 3932. Dr R. E. Dale (Christie Hospital, Manchester) addressed Prof.Levine. (1) For DPH in the multibilayer systems you report a qualitative change in the derived orientational distribution function between interpretation in terms of P2( cos p ) and both P,(cos p ) and P~(COS p). How physically realistic can the latter result then be considered, since presumably inclusion of P,(cos p), if it were possible, would result in a further qualitative change in the derived distribution? (2) You have reported heterogeneity of probe lifetimes in these systems. This might arise from heterogeneity of probe environment in the membrane. Would you expect these two putative populations to have different orientational distributions and rotational diffusion coefficients, and is there any possibility of differentiating them, perhaps by time-resolved AFD? In the case of DPH at least, might they be correlated with the ‘parallel’ and ‘perpendicular’ populations indicated by the minima in the derived orientational distribution functions ? (3) More generally, is there substantially more information to be gained from time-resolved as opposed to steady-state AFD experiments? Prof.Y. K. Levine (University of Utrecht, The Netherlands) replied. (1) It is difficult to assess quantitatively the importance of the higher-order param- eters, ( PL) with L > 4, on the reconstruction of the orientational distribution function f( p ) without knowledge of the function one is trying to approximate. There are, however, some indications that knowledge of (P6) and higher terms will not change the form of the distribution substantially.It can be shown on simple mathematical grounds that if f(p) is evaluated up to the Lth order parameter ( L even), it will have at most L / 2 - 1 maxima in the interval 0 < p < ~ / 2 . On the other hand, X-ray diffraction experiments on the orientation of the hydrocarbon chains in oriented bilayer systems indicate the presence of only one maximum in this interval and that for the gel Lpr phase. This supports the view that knowledge of (PJ and (P4) is sufficient to describe the overall features of f(P). The question which remains to be answered is whether this knowledge yields a correct description of the detailed form off@) around /3 = ~ / 2 . I.e. how sure are we that the DPH molecules can lie parallel to the bilayer surface.In the paper we present arguments based on the consistency of this description for a number of systems and the chemical structure of the two probe molecules. One can say that at best this is circumstantial evidence. However, we can also note that the longer, but chemically similar, p-carotene molecules are also found to have a propensity to lie parallel to theGeneral Discussion 77 bilayer plane in many membrane systems. This behaviour is quite clearly demonstrated by linear dichroism experiments' which only yield (P2) and resonance Raman experi- ments2 which yield both (P2) and (P4). (2) We have been unable to find any correlation between the heterogeneity of probe lifetimes and a putative heterogeneity in the probe populations. In fact, our results indicate that the steady-state fluorescence depolarization experiments are essentially sensitive only to the long-lifetime component of the fluorescence decay. This arises simply from the small contribution of the short-lifetime component to the total time- integrated fluorescence intensity.Thus the heterogeneity indicated is associated in its entirety with the long-lifetime component for both DPH and TMA-DPH molecules. It is important to note in this context that our analysis is based on the assumption of a dynamically homogeneous probe population. The fact that for DPH one obtains an orientationally heterogeneous population strongly suggests that here our description of the probe behaviour is oversimplified. The point is that steady-state AFD experiments do not yield sufficient information to resolve this problem.Time-resolved experiments, however, may provide the answer. (3) In principle time-resolved AFD experiments provide means for following directly the decay of the time autocorrelation functions Gk( t ) . In this sense they are potentially more interesting than steady-state measurements, which only yield the time-integrals of these functions. For this reason we are currently carrying out such experiments with the synchrotron radiation source in Daresbury. Our expectations, however, have taken something of a knock as a result of the cumbersome analysis of the single-photon counting experiments. The main problem being the use of iterative least-squares methods for the deconvolution of the data and the setting up of statistical criteria for the goodness-of-fit.So far we have been able to show that the order parameters obtained from the time-resolved experiments are in good agreement with those obtained from steady-state ones, but are having difficulty in obtaining the kinetics of the decays in a consistent manner. This work is being continued. On the other hand, we have shown that steady-state AFD experiments coupled with measurements of the fluorescence decay afford a convenient, reliable and fairly quick method for obtaining quantitative informa- tion about the dynamic behaviour of the probe molecules. 1 B-A. Johansson, G. Lindblom, A. Wieslander and G. Arvidson, FEBS Lett., 1981, 128, 97. 2 M. Van de Ven, M. Kattenberg, G. van Ginkel and Y. K. Levine, Biophys. J., 1984, 45, 1203.Prof. G. Lindblom (University of Umea", Sweden) said: I would just like to point out that we have found that the lateral diffusion coefficient of lipids in lecithin bilayers does not depend on the molecular order (e.g. by increasing the cholesterol concentration).' I think that this supports Prof. Levine's ideas. 1 G. Lindblom, B-A. Johansson and G. Arvidson, Biochemistry, 1981, 20, 2204. Prof. J. F. Holzwarth (Fritz Haber-Institut, Berlin, West Germany) asked Prof. Levine: In your paper about e.s.r. and angle-resolved fluorescence depolarization ( AFD) experiments you mention the influence of probes on the bilayer structure. In fig. 7 we have demonstrated this influence for DMPC and could show that even at 1/250 ratios there is still a clear reduction of the transmission at temperatures above the phase- transition temperature T, , even if the normalized order-parameter/ temperature depen- dence still shows inside 0.5 "C the expected midpoint T, for pure lipids, as is demon- strated in fig.8. Another problem which is associated with the AFD measurements is that you sense the time range of nanoseconds, but the major part (80% with respect to AHHpT) of the phase transition occurs in the ps to ms time regime, as we have demon- strated' (see also the following contribution).78 General Discussion h W c -e 0 -2 a v - 0.1 1 I I I 15 20 25 T / "C Fig. 7. Turbidity us. temperature dependence of unilamellar vesicles of dimyristoylphos- phatidylcholine (DMPC) containing different concentrations of the probe molecule diphenyl- hexatriene (DPH): (---) 1/135, ( - - - ) 1/250, (-) none. 2.7 mmol dm-3; A = 300 nm. R == 50 nm; [DMPC] = I I 20 25 T / "C Fig. 8. Normalized phase transition us. temperature dependence of fig. 7: (- - -) 135/1, T, = 23.5 "C; (-) 500/1, T, = 24.2 "C. This disadvantage of fluorescence polarization lifetime measurements, and also with respect to the time range of the e.s.r. experiments, can be overcome by using the time dependence of the fluorescence anisotropy as we did between 5 ps and 100 ms.' My question is how important is the influence of the probe molecules with respect to the disturbance of the bilayer structure, especially if one senses the immediate environment of the probe? 1 A. Genz and J. F. Holzwarth, Eur. Biophys. J., 1986, 13, 323. 2 J. F. Holzwarth, V. Eck and A. Genz, in Spectroscopy and Dynamics of Molecular Biological Systems, ed. P. Bayley and R. Dale (Academic Press, London, 1985), pp. 351-377.General Discussion 79 Prof. Y. K. Levine ( University of Utrecht, The Netherlands) replied: The experiments reported in the paper were carried out at temperatures well above those of the phase transitions of the various lipid systems. Control AFD experiments yields the same fluorescence depolarization ratios for probe/lipid ratios between 1/ 1000 and 1/ 125. The ratio of 1/250 was chosen arbitrarily for experimental convenience. Similarly, no dependence of the e.s.r. lineshape on the probe/lipid ratio was observed over the same range. As we carry out our experiments above the phase transition we expect the probe molecules to be uniformly distributed in the lipid multibilayers. We have no experimental evidence of clustering or phase-separation processes which may well play a role around the phase-transition temperature of the bilayers. The point we are making is that two probe techniques, utilizing different probe molecules, yield the same information about the orientational order and dynamics in a variety of lipid bilayer systems. To my mind this strongly suggests that the perturbation of the bilayer structure by the probe molecules is much less important than has been suggested.