Chapter 1 Protein-lipid interactions and membrane heterogeneity

1

A. Watts (Ed.), Protein-Lipid Interactions 0 1993 Elsevier Science Publishers B.V. All rights reserved

CHAPTER 1

Protein-lipid interactions and membrane heterogeneity Ole G. MOURITSEN'>* and Rodney L. BILTONEN2 'Department of Physical Chemistv, The Technical University of Denmark, Building 206, DK-2800 Lyngby, Denmark, 2Department of Biochemistry, University of nrginia, Charlottesville, VA 22908, W . A .

Abbreviations AC DS DSC DMPC DPPC DSPC LUV

alternating current distearoyl differential scanning calorimetry dimyristoyl phosphatidylcholine dipalmitoyl phosphatidylcholine distearoyl phosphatidylcholine large unilamellar vesicles

MLV NMR PA PC PE PS SUV

multilamellar vesicles nuclear magnetic resonance phosphatidic acid phosphatidylcholine phosphatidylethanolamine phosphatidylserine small unilarnellar vesicles

1. Perspectives and overview 1.1. Lipids, proteins, and the biological membrane

The conventional picture of the biological membrane is that of Singer and Nicolson [ 1,2] who suggested viewing the biological membrane as a fluid mosaic bimolecular lipid layer in which the various membrane components, such as proteins, enzymes, and polypeptides, are embedded or attached to. The crucial property of this molecular assembly is the bilayer fluidity which assures sufficient lateral mobility of the membrane components to support biological function. A fuller picture of the biological membrane considers the fluid bilayer as only part of a composite structure, see Fig. 1, consisting, in the case of eucaryotic cells, of a glycocalyx structure on the outside and of a cytoskeletal scaffolding of proteins on the inside. The particular engineering of this composite and its various parts has imparted the biological membrane with unique physical properties [ 3 ] . The *

Associate Fellow of The Canadian Institute of Advanced Research.

2

Fig. 1. Schematic illustration of an eucaryotic cell membrane which highlights the membrane as a composite of a fluid lipid bilayer sandwiched between the carbohydrate glycocalyx on the outside and the cytoskeleton on the inside. Intercalated in the lipid bilayer are shown schematically various integral and peripheral proteins and polypeptides (illustration by Ove Broo Ssrensen).

membrane and its stability are intimately related to the biological solvent, water, to which the membrane owes it existence via the hydrophobic effect. It is the viewpoint of the present chapter that the lipid-bilayer matrix is not only a mosaic structure as implied by the Singer-Nicolsen model but furthermore is highly heterogeneous due to strong correlations between the various molecular constituents. These correlations lead to biologically differentiated regions on different lengthscales. The heterogeneity is of a highly dynamic nature. The biological functioning of the membrane is governed by the structure and the dynamics of the membrane composite in general and by the structure and molecular organization of the fluid bilayer in particular. To a large extent, the relationship between function and structure is controlled by physical forces, i.e. the mutual interactions between the molecular constituents of the bilayer, the forces between the bilayer and the cytoskeleton, as well as the interactions between the clycocalyx and extramembrane structures such as macromolecules and other membranes. In the present chapter we shall restrict ourselves to phenomena within or closely related to the fluid bilayer component of the membrane and discuss a certain subset of the physical forces which are responsible for the interaction between the lipid bilayer and integral membranebound proteins and enzymes. We shall study how these interactions manifest themselves in terms of the protein’s influence on the lipid structure and the

3

influence of the lipid-bilayer structure on the hnctioning of certain enzymes. In particular, we shall address cooperative effects due to lipid-lipid and protein-lipid interactions, which may lead to dramatic density fluctuations and compositional fluctuations within the bilayer. We shall discuss how such effects lead to a heterogeneous organization of the membrane which in turn may control membrane function. Furthermore, we shall pursue this line of reasoning by investigating how the hnctional effects of various membrane-active agents, such as drugs, may be related to their ability of physically modulating the membrane heterogeneity. 1.2. Phase transitions and membrane heterogeneity Lipid bilayers composed of a single lipid species have several different types of phase transitions [4-61. The transition which is believed to have a substantial effect on membrane function is the main gel-to-fluid phase transition which takes the membrane from a low-temperature solid (gel) phase, characterized by acyl-chain order, to a high-temperature liquid (fluid) phase in which the lipid-acyl chains have a substantial degree of conformational disorder and fast lateral diffusion. For lipid bilayers composed of more than a single lipid species, complex phase behavior arises including membrane states with lateral phase separation. The phase behavior is further complicated when integral membrane proteins are present. This phase behavior and its consequences for membrane organization are conveniently studied by means of simplified model membrane systems, cf. section 2, below. The main gel-to-fluid phase transition of lipid bilayer membranes provides a ubiquitous phenomenon to probe the physical effects of the interaction between lipids and integral membrane proteins. By being extremely sensitive to the molecular interactions of the bilayer membrane system, the appearance of this transition provides information on the fundamental principles underlying protein-lipid interactions in model membranes and ultimately in biological membranes. It is an often overlooked fact in membranology that the manyparticle character of the membrane bilayer supports strong correlations that are associated with the cooperative phenomena. These correlations may be long-ranged and manifest themselves in terms of phase transitions or phase separation, or they may be short- or medium-ranged in the neighborhood of phase transitions. The correlations between the molecules of the bilayer lead to a highly non-trivial lateral organization of the membrane, often characterized by a substantial degree of static or dynamic membrane heterogeneity. In both static and dynamic lipid-bilayer heterogeneity, the membrane displays biologically differentiated regions which are considered important for a variety of functions associated with membranes.

4

2. Membrane heterogeneity 2.1. Static membrane heterogeneity Static membrane heterogeneity [2,7-91 is generally caused by some static stabilizing condition which can be of either mechanic or thermodynamic origin. Examples of static lateral heterogeneity in membranes include (i) macroscopic morphologically different domains induced, for example, by coupling between different neighboring membranes in a stack, (ii) aggregation of proteins in single membranes due to thermodynamic phase separation or due to specific couplings to the cytoskeleton, and (iii) lateral phase separation of the lipids in the bilayer driven either by intrinsic thermodynamic forces or by external fields or concentration gradients. 2.2. Dynamic membrane heterogeneity The phenomenon of dynamic membrane heterogeneity is a consequence of the many-particle nature of the bilayer molecular assembly. Large assemblies of mutually interacting molecules are subject to spontaneous fluctuations, e.g. in lateral density and in local molecular composition. Such fluctuations prevail near phase transitions, e.g. the gel-to-fluid transition of lipid bilayers 161. Microscopically and mesoscopically these fluctuations manifest themselves in the dynamic formation of domains or clusters of correlated lipid molecules of a structure different from that of the bulk lipid matrix. This type of dynamic, fluctuation-induced lipid-bilayer heterogeneity is associated with a time scale characteristic of the lateral fluctuations in density or composition. Obviously this time scale is strongly dependent on membrane composition and on the thermodynamic conditions of the membrane. A similar dynamic heterogeneity can be defined in relation to protein-lipid interactions in membranes, where it is often found that on a certain time scale, the lipid environment adjacent to an integral membrane protein has a structure and composition which is different from that of the bulk lipid matrix [ 10-121 even though there is no specific chemical binding of the lipids to the protein. Insofar as these ‘boundary’ lipids may be in dynamic equilibrium with the bulk lipids, the nature and character of the lipid region next to the proteins are strongly coupled to the overall state of the membrane. A second type of dynamic membrane heterogeneity [13] occurs in certain membrane-enzyme systems where the products of the enzymatic process induce phase separation and membrane heterogeneity, which for enzymes like phospholipase A2 [ 141 catalyzes and further enhances the reaction. We shall describe this rather general phenomenon in section 6.

5

The problem of which factors control dynamic membrane heterogeneity has received little attention. An exception is the heavily debated question of the possible existence of a lipid annulus around integral membrane proteins, a question which has now been solved via proper time-scale considerations [lo]. There are several reasons for the limited interest in dynamic lipid-bilayer heterogeneity in membranes. Firstly, many membranologists consider the lipid bilayer as a mere anchor place for more interesting membrane components, such as proteins, receptors, and enzymes, and assume that the lipid properties have little influence on hnction. Secondly, it is extremely difficult to measure experimentally dynamic heterogeneity directly since most techniques involve implicit averaging procedures which do not allow a resolution of the lateral structure. This is true of thermodynamic and thermomechanic measurements as well as of most spectroscopic techniques. One of the obstacles in obtaining direct information on dynamic heterogeneity of the lipid bilayer is that this type of heterogeneity, which occurs on a time scale of typically 10-4-10-' s, is manifested on mesoscopic length scales of the order of lOO-lOOOA which is not easily accessible by using current experimental techniques. The most direct evidence of dynamic heterogeneity in lipid bilayers has been provided by fluorescence lifetime heterogeneity measurements [151. However, dynamic heterogeneity may be inferred indirectly from measurements of membrane response functions, such as the specific heat or the lateral compressibility, which are bulk, integral measures of the thermal density fluctuations [5,16,17]. We shall discuss the indirect evidence obtained from thermodynamic measurements in section 3.1. Some theoretical evidence in favor of thermally induced dynamic membrane heterogeneity has been obtained by theoretical calculations on specific molecular interaction models of lipid bilayers with and without proteins [6,12,17-201. The theoretical calculations involve computer simulations on large arrays of lipid molecules taking into account the acyf-chain conformations and their mutual interactions. The type of information which can be obtained from such computer simulations on phospholipid bilayers is illustrated in Fig. 2. Dynamic heterogeneity may be described as a consequence of a lipid-domain formation process as illustrated in Fig. 2. The domains can be thought of as membrane defects on the nanoscale. Associated with the lipid domains is a network of interfaces bounding the domains within the membrane. These interfaces have very special molecular packing properties; they are 'soft' and have a low interfacial tension. The hypothesis of the present chapter is that these lipid domains and the domain interfaces, in their capacity of being large defects and lines of defects, respectively, may support specific passive and active membrane hnctions.

6

Fig. 2. Dynamic membrane heterogeneity: Planar organization of lipid-bilayer model systems as obtained from computer-simulation calculations. The one-component DPPC bilayer in (a) displays density (area) fluctuations and the binary DMPC-DSPC lipid mixture in (b) displays compositional fluctuations. In (a), dark and light areas denote the gel and the fluid phase, respectively; in fb), dark and light areas denote fluid-phase DMPC and fluid-phase DSPC, respectively. Courtesy of Kent Jargensen and M.M. Sperotto.

3. Evidence of heterogeneity in lipid bilayers One-component and few-component phospholipid bilayers constitute a simple class of model-membrane systems which are well defined and whose physical properties can conveniently be studied by powerful experimental and theoretical techniques [3,4]. These phospholipid bilayers can be reconstituted with proteins and enzymes, and the manifestations of protein-lipid interactions can then be studied in a setting which often is sufficiently simple to permit a clear interpretation. In the following we shall focus on two approaches to study membrane heterogeneity, thermodynamic experiments and microscopic modelling. The results presented form the basis for understanding the relationship between protein-lipid interactions, membrane heterogeneity, and membrane functions which are discussed in subsequent sections.

3.1. What can thermodynamics tell us? The structural and compositional heterogeneity of a lipid matrix is determined by the nature and strength of intramolecular and intermolecular interactions and their dependence upon the environment in which they exist. The general information describing the existence of distinct phases and regions of phase coexistence are contained in the phase diagram of the system. The structural detail of the lipid within specific regions is best determined by techniques such as X-ray diffraction and NMR [4]. The thermodynamics and cooperativity of

7

lipid phase changes, which are manifestations of the energetic details of the accessible structures of the lipid bilayer, are most directly and precisely obtained by high-sensitivity differential scanning calorimetry (DSC).

3.I . I . Differential scanning calorimetry Differential scanning calorimetry (DSC) has been widely used to obtain estimates of the enthalpy change, A H " , the melting temperature, T,, and hence entropy change, ASo, of various transitions of lipid systems. The types of phase changes investigated include the subphase-to-gel transition [21], the gel to 'ripple phase' transition [22], the lamellar-to-hexagonal phase transition [23], as well as the gel-to-fluid transition [24]. The last of the above-mentioned transitions is most easily investigated by DSC under essentially equilibrium conditions. For this reason we will limit this discussion to the gel-to-fluid transition of bilayer systems with emphasis on the phosphatidylcholines having identical saturated acyl chains. The thermodynamic changes associated with the gel-to-fluid transition of phosphatidylcholines are dominated by chain melting. In fact, AH" and AS" are, to the first approximation, linear functions of the acyl-chain length [25]. This implies that the measured changes in energy are the result of the disruption of intra- and intermolecular van der Waals interactions and transgauche isomerization and that ASo is primarily the result of increased rotational entropy about the carbon-carbon bonds of the acyl chain. The large change in entropy is a reflection of the increased degeneracy of the energy states of the lipid in the fluid phase. A major increase in surface area and decrease in bilayer thickness is associated with this transition, resulting in a positive volume change, making the transition extremely sensitive to pressure changes [26]. The interactions which dominate the gel-to-fluid transition in bilayers of phosphatidylcholines having identical acyl chains appear to be limited to a single monolayer. Thus, there is no thermodynamic coupling between the two monolayers of the bilayer. This was shown by Schmidt et al. [27] and by Sillerud and Barnett [28] using small unilamellar vesicles (SUV) for which the NMR signals of the inner and outer choline methyl protons are distinguishable. Using a paramagnetic lanthanide ion on one side of the vesicle to change the value of T , and to alter the proton-chemical shift of that monolayer, they demonstrated that the inner and outer monolayers exhibited distinguishable melting profiles. Modelling of the transition of bilayer lipids with identical acyl chains need thus only consider it as a monolayer transition, at least to the first approximation. This is not the case for lipids which have dissimilar acyl chains or for lipids with identical acyl chains in the presence of ethanol[29] which promotes interdigitation at high concentration. The thermodynamics associated with the gel-to-fluid transition of such systems has been investigated by Huang

8

and co-workers [30] using DSC, who found a relatively simple mathematical relationship between the difference in acyl-chain length and T , [30]. The aggregation state of the lipids varies widely depending upon the details of preparation. These include at least three types of bilayer systems. Multilamellar vesicles (MLV) consisting of several bilayers separated by an aqueous space of about 3nm[31]. The largest outside lamellae have a radius of 150nm or more and contain 105-106 molecules. Small unilamellar vesicles, made by sonication of the multimalellar vesicles, have a diameter of 20-30 nm and contain about 4000-5000 molecules in the outer monolayer [32]. Large unilamellar vesicles (LUV), having a diameter of about lOOnm and containing about lo5 molecules in the outer monolayer can be made by fusion of S U V [33,34] or by extrusion of multilamellar dispersions [35]. All three types of bilayer systems undergo a well-defined gel-to-fluid transition. The primary differences among these various preparations are their melting temperatures and the breadth of the transition. For DPPC MLV, T , is about 41.3"C, with a half-width at half-height of the heat capacity function of less than 0.1"C [24,33]. LUV melt at approximately the same temperature, but have a half-width at half-height of the heat capacity function on the order of a few tenths of a degree[33]. The T , for S U V is about 4" lower than for MLV and the half-width at half-height is on the order of 3°C [5,33]. The enthalpy change for MLV and LUV appears to be approximately identical within experimental error, whereas it has been reported that the enthalpy change of small unilamellar vesicles is somewhat smaller [33]. This latter result may, however, be incorrect as will be discussed later. The heat capacity has been analyzed in terms of a simple model in which the lipid monolayer exists as clusters of either gel or liquid in a sea of the other. Assuming that these clusters do not interact, the average cluster size, for example, can be estimated from double integration of the heat capacity function which yields the apparent partition function of the system [ 16,361. Analysis of the data for DPPC SUV provided an estimate of the average cluster size at the transition temperature on the order of 80 lipid molecules. However, no information about the shape of the clusters could be obtained. 3.1.1.1. A simple two-state model. The heat capacity function C J T ) for any transition contains all the information necessary to describe the changes in the energy distribution of a system as the transition progresses. Unfortunately, this information is difficult to extract in a model-independent fashion. In order to provide a basis with which to quantitatively interpret experimental heat capacity data, we will introduce a simple two-state model for the gel-to-fluid transition of single component lipid bilayers [37]. This model is analogous to the one employed by Marsh et al. [38] to describe the transition. This model assumes that each lipid exists in either an all-gel (a) or an all-liquid (b) state. If we assign a statistical weight of unity for a vesicle of size N in the all-gel state, the Gibbs energy (relative to an all-gel vesicle) of a vesicle containing n b liquid molecules

9

with nab unlike nearest neighbors is

where

Q ( N , nb, nab) is the number of ways of distributing nb molecules with nab unlike nearest-neighbors over a lattice of size N , and gas, g b b and g a b are Gibbs energies associated with the various nearest-neighbor interactions. It should be noted that for a hexagonal lattice

where g b and ga refer to the Gibbs energies associated with intramolecular interactions and trans-to-gauche isomerization of single lipid molecules, respectively. Although this model contains no structural information and is much simpler than that used in the Monte Carlo calculations described in the next section, it is capable of reproducing the major features of the more complex model as well as the experimental data. Its main advantage is that only one unknown parameter, E , is required since Aho and Aso are determined experimentally. Furthermore, it allows a quantitative interpretation of C J T ) in terms of specific interaction terms. For example, if a vesicle undergoes an allor-none transition (the equivalent of a first-order transition for a finite system), E . = co.In general, the abruptness (cooperativity) of a transition is directly related to the magnitude of e. Whether or not the transition of the bilayer vesicles from the gel to the fluid state is first order is a question which has received much attention. The conjecture that indeed it is a true first-order phase transition is based upon the fact that the temperature span for gel-to-fluid multilamellar vesicles of very pure lipid is on the order of O.l"C, first shown by Albon and Sturtevant [24]. An example of a DSC scan of DPPC MLV is shown in Fig. 3. These data were obtained at a scan rate of O.l"C/h, eliminating any effect of scan rate on the shape of the heat capacity function. As can be seen, the transition is continuous and shows no evidence of a discontinuity at the transition temperature although CJT,) is very large, lo5 kcal/mol deg. The temperature span of the data in Fig. 3 is only 0.2K and the transition half-width at half-height is 0.076K. It has been argued that the transition is first order but that the lipid contains a very small amount of contamination [29]. However, if the transition is first order, extremely

10

901 80

Fig. 3. The excess heat capacity of a dispersion of DPPC multilamellar vesicles calculated from a DSC scan at a scan rate of O.l"C/h.From Biltonen (1990) [ 5 ] , with permission.

E

1500

500

290

295

300

305

310

315

TEMPERATURE (K)

320

325

2 10

Fig. 4. Heat capacity function of DPPC S U V obtained at a heating scan rate of 10"C/h. The small, but sharp peak at about 41" is due to a residual amount (5%) of MLV The solid points are calculated from Monte Carlo calculations. Aho = 8.7 kcal/mole, T , = 3 10 K, E = 0.45RT.

sharp transitions should also be observed with the unilamellar vesicles since they are contaminated to the same extent. This, hovewer, is not the case as seen in Fig. 4,where the excess heat capacity function for DPPC S U V is presented. A simple calculation demonstrates that the experimental C, results for small unilamellar vesicles do not exhibit the equivalent of first-order behavior. Systems of finite size N (i.e. a vesicle) will undergo a transition in which all vesicles exist in either a totally gel state or a totally fluid state at equilibrium near T , if the transition is truly first order. This is equivalent to stating that E = GO, but

11

a finite width to the transition will be observed because of the system’s finite size. The equilibrium poise of the macrosystem is defined by a fraction,fl, of vesicles in the fluid (liquid) state: exp(-AGo/RT) = 1 + exp(-AGO/RT)’

(5)

AGO = N(Aho - T AS’).

(6)

where

The excess heat capacity function of this system is

For an S U V (N Cp(Tm) =

21

4 x lo3) of DPPC:

N (Ah’)’ 4RTk

=

1.5 x lo5 caVmol deg.

This value of Cp(T) per mole of lipid is about equal to the largest value measured for MLV and approximately two orders of magnitude larger than that measured for S W (see Fig. 3). Thus, it has been suggested that the rather broad transition for SUV is the result of heterogeneity of the vesicles. However, fractionation of SUV on a sizing column indicates a constant thermodynamic character of the small unilamellar vesicles [5]. That is, different fractions of the S U V preparation have essentially identical melting temperatures and identical half widths. In fact, it appears that there are only two types of DPPC bilayer vesicle populations based upon T,, those that melt at about 37°C and those that melt at about 41°C. On this basis it has been argued that the S U V are homogeneous and that their gel-to-fluid transition is not of the all-or-none type ~51. The ability of the model defined by Eq. (1) to determine the heat capacity function for SUV is demonstrated in Fig. 4. The solid points in this figure represent the results of a Monte Carlo calculation assuming that each lipid molecule on a hexagonal lattice of size 65x65 exists in either a gel or a liquid state. This lattice is the approximate size of the outer monolayer of SUV The results shown in Fig. 4 were calculated assuming T,n = 310.3 K, Aho = 8.7 kcal/mole lipid (the experimental value obtained for MLV) and E = 0.045RT. As can be seen, the fit to the experimental data is reasonable and the significant ‘wings’ on each side of Tm, predicted by Monte Carlo calculations using a more complex model [ 171 (cf. section 3.2), are observed.

12

The existence of these wings, described in detail in the following section, is an important feature of these curves. They indicate the existence of significant thermodynamic fluctuations at temperatures distant from T , . More important from an experimental view point is that they must be taken into account to accurately calculate Ah" for the transition. The Aho for the gel-to-fluid transition of S W was originally estimated to be 6.1 kcal/mol[33]. However, using this value of Aho, no value of E which yielded a good representation of the data was found. However, when the DSC baseline was carefully determined over a very broad temperature range ( T , f 15"C),the existence of the wings became apparent and the integrated C J T ) now produced a value similar to the Aho for h4LV and L W and the results of the Monte Carlo calculation agreed well with the experimental data, as shown in Fig. 4. The results are consistent with the idea that the transition is a weakly first-order (or continuous) transition with large thermodynamic fluctuations existing both above and below the transition temperature. This experimental result is consistent with the predictions of Monte Carlo calculations discussed in section 3.2. As mentioned, thermodynamic fluctuations distant from T , can be significant. The fluctuations, which appear as 'wings' on the C,(T) curve, are most evident in lipid systems which exhibit broad transitions, such as SUV. They are more readily seen in mixed lipid systems. As an example, Fig. 5 shows the excess C J T ) hnction for MLV of a 1:l DMPC-DSPC mixture [32]. The physical basis of these fluctuations will be discussed in the following section. What is important to note here is that they reflect the existence of microscopic heterogeneity in the lipid structure over a very broad temperature range. The experimental DSC results with DPPC vesicles coupled with Monte Carlo analysis using a very simple model lead to three important conclusions: (i) the gel-to-fluid transition is best described as a continuous, or weakly first-order with significant thermodynamic fluctuations at temperatures distant from T , ; (ii) these fluctuations are readily detectable in the excess C,(T) of S U V and of mixed lipid systems; and (iii) the overall Aho for DPPC S U V has been underestimated in the past because the contributions of the fluctuations to the excess C, were not properly taken into account. It appears that the Ah" for DPPC Sw LUV and MLV are essentially identical. While this discussion has focused on experimental results of saturated phosphatidylcholines, it is likely that the general aspects of these conclusions apply to most bilayer vesicle systems.

3.1.2. Volume perturbation calorimetry The large volume change for the gel-to-fluid transition makes it very sensitive to changes in hydrostatic pressure. In the case of DPPC, the transition curve is shifted to higher temperature without any significant change in the shape of the heat capacity function [5,39]. This shift in T , is given by the Clausius-

14

13

1

.

0 0

-

h

I-

20

30

40

50

TEMPERATURE ( C ) O

Fig. 5. Heat capacity function of a 1:l mixture of DMPC-DSPC in multilamellar form. Data obtained at a scan rate of 10"Cih.

Clapeyron equation dTm - Aho - 0.025"C/atm. (9) dP Aso Thus examination of the pressure dependence of T , will yield the value for A V o if Aho = T,Dso is known. The pressure dependence of the transition has been used as a means to perturb the lipid system and monitor the progress of the transition as it strives to attain the new equilibrium position. In this manner estimates of the time scale for the relaxation kinetics of MLV using Raman spectroscopy[40] and X-ray diffraction [41] have been obtained. These methods, however, used a large perturbation so that the relaxation process occurred at a condition far removed from Tm and the time scales do not necessarily correspond to equilibriumfluctuation time scales. Johnson and co-workers [42] have devised an experimental approach that uses very small pressure perturbations. This technique, called volume perturbation

14 7,

313 6.5 314.15 314.65 315.15 T iK

Fig. 6. The relative amplitude A of the temperature response of a dispersion of DPPC multilamellar vesicles to a periodic compression-decompressionas a function of temperature adjusted for changes in the average pressure as the lipid melts. The frequency of the perturbation was 0.1 Hz. From Biltonen (1990) [ 5 ] , with permission.

calorimetry, yields relaxation times which should resemble the characteristic times for the lipid fluctuations in the gel-to-fluid transition region under equilibrium conditions. The volume perturbation calorimeter is analogous to an AC calorimeter [43], but induces a small hydrostatic pressure change rather than a temperature change of the system in a periodic fashion. The change in the equilibrium poise of the system results in an absorption of heat from or release of heat to the solvent, which is monitored as a temperature change. Because the pressure perturbation is isotropic throughout the solution, a wide frequency of pressure oscillations can be used. The time-dependent amplitude of the temperature change is monitored as a function of the perturbation frequency. At very low frequency, the amplitude of temperature response to the pressure perturbation as a function of temperature is proportional to the equilibrium heat capacity function, as seen in Fig. 6. The apparent T , and the half-width of the capacity function estimated from these data are in good agreement with the parameters obtained from DSC results, cf. Fig. 3. The frequency dependence of the response function, an example of which is given in Fig. 7, is consistent with a single relaxation process with a characteristic time, T, in the range of 50 ms to 5 s throughout the transition. These characteristics are qualitatively identical for all phosphatidylcholines with saturated identical fatty acid chains. Their mean relaxation times are of the same order of magnitude and all show a maximum in T in the vicinity of 75% through the transition [44]. It should be noted that the conditions of these experiments result in only a 2-3% change in the degree of melting of the lipid, so that the relaxation is in the linear response region. Therefore, the fluctuationdissipation theorem should apply and the time scales that are observed should be

15

I

T

-

I

-2.0

-1.0

0.0

1.0

1 1

J

2.0

LOG (v)

Fig. 7. The amplitude of the temperature response of a DPPC vesicle dispersion to periodic pressure changes as a function of frequency at 3 14.45 K. The best-fit curve is calculated by assuming a single relaxation process for the lipid transition. From Biltonen (1 990) [ 5 ] , with permission.

those that reflect isothermal fluctuations in the lipid structure under equilibrium conditions. Large unilamellar vesicles do not exhibit a pronounced maximum in T and have a characteristic time of about 80ms. This latter result has been confirmed by Freire and co-workers [43] using a multifrequency heat capacity calorimeter. Ye[45] discovered that the frequency of the relaxation spectrum of MLV was much sharper than expected for a simple exponential relaxation process. Following Yang and Nagle [46] who used Kologomov-Avrami theory to analyze the kinetics of the subgel-to-gel transition, Ye [45,47] applied the same model to the analysis of the relaxation spectrum in the frequency domain. In this model, it is assumed that constant radial growth occurs following the pressure perturbation and that the relaxation time has a dimensionality related to the geometrical shape of the cluster. Ye showed that the data for the phosphatidylcholines analyzed in this manner were consistent with a dimensionality significantly greater than 1. The dimensionality varied from about 2.0 for DMPC, to about 1.8 for DPPC to about 1.5 for DSPC. These results suggested that, first, the theory was applicable to the relaxation process and that, secondly, the frequency spectrum of the relaxation spectra provided a measure of the geometric shape of the cluster. In terms of Eq. (1) the geometric shape is dependent primarily on the magnitude of interaction energy, E , between lipids in the fluid and the gel states. The more repulsive (or less attractive)

16

this interaction, the larger and more circular will be the clusters. As the dimensionality decreases the clusters become smaller and more ramified. This analysis provides, in principle, a means to compare theory with both equilibrium and kinetic results. 3.1.3. The effect of anesthetics on the gel-to-JEuid transition The effect of anesthetics on lipid behavior and their mechanism of action is a much debated issue. What appears to be clear from calorimetric [39,48,49] and other studies[50,51] is that both general and local anesthetics reduce T , of the gel-to-fluid transition without changing AH'. DSC studies have shown that anesthetics also cause a broadening of the transition C, curve. This latter result could be due to changes in the chemical potential of the anesthetic in the aqueous phase as the transition progresses, due to the stabilization of the unlike nearest-neighbor interactions (reduction of c) or both. Ueda [52] has suggested that the first effect is the cause of such broadening in most DSC experiments because of the relatively large lipid concentration used in the experimental study. This conclusion was based on the observation [52] that halothane did not apparently cause broadening of the transition in DPPC unilamellar vesicles as measured by turbity measurement. DSC experiments with 1.4 mM hexanol at quite low lipid concentration ( < 0.1 mM) using LUV indicates that the transition is about two times as broad as in the absence of the alcohol (IS.Thompson and R. Biltonen, unpublished). The discrepancy between the two types of experiments is not clear. The effect of anesthetics on lowering the melting temperature is the result of preferred solubility in the liquid state. In terms of the two-state model described in section 3.1.1.1 , this effect is interpreted as a reduction in g b b relative to gas. The anesthetic-induced broadening of the phase transition can be interpreted as a reduction in c near T,. This interpretation is consistent with the Monte Carlo simulations of Jmgensen et al. [53]. It is interesting to note that the anesthetic dibucaine has very little affect upon the kinetic behavior of DPPC MLV except to reduce the dimensionality in a continuous fashion from 1.8 to 1 as anesthetic is added[54]. This result suggests that the clusters become smaller and more ramified. This result is also consistent with an interstitial model for the effects of anesthetics on lipid transitions [53].

3.2. What can microscopic modelling tell us?

A powerful approach to study cooperative phenomena which leads to information

on the microscopic as well as the macroscopic level in membrane systems is Monte Carlo computer simulation of molecular interaction models [ 171. The simulation techniques fully allow for the correlations in the thermal fluctuations

17

and they can provide a direct picture of the microscopic phenomena which underlie macroscopic events, e.g. in terms of lateral membrane organization. A number of computer-simulation studies have been carried out on a class of molecular interaction models which take accurate account of the acyl-chain conformational degrees of freedom (for recent reviews, see Mouritsen [6,17]) as well as the lateral mobility of the membrane components within the membrane plane. Fig. 2 (above) shows examples of typical membrane structures as obtained from computer simulations near the main transition in a pure DPPC lipid bilayer and for a binary DMPC-DSPC lipid mixture in the fluid phase. The two frames illustrate the effect of density (area) fluctuations and compositional fluctuations, respectively. The density fluctuations in Fig. 2a manifest themselves in the formation of domains or clusters of correlated lipids of a structure and a density which is different from that of the bulk equilibrium lipid matrix. These domains are dynamic and highly fluctuating entities which are consequences of the cooperative fluctuations of the membrane. The range over which the fluctuations are operative is described by a coherence length which is a measure of the average domain size. The average domain size depends on temperature and attains a maximum at the gel-to-fluid phase transition as shown in Fig. 8a in the case of DMPC, DPPC, and DSPC. For example, for DPPC bilayers, the actual lipid-domain size may become as large as several hundred lipid molecules. Hence, the dynamic membrane heterogeneity leads to membrane organization on the mesoscopic length scale. The average domain size depends on the lipid species in question and increases as the acyl-chain length decreases [20]. The compositional fluctuations in Fig. 2b are a striking characteristic of lipid mixtures in the fluid phase [12,19,55], here revealed by a computer-simulation calculation. Hence, thermodynamic one-phase regions of mixtures may imply a considerable lateral structure of the membrane and a dynamically heterogeneous organization. The coherence length of the compositional fluctuations increases as the temperature approaches the phase boundary. Moreover, the local structure in the fluid phase is more pronounced the more non-ideal the lipid mixture is [19]. As an example of the macroscopic consequences of the dynamic membrane heterogeneity, Fig. 8b shows the computer-simulation data for the specific heat as a function of temperature and acyl-chain length. The specific heat has a sharp peak at the main transition. The peak intensity is decreased for decreasing chain length whereas the intensity in the wings of the transition is increased. Hence, away from the main transition, shorter acyl-chain lengths lead to a higher degree of membrane heterogeneity. Another consequence of the dynamic heterogeneity is an enhancement of the passive membrane permeability [ 18,561. It has been suggested [I81 that the domain interfaces make the membrane leaky and that the thermal anomaly found experimentally [56] in ion permeability at

18 10 40 I

I

I

I

1

kBT’Cp(?”

I

10:

:DMPC

101

in

I

00 0.94

I 0.96

1

0.98

I

1

I

1.02

I

1.04

TITrn

06 1

I

0.96

I

0.98

I

1

1

I

1.02

1.0‘

TITrn

Fig. 8. Computer-simulation data for (a) the average lipid domain size I(T) (in units of numbers of molecules) and (b) the specific heat per molecule, C,,(T), in the transition region for DMPC, DPPC, and DSPC bilayers. The temperature axes are scaled to the value of the pertinent main transition temperature T,. From Ipsen et al. (1990)[20], with permission.

the main transition can be explained by the analogous anomaly in the interfacial area. The effect of cholesterol incorporated into pure lipid bilayers in large amounts, 220%, is to lower the passive permeability. However, small amounts of cholesterol have been found both experimentally and theoretically to increase the passive ion permeability in the neighborhood of the phase transition [57,58]. The computer simulations have revealed that this increased permeability is related to enhanced density fluctuations and a higher degree of dynamic heterogeneity, as illustrated in Fig. 9b [57]. Similarly, molecular agents active at membranes, such as drugs [39,53,59], have been found to alter membrane heterogeneity, and it has in fact been suggested that the mechanism of anesthesia is related to modulation of membrane heterogeneity [59]. Fig. 9c shows the results of computer-simulation calculations illustrating the effect on dynamic membrane heterogeneity of an anesthetic (e.g. halothane). For this system, the specific heat is found theoretically [53], and in accordance with experiments involving halothane [39], cf. section 3.1.3, to shift to lower temperature and to broaden substantially. The broadening reflects increased density fluctuations away from the phase transition.

19

Fig. 9. Dynamic membrane heterogeneity and how it is affected by cholesterol and drugs incorporated into lipid bilayers. The configurations are obtained from computer simulations on a molecular interaction model of DPPC bilayers at a temperature in the fluid phase. Gel regions are denoted as dark areas in the light fluid phase. (a) The pure lipid bilayer; (b) the lipid bilayer in the presence of 9.5% cholesterol; (c) the lipid bilayer in the presence of 28% general anesthetics. In all three cases the temperature is taken to be 3" above the respective midpoint of the phase transition. Adapted from Mouritsen and Jergensen (1992) [9], with permission.

4. Eflects of proteins on membrane heterogeneity The overall dominating effect of incorporation of integral proteins into membranes is a dramatic change in the phase equilibria. Usually, proteins are predominantly soluble in the fluid lipid-bilayer phase and a protein-induced phase separation results in the gel phase[60]. In the fluid phase, which we shall be concerned with here, the presence of the proteins leads to structural changes in the adjacent lipid molecules. The discussion of the physical effects on lipid-membrane structure of protein-lipid interactions has to a large extent been dominated by the experimentalists' preoccupation with spectroscopic order parameters [10,611. The spectroscopic order parameters refer to acyl-chain conformational order (or hydrophobic membrane thickness) and the influence

20

of the proteins is sometimes signalled by the occurrence of additional spectral features. Obviously, such features will be dependent on the intrinsic time scale of the spectroscopic technique used as well as on the diffusional characteristics of the spectroscopic probe. Therefore, it is difficult to infer the nature of the local structure and heterogeneity in membrane-protein systems from spectroscopic experiments alone (see also chapter 6 by Sankaram and Marsh in the present volume). We shall here describe some of the results which have been obtained from computer-simulation calculations on a molecular interaction model, the mattress model [62], of protein-lipid interactions in membranes. The model is built upon the concept of hydrophobic matching [60,62]. According to this concept, a major contribution to protein-lipid interactions is controlled by a hydrophobic matching condition which requires the hydrophobic lipid-bilayer thickness to match the length of the hydrophobic domain of the integral membrane protein. From the results described above on the occurrence of membrane heterogeneity in phospholipid bilayers of different kinds, cf. Fig. 2, it can be anticipated that integral membrane proteins, which, via a hydrophobic matching condition, couple to the membrane lipid acyl-chain order (area density) and/or the membrane composition, are going to influence the degree of heterogeneity. Conversely, a certain degree of membrane heterogeneity will couple to the conformational state of the individual proteins as well as to the aggregational state of an ensemble of proteins.

4.1. Perturbation of lipid acyl-chain structure by integral membrane proteins A microscopic version of the mattress model has been studied by computer simulations [l I] to determine the coherence length, [p(T,xp), for the spatial fluctuations of the lipid order parameter profiles around integral membrane proteins for a fixed distribution of proteins in a large lipid bilayer array in the transition region. The model is studied at low protein concentrations where the overlap between the lipid profiles from neighboring proteins is negligible. The protein is characterized by a hydrophobic length, dp, and a cross-sectional area, np, measured in units of the typical cross-sectional area of a lipid-acyl chain. Schematically, the lipid conformational-order parameter profile, which is related to variations in the local lipid-bilayer thickness, may appear as shown in Fig. 10 in the case of a protein whose hydrophobic length is larger than that of the hydrophobic thickness of the unperturbed lipid bilayer. The profile is exponential, except very close to the main transition. The coherence length, which can be determined from the profile, is found to have a dramatic temperature dependence with a sharp peak at the transition, as illustrated in Fig. 11. The systematics revealed in Figs. 1la-c underline the importance of the degree of hydrophobic matching for the coherence length.

21

12345678

r

Fig. 10. Schematic illustration of the lipid acyl-chain order parameter profile (length profile) near an integral membrane protein of hydrophobic length d p . Here d: is the hydrophobic bilayer thickness and r denotes the distance from the protein in units of lipid acyl-chain diameters.

First of all, the protein-induced disturbance of the lipid bilayer is seen to extend beyond the first few molecular layers over a wide range of temperatures. Secondly, close to the transition the coherence length becomes very large. The overall shape of &(T) is very similar to that of the correlation length of density fluctuations in the pure lipid bilayer. The protein simply couples to the fluctuations in the lipid-bilayer thickness. The data in Fig. 11 show that the effect of the protein on the local structure of the lipid bilayer depends in a detailed manner on the temperature, the size of the protein, as well as the protein hydrophobic length relative to the hydrophobic thickness of the lipidbilayer phases. The long coherence length found in these calculations provides a mechanism for indirect lipid-mediated protein-protein long-range attraction and may hence play an important role in regulating protein segregation as discussed in the following section.

4.2. Lateral distribution of proteins in membranes Several factors control the lateral distribution of proteins in the lipid membrane plane: (i) protein concentration xp; the higher xp is, the higher is the probability for a protein to be next to another protein and hence to form an aggregate; (ii) temperature; the higher the temperature is, the stronger is the relative effect of the entropy which tends to randomize the protein distribution; (iii) proteinlipid interactions and lipid-mediated protein-protein interactions; and (iv) direct protein-protein interactions which may be of long range due to extramembrane moieties. The microscopic version of the mattress mode1[63] has been used to systematically study the lateral protein distribution in model membranes as controlled by factors (ii) and (iii). The results suggest that the formation of protein aggregates in the membrane plane is predominantly controlled by the strength of the direct van der Waals-like protein-lipid interaction. It is found that, whereas the hydrophobic mismatch is of prime importance for determining

22

Fig. 11. Temperature dependence of the coherence length

the phase equilibria, a mismatch may not be the only reason for protein aggregation within each of the individual phases: depending on the strength of the van der Waals-like interaction associated with the direct lipid and protein hydrophobic contact, the proteins may remain dispersed in the fluid phospholipid bilayer, even if the mismatch between the protein and the bilayer thickness is as high as 12A. The type of data on which these conclusions are based is exemplified in Fig 12 which shows a collection of results for a small, rather short protein, dp = 24 A, in a DPPC bilayer membrane at low concentration. The phase diagram in Fig. 12a shows that massive phase separation occurs below the main

23

phase-transition temperature T,. In the phase-separated region, the proteins are dissolved almost exclusively in the fluid-like regions of the bilayer. This is due to the fact that the protein length is closely matched to the fluid bilayer thickness, and solution of proteins in the gel phase would therefore be very costly. However, since the attractive interaction between the lipids and the proteins in this case is assumed to be very low, the solubility of the protein is also low in the fluid phase, and therefore, one might expect a tendency for protein aggregation within the coexistence region at low temperatures where the entropy is low. These expectations are confirmed by the results from the simulations. At T = 295K, for the chosen protein concentration x p = 0.095, the system is in the phase separation region. The aggregate-size distribution function, N ( A ) , in Fig. 12c shows that the number of isolated proteins is low and large protein aggregates are formed. N ( A ) is a measure of the number of protein aggregates consisting of A proteins. The appearance of the protein aggregates almost exclusively in the fluid region of the lattice is demonstrated by the microconfiguration in Fig. 12b. As the temperature is raised toward T = 313K, the system leaves the phase separation region. The proteins are no longer only dissolved in a limited region and the number of isolated proteins increases strongly, as can be seen from Fig. 12d. A number of small protein aggregates remain at the temperature just above the pure lipid transition temperature. At T = 335K, the entropy disordering effect allows only a small number of protein dimers and trimers to be present in the system, as shown in Fig. 12e. The simulation results [63] indicate that, when the direct protein-lipid interaction parameter is sufficiently small, protein aggregates form in the fluid region of the phase diagram just above the phase boundary due to dynamic aggregation induced by the lipid-density fluctuations. This effect is therefore a consequence of the way the proteins couple to the dynamic membrane heterogeneity. Hence by this mechanism, lipid fluctuations can induce dynamic protein aggregation which should be most pronounced close to the phase boundaries. As the strength of the direct protein-lipid interaction is increased, the tendency for formation of protein aggregates via this mechanism is diminished. Larger proteins would, however, induce stronger lipid-mediated attractive protein-protein interactions, cf. Fig. 11d, which in turn would enhance the tendency for aggregation, in particular close to the phase boundaries where the coherence length of the lipid-mediated force is maximal. For larger proteins, a further complication is that the aggregates would be complexes of proteins with lipids trapped in the interstitial regions between several proteins.

24

*

T = 295[K]

I d

0

5

10

15

(d) T = 313[K]

20

30

Fig. 12. (a) Phase diagram in temperature, T , versus protein concentration, x p , for a mixture of DPPC lipids and small proteins with a hydrophobic length of 24A and a very weak hydrophobic protein-lipid interaction. T , = 3 14 K is the transition temperature of the pure lipid bilayer. The labels f and g refer to the fluid and gel lipid phases, and the shaded region f + g indicates the fluidgel coexistence region. The points indicated by asterisks denote the points in the phase diagram investigated by computer simulations on the microscopic version of the mattress model. (b) Snapshot of a typical microconfiguration of the lattice at T = 295 K. The proteins are indicated by dots, and gel and fluid lipid regions are denoted by grey and white areas, respectively. (c-e) Protein cluster sizedistribution, N ( A ) , as a function of temperature for a lipid bilayer matrix with 40 x 40 acyl chains, 80 of which have substituted with small proteins of a hydrophobic length of 24 A. From Sperotto and Mouritsen (1991) [63], with permission.

25 ,

0.4

I

,

l

l

,

1

I

1

5

6

7

8

9

I

DSPC

,,

0.3 1

2

3

4

10 I '

Fig. 13. Example of protein-lipid interface enrichment and physical lipid specificity in a binary lipid mixture. Lipid concentration profiles P(r) for the two lipid species are shown as a function of distance r from a very large integral membrane protein. The data refer to computer simulation on a equimolar binary mixture of DMPC and DSPC at 340 K which is well above the coexistence region. The protein hydrophobic length is taken to be close to the hydrophobic thickness of a fluid DMPC lipid bilayer. From Mouritsen and Sperotto (1 992) [60],with permission.

4.3. Compositional membrane heterogeneity induced by protein-lipid interactions: lipid enrichment and selectivity

The microscopic model calculation described in sections 4.1 and 4.2 above for a one-component lipid membrane incorporated with a very dilute static dispersion of proteins has been extended to membranes with two different lipid species characterized by different acyl-chain lengths [ 121. This extended model was considered with a view to determining to which extent bare physical effects may be responsible for lipid selectivity and lipid specificity of membrane proteins. The basic idea behind the calculation is that, via the hydrophobic matching condition mentioned above, lipid chains of varying length will feel the perturbation of the protein surface to different extents and the lipid species which can most easily adapt to the matching condition will be selected, on a statistical basis, and have an increased probability of being close to the proteinlipid interface. This is an example of interface enrichment. The fact that such a selectivity can be a consequence of the hydrophobic matching condition is demonstrated by the data in Fig. 13 which are derived from computer simulations on the microscopic mattress model, now appropriately extended to account for two different lipid species, DMPC and DSPC. The data in Fig. 13 refer to a very large protein (np M 00) of length dp = 20 and an equimolar lipid mixture at a temperature, T = 340 K, well above the coexistence region of the mixture, i.e. in the fluid lipid phase. The value of dp is chosen to be close to the hydrophobic thickness of fluid DMPC bilayers. Fig. 13 shows the lipid-concentration profiles of DMPC and DSPC as a fkction of distance r from the protein. The protein is seen to select the lipid species (in this case DMPC) which most easily wets the hydrophobic surface of the protein. Conversely, the protein-lipid interface is depleted in the other species. Hence, the protein-lipid

A

26

....... D3PC

.........

XDSPC= 0.5 0

5

10

15

20

25

r

30

Fig. 14. Transient oscillatory behavior of the lipid concentration profiles P(r) for an equimolar mixture of DMPC and DSPC, as a function of the distance r from a very large integral membrane protein of hydrophobic length d p = 26A. The data are obtained from computer simulations on a system with 60 x 60 lipid chains and refer to a temperature of 7' = 325K. From Sperotto and Mouritsen (1992) [12], with permission.

interactions lead to compositional heterogeneities. In other words, the proteinlipid interactions couple to the compositional fluctuations of the binary lipid mixture in the fluid phase, cf. Fig. 2b. The most striking observation made from the model simulations[12] of protein-induced compositional heterogeneity is related to a non-equilibrium transient effect found in the different concentration profiles of the two lipid species as these profiles establish themselves in the course of time. This effect, which may have some important consequences for steady-state membrane organization, refers to a situation where a thermally equilibrated binary lipid mixture is prepared in the fluid phase and then suddenly is made subject to the boundary condition imposed by the presence of the proteins. In response to the presence of the proteins, the mixture has to reorganize itself laterally and decompose locally, as illustrated by the concentration profiles in Fig. 13. This reorganization proceeds via long-range diffusional processes. The interdiffision of the species is, however, limited by the conservation law imposed by the global composition of the mixture. Fig. 14 shows that the mixture, on its way to equilibrium in the presence of proteins, displays a pronounced oscillatory behavior in the concentration profiles. This behavior is dictated by the diffusional processes and the mass-conservation law: after introduction of the proteins to the initially equilibrated mixture, the protein surfaces are, on a time scale corresponding to short-range diffusion, enriched in the appropriate species whose hydrophobic acyl-chain length is compatible with the protein thickness. However, on this time scale the mixture does not have time to fully reorganize and compensate for the excess mass of the enriched species. Therefore, a depletion layer of the same species is formed next to the enrichment layer near the protein. Since the other species have to follow suit by the opposite series of local

21

depletion and enrichment layers, a full oscillatory behavior develops as seen in Fig. 14. As time elapses, the nodes of the oscillations move towards larger values of Y (as 4 according to diffusion) and eventually dampen out, and the equilibrium concentration profiles of Fig. 13 are recovered. The results presented above refer to the case of immobile model proteins, such as proteins bound to specific positions of the membrane, for example via the cytoskeleton, or to proteins which diffuse very slowly relative to the lipids. However, in the case of mobile proteins it can be anticipated from the general nature of the results for static proteins that the structured concentration profiles, cf. Fig. 13, will facilitate a medium-range lipid-mediated indirect proteinprotein attraction which will influence the state of protein aggregation. This observation may have biological relevance for those proteins whose biological activity depends on their aggregational state. It is interesting to note that for a non-equilibrium system, say a proteinlipid membrane driven be external sources of energy which couple to protein conformational changes, the oscillatory profile in Fig. 14 may be dynamically maintained. The mobile proteins in the driven system may organize themselves laterally to fit into the part of the profile which is enriched in the lipid species with the higher affinity for the protein. This picture may straightforwardly be generalized to systems of different proteins with different lipid selectivity. It should be pointed out that the possibility exists that there may be parts of the phase diagram in which the enrichment equilibrium profiles in Fig. 13 develop into a complete wetting phenomenon which implies that the enriched layer grows macroscopically large. Wetting phenomena would have a pronounced effect on the heterogeneous membrane structure.

5. The effect of lipid structure on protein state and functions It is clear that cell surface membranes are organized into domains of distinct composition, structural character and indeed function. Such domains vary in size and have been observed directly by light and electron microscopy [64]. The existence of such domains has also been inferred from fluorescent lifetimes and diffusion properties of molecular probes. These domains have been characterized in terms of lipid composition as, for example, the glycolipid clusters studied by Thompson and Tillack [65]. Yechiel and Edidin [66,67] have investigated domain structure using both fluorescent labelled proteins and lipids and concluded that in human fibroblast plasma membranes, regions rich in protein and regions rich in lipid coexist. The existence of such domains is the net result of the existence of the cytoskeleton, protein-lipid interactions, lipid-lipid interaction, and proteinprotein interactions. However, it is naive to think that any interaction is

28

singularly the driving force for development of these domains. For example, a thermodynamic tendency for proteins to associate (or dissociate) is governed by the difference in the energetics of the protein-protein and lipid-lipid interactions compared to protein-lipid interactions. I: is not necessarily the result of strong protein-protein interactions. If for some reason (e.g., temperature or the presence of a small molecule), the protein-lipid interface becomes stabilized, the proteins will tend to ,dissociate. That is, the systems will tend toward a state which increases protein-lipid interfacial area. Similarly, the energetics associated with either lipid or protein aggregation will be, in part, determined by lipid composition. Unfortunately, there are few experimental studies which focus directly on questions such as those posed in this discussion. One example which suggests strong lipid specificity in both protein-lipid interaction and protein function is the study of protein kinase C by Orr and Walton [68]. These authors demonstrate that the binding of a soluble protein kinase C to small unilamellar vesicles, which are predominantly phosphatidylcholine, is very dependent on PS content. Furthermore, the rate of autophosphorylation is dependent on PS content in an even more complex way, cf. Fig. 15. This latter effect is very dependent on ionic strength and lipid composition. Although data such as these have been interpreted in terms of specific PS lipid-binding sites, they suggest that the forces which drive compositional phase separation in DS-PC systems[69] may be at work in promoting activation of the enzyme as measured by autophosphorylation. In any case, lipid-binding per se is insufficient for activity. A detailed study by Freire and coworkers[70] on the thermal stability of membrane reconstituted yeast cytochrome C oxidase provides additional support for specific lipid-protein interactions. This study, using DSC, differential thermal gel analysis, and enzyme activity, dissected the energetics of the irreversible unfolding into contributions from various subunits. It was found that DMPC and DEPC had different effects on thermal stability. They suggest that stabilization of some membrane proteins may not be entirely a thermodynamic phenomenon, but might be a process modulated by kinetic constraints. This latter speculation raises the question of the role of lipid composition and structure in dynamic processes occurring on and in the membranes. Although direct evidence of the effect of lipid structure on protein state and fbnction is limited, it seems clear that the effect is important. In the next section we will describe in some detail the effect of lipid structural changes on the activation of phospholipase A*, an enzyme whose catalytic efficiency is clearly related to the ‘quality’ of the water-lipid interface and the structural organization to the lipid.

29

100

1

0

20

40

60

[PSI, mol o/o

Fig. 15. Phosphatidylserine dependence of binding and activation of protein kinase C in the presence of vesicles containing 20mol% phosphatidic acid. Protein kinase C (type I) binding (solid circles) and auto-phosphorylation (open circles) were measured in the presence of small unilamellar vesicles composed of 2 mol% DG, 20 mol% PA, 10 mol% dansyl-PE, 0-50 mol% PS, and 18-68 moI% PC. The presence of negatively charged lipids other than PS is required for protein binding in the absence of PS. Adapted from Orr and Newton (1992) [68], with permission.

6. Lipid microheterogeneity and the activation of phospholipase A2 The preceding discussions regarding microheterogeneity of the lipid matrix of biological membranes are based primarily on experimental and theoretical studies of the phase structure of model bilayer lipid systems. The correlation of the results of these experiments and calculations with experimental data of physiological relevance is difficult because of the limited amount of experimental work in this area. A quantitative correlation between the temperature dependence of the rate of passive diffusion of sodium ions across the bilayer and the average size of the domain or cluster of the minor lipid component for single component vesicles has been noted [ 181. The broadest range of correlations between thermodynamic fluctuations or dynamic structural changes of lipid bilayers and protein function exists with phospholipase A2. In fact, many experiments with phospholipase A2 have been designed assuming that it is the shape and size distribution of lipid clusters describing topological heterogeneity in the bilayer that is the primary property of the substrate that relates to the enzymes catalytic function. Phospholipase A2 is a ubiquitous enzyme which hydrolyses the ester linkage at the sn-2 position of a variety of phospholipids (for a general review see Verheij et al.[71]). The enzyme is secreted by a variety of tissues such as the pancreas and is found in both soluble and membrane-bound forms. A major

30

0

2000 3000 4000 Time (seconds1

1000

Fig. 16. Time courses of phospholipase A2 hydrolysis of DPPC large unilamellar vesicles by phospholipase A2 at (a) 39"C, (b) 41"C, and (c) 45°C. From Orr and Newton (1992)[68], with permission.

source of phospholipase A2 is snake venom. This water-soluble protein catalyzes the hydrolysis of monomeric phospholipids, but prefers the substrate to be in aggregated form [72]. Although its activity is intimately related to the physical state of the aggregated substrate, phospholipase A2 will instantaneously hydrolyze the phospholipid substrate in the form of mixed micelles or small unilamellar vesicles. When the bilayer is in the form of large unilamellar or multilamellar vesicles having a large radius of curvature, activity toward zwitterionic substrates is seen primarily in the gel-to-fluid transition region [14,73]. With these types of substrates, the reaction is characterized by a slow lag phase and then a very rapid burst in catalytic activity as shown in Fig. 16. Depending upon the exact temperature, lipid and enzyme species, the burst in activity can be as large a thousand-fold developing over a period of a few seconds following a lag phase of minutes to hours. One model describing the early time course of phospholipase A2 hydrolysis is activation of an inactive phospholipase on the surface of the lipid bilayer where the lag time T is a measure of the rate of that activation. Detailed studies of the enzyme and substrate dependence of this early phase led to a suggested mechanism that involves dimerization of the enzyme on the lipid surface followed by a quasi-irreversible structure change in the enzyme-substrate complex [75]. Most interestingly T achieves a minimum value near T , of the gel-to-liquid transition region [14]. Mouritsen and coworkers [9] have suggested that T-' correlates with the average cluster size as shown in Fig. 17. This result suggests that the events leading to the burst in activity at T are coupled strongly to the magnitude of the thermodynamic fluctuations of the lipid substrate. The temporal sequence of the events for A . piscivorus piscivorus phospholipase A2 were deduced from time-dependent changes in enzyme fluorescence, rate of hydrolysis of substrate, and the emission spectrum of fluorescence probes

31

312

314

316

I TWI

required to reach halfFig. 17. Rate of activation (circles) ( N inverse time, T-’ [min-’ x maximum activity) for hydrolysis of large unilamellar dipalmitoyl phosphatidylcholine vesicles by porcine pancreatic phospholipase A2 in the neighborhood of the gel-to-fluid phase transition of the vesicles at 314 K. Adapted from Biltonen (1990)[5], with permission. T is the time required to reach half-maximum activity of the enzymatic process. Also shown are the results of a model calculation of the average lipid-domain size Z(T) (solid curve) in a dipalmitoyl phosphatidylcholine bilayer (cf. Fig. 8a). Z(T) is in units of number of lipid-acyl chains. Adapted from Ipsen et al. (1990) [20], with permission.

of the lipid structure [76,77]. It was suggested that the first event is binding to the lipid surface followed by slow activation and slow hydrolysis until a critical mole fraction of reaction products (fatty acid and lysolecithin) is reached. At that point a major change in the structure of the lipid, promoting a change in protein structure, occurs. These events are followed by rapid hydrolysis. An important result of these studies is that T exhibits a minimum as a hnction of substrate concentration. The only model of activation consistent with this observation is that the reaction activating the enzyme is quasi-irreversible. More recent studies [78] have demonstrated clearly that the process occurring on the surface following production of a small amount of reaction products is formation of domains of reaction products in a sea of the phospholipid molecules, a possibility that had been suggested by Jain [79]. A pyrene-labelled fatty acid was used to probe the structural state of the bilayer during the course of the hydrolysis reaction. The fluorescence properties of this pyrene label are such that it can be used as a measure of the local concentration of the labelled fatty acid. The experiment was based upon the following hypothesis: Initially the lipid matrix, containing less than 1% of the fatty acid probe, exists in a random distribution throughout the lipid matrix. As the reaction products, fatty acid and lysolecithin, accumulate during the lag phase, little change in the fluorescence

32 01-

0.14c1/;'

\

E/M

['""

Fig. 18. The excimer/monomer fluorescence of pyrene-labelled fatty acid and the amount of product formed as a function of time following phospholipase A2 addition. The maximum in the excimer/monomer ratio occurs just prior to coincident with the maximum velocity. It is at this point that the labelled fatty acid is most concentrated in the lipid matrix.

of the pyrene will be observed. However, as the reaction product concentration in the bilayer reaches a critical value, they separate into compositionally distinct regions. Assuming the labelled fatty acids segregate into domains containing the minor components, their local concentration will increase and the fluorescence signal of the probe will change in an appropriate fashion. As more product is formed, the concentration of the labelled fatty acid in the reaction-productrich regions should decrease and the fluorescence signal should return to its original value. The time dependence of excimer to monomer ratio fluorescence of a pyrene-labelled fatty acid and product formation following addition of phospholipase A2 is shown in Fig. 18. These data are consistent with the hypothesis. The preceding discussion provides two examples from which it appears that the activation of phospholipase A2 is coupled to microheterogeneity of the lipid matrix. One is produced by the gel-to-fluid transition and the other is a phaseseparation phenomenon produced by the production of a second component. Other studies with phospholipase A2 and lipid substrates containing other components are consistent with this picture. The anesthetic dibucaine is able to both activate and inhibit the activation process of the enzyme with no observable effect on the catalytic efficiency of the activated enzyme [80]. The effect of the anesthetic is quantitatively correlated with the anesthetic reducing the phase transition temperature of the lipid model membrane and with the idea that the anesthetic molecules stabilize the interfacial regions between gel and fluid clusters, thereby promoting smaller clusters and, at a constant degree of melting, reducing the magnitude of the thermodynamic fluctuations. This inhibitory effect of dibucaine can be observed at temperatures significantly above T, and may be related to the phase separation phenomenon described above or could be the result of the anesthetic's effects on lipid fluctuations existing above T,. These results suggest that microheterogeneity of the lipid bilayer may play an important role in phospholipase A2 activation on membrane surfaces. The

33

temperature dependence of the lag period indicates that thermal fluctuations can be important and the effect of reaction products suggests that phase separation is important. Whether or not microheterogeneity as described here is relevant in biological situations is not, however, clear. Nevertheless, it is quite clear that domains can form from either proteins, lipid, or both in real biological membranes. These domain structures can greatly influence diffusion [81-831 within their plane of the bilayer and could in fact have great effects on chemical reactions and protein interactions within the bilayer as well. Thompson et al. [8] have recently described a relatively simple analysis of the effect of domain structure on bimolecular reactions in the plane of bilayer in which the reactants are confined to one phase or the other. Such phenomena may be important but many more experiments both with model systems and with the more experimentally difficult biological membranes must be pursued

7. Eflects of drugs on protein-lipid interactions and membrane heterogeneity The presence of foreign molecules such as anesthetics are known to alter the phase transition characteristics of lipid bilayers [ 171 or to change the ‘fluidity’ of more complex membrane systems [51]. These effects manifest themselves as changes in the melting temperature (generally, but not always a reduction in T,) and a broadening of the phase transition. This latter effect has been suggested to be the result of thermodynamic stabilization of the interface between gel and fluid domains [5,53]. Monte Carlo calculations have shown that in such a case the foreign molecules can accumulate in these dynamic interfacial regions [53]. This phenomenon has not been investigated in any detail experimentally, but the temperature dependence of the lipid-water partition coefficient of different local anesthetics, such as cocaine derivatives [84], as well as insecticides, such as lindane[85], exhibits a maximum near T,, a result consistent with this possibility. The gel and fluid cluster interdomain regions are prototypical of any type of lipid mismatch region, whether it be an interface between structurally distinct lipids, chemically distinct lipids, or protein-lipid interfaces. Accumulation of a foreign molecules at any of these interfaces implies stabilization of it. An example of the first type is induced broadening of the phase transition without significant change in A H o . The inhibitory effect of the anesthetic dibucaine on phospholipase A2 activation at temperatures well above T , [80] could be the result of stabilization of the boundaries between phospholipid and reaction product domains, and the accumulation of spin-label probes at protein surfaces adjacent to the lipid matrix is an example of the last type.

34

Several studies have shown that spin labels [86-881 which are used as probes of the state of the lipid matrix and protein-lipid interactions tend to accumulate at the protein-lipid interfacial region. It should be noted that the accumulation as measured by immobilization of the spin system is not a static event. Rather, relatively rapid exchange of the label between the interfacial region and the bulk lipid phase occurs. The accumulation of foreign molecules at these structurally mismatched regions is not really surprising since they are probably regions of high energy (reduced van der Waals contacts) and low entropy (limited rotational freedom). Thus small foreign molecules, at least, can be accommodated in these mismatched regions relatively easily. The observation that anesthetics appear to displace spin-label probes from protein-lipid interfaces is a result of competitive binding to the structurally mismatched region. It does not necessarily imply that protein-lipid interfaces are the site of anesthetic action, but rather the anesthetic energetically prefers such regions relative to the bulk lipid phase. Nevertheless, these studies support the important proposition that assuming anesthetics do not assert their influence by binding to specific protein sites, they most likely do it by binding and thus stabilizing interfacial regions in the lipid matrix. A mechanism of how this might influence protein function is still open to question.

References [ 11 Singer, S.J. and Nicolson, G.L. (1 972) The fluid mosaic model of the structure of cell membranes. Science 173, 72&73 1. [2] Gennis, R.B. (1989) Biomembranes. Molecular Structure and Function. Springer, London. [3] Bloom, M., Evans, E. and Mouritsen, O.G. (1991) Physical properties of the fluid-bilayer component of cell membranes: a perspective. Q. Rev. Biophys. 24, 293-397. [4] Cevc, G., and March, D. (1987) Phospholipid Bilayers. Physical Principles and Models. WileyInterscience, New York. [5] Biltonen, R.L. (1 990) A statistical-thermodynamic view of cooperative structural changes in phospholipid bilayer membranes: their potential role in biological function. J. Chem. Thermodyn. 22, 1-19. [6] Mouritsen, O.G. (1991) Theoretical models of of phospholipid phase transitions. Chem. Phys. Lipids 57, 178-194. [7] Edidin, M. (1990) Molecular associations and lipid domains. Current Topics in Membranes and Transport 36, 81-96. [8] Thompson, T.E., Sankaram, M.B. and Biltonen, R.L. (1992) Biological membrane domains: functional significance. Comm. Mol. Cell. Biophys. 8, 1-15. [9] Mouritsen, O.G. and Jergensen, K. (1992) Dynamic lipid-bilayer heterogeneity: a mesoscopic vehicle for membrane function? BioEssays 14, 129-136. [lo] Marsh, D. (1985) ESR spin label studies of lipid-protein interactions. In: Progress in ProteinLipid Interactions, Val. 1 (Watts, A. and De Pont, J.J.H.H.M., Eds.), pp. 143-172, Elsevier, Amsterdam.

35 [ I l l Sperotto, M.M. and Mouritsen, O.G. (1991) Monte Carlo simulation studies of lipid order parameter profiles near integral membrane proteins. Biophys. J. 59, 261-270. [I21 Sperotto, M.M. and Mouritsen, O.G. (1992) Lipid enrichment and selectivity of integral membrane proteins in two-component lipid bilayers. Biochim. Biophys. Acta (submitted). [13] Yuan,Q. and Biltonen, R.L. (1991) Evidence for compositional phase separation in lipid vesicles during phospholipase A2 catalyzed hydrolysis. Biophys. J. 59, 509a. [14] Menashe, M., Romero, G., Biltonen, R.L. and Lichtenberg, D. (1986) Hydrolysis of dipalmitoyl phosphatidylcholine small unilamellar vesicles by porcine pancreatic phospholipase A2. J. Biol. Chem. 261, 5328-5333. [15] Ruggiero, A. and Hudson, B. (1989) Critical density fluctuations in lipid bilayers detected by fluorescence lifetime heterogeneity. Biophys. J. 55, 11 11-1 124. [16] Freire, E., and Biltonen, R.L. (1978) Estimation of molecular averages and equilibrium fluctuations in lipid bilayer systems from excess heat capacity function. Biochim. Biophys. Acta 514, 54-68. [17] Mouritsen, O.G. (1990) Computer simulation of cooperative phenomena in lipid membranes. In: Molecular Description of Biological Membrane Componentsby Computer Aided Conformational Analysis, Vol. 1 (Brasseur, R., Ed.), pp. 3-83, CRC Press, Boca Raton, Florida. [18] Cruzeiro-Hansson, L. and Mouritsen, O.G. (1988) Passive ion permeability of lipid membranes modelled via lipid-domain interfacial area. Biochim. Biophys. Acta. 944, 63-72. Ipsen, J.H. and Zuckermann, M.J. (1993) Phase [19] Jergensen, K., Sperotto, M.M., Mouritsen, 0.6, equilibria and local structure in binary lipid bilayers. Biochim. Biophys. Acta (submitted). [20] Ipsen, J.H., Jsrgensen, K. and Mouritsen, O.G. (1990) Density fluctuations in saturated phospholipid bilayer increase as the acyl-chain length decreases. Biophys. J. 58, 1099-1 107. [21] Koynona, R.D., Tenchov, B.G., Quinn, P.J. and Laggner, P. (1988) Structure and phase behavior of hydrated mixtures of L-palmitoylphosphatidylcholineand palmitic acid. Correlations between structural rearrangements, specific volume changes and endothermic events. Chem. Phys. Lipids 48,204-214. [22] Menashe, M., Lichtenberg, D. and Biltonen, R.L. (1984) Characterization of the pre-transition in unilamellar dipalmitoylphosphatidylcholinevesicles. Lipids 19, 293-301. [23] Seddon, J.M., Cevc, G., and Marsh, D. (1983) Calorimetric studies of the gel-fluid and lamellar-inverted hexagonal phase transition of dialkyl- and diacylphosphatidylethanolamines. Biochemistry 22, 1280-1289. [24] Albon, N. and Sturtevant, J.M. (1978) Nature of the gel to liquid crystal transition of synthetic phosphatidylcholines. Proc. Natl. Acad. Sci. USA. 75, 2258-2260. [25] Lewis, R.N.A.H., Mak, N., and McElhaney, R.N. (1987) A differential scanning calorimetric study of the thermotropic phase behavior of model membranes composed of phosphatidylcholines containing linear saturated fatty acyl chains. Biochemistry 26, 61 184126. [26] Wilkinson, D.A. and Nagle, J.F. (1979) Dilatometric study of binary mixtures of phosphatidylcholines. Biochemistry 18, 4244-4249. [27] Schmidt, C.F., Barenholz, Y., Huang, C. and Thompson, T.E. (1978) Monolayer coupling in sphingomyelin bilayer systems. Nature 27 1, 775-777. [28] Sillerud, L.O. and Barnett, R.E. (1982) Lack of transbilayer coupling in phase transitions of phosphatidylcholine vesicles. Biochemistry 2 I , 1759-1 776. [29] Zhang, F. and Rowe, E.S. (1992) Titration calorimetric and differential scanning calorimetric studies of the interactions of n-butanol with several phases of dipalmitoylphosphatidylcholine. Biochemistry 31,2005-201 1. [30] Wang, Z.-Q., Lin, H.-N., and Huang, C.-H. (1990) Differential scanning calorimetric study of a homologous series of fully hydrated saturated mixed-chain C(X):C(X+6) phosphatidylcholines. Biochemistry 29, 7072-7076.

36 [31] Bangham, A.D., Standish, M.M. and Watkins, J.C. (1965) Diffusion of univalent ions across the lamellae of swollen phospholipids. J. Mol. Biol. 13, 238-252. [32] Huang, C.-H. (1969) Studies on phosphatidylcholine vesicles. Formation and physical characteristics. Biochemistry 8, 344-352 [33] Suurkuusk, J., Lentz, B.R., Barenholz, Y., Biltonen, R.L. and Thompson, T.E. (1976) A calorimetric and fluorescent probe study of the gel-liquid crystalline phase transition in small, single-lamellar dipalmitoylphosphatidylcholinevesicles. Biochemistry 15, 1393-1401. [34] Wong, M., Anthony, F.H., Tillack, T.W., and Thompson, T.E. (1982) Fusion of dipalmitoylphosphatidylcholine vesicles at 4°C. Biochemistry 21, 41264132. [35] Barenholtz, Y., Amselem, S., and Lichtenberg, D. (1979) A new method for preparation of phospholipid vesicles (liposomes) - french press. FEBS Letts. 99, 210-214. [36] Biltonen, R.L. and Freire, E. (1980) Statistical thermodynamic analysis of the excess heat capacity functions of macromolecular systems. Pure and Applied Chemistry 52, 41 9-43 1. [37] Sugar, I., Mitchard, N. and Biltonen, R.L. (1992) Two-state model of the gel-liquid crystalline transition of small unilamellar vesicles of dipalmitoylphosphatidylcholine. Biophys. J. 61 a, A238.20. [38] Marsh, D., Watts, A. and Knowles, P.F. (1977) Cooperativity of the phase transition in singleand multibilayer lipid vesicles. Biochim. Biophys. Acta 465, 500-5 14. [39] Mountcastle, D.B., Biltonen, R.L. Halsey, M.J. (1978) Effect of anesthetics and pressure on the thermotropic behavior of multilamellar dipalmitoylphosphatidylcholine liposomes. Proc. Natl. Acad. Sci. U.S.A. 75, 4906-4910. [40] Yager, F! and Peticolas, W.L. (1982) The kinetics of the main phase transition of aqueous dispersions of phospholipids induced by pressure jump and monitored by Raman spectroscopy. Biochim. Biophys. Acta 688, 775-785. [41] Caffrey, M., Hogan, J. and Mencke, A. (1991) Kinetics of the barotropic ripple (Pgt)llamellar liquid crystal (La) phase transition in fully hydrated dimyristoylphosphatidylcholine (DMPC) monitored by time-resolved X-ray diffraction. Biophys. J. 60, 456-466. [42] van Osdol, W.W., Biltonen, R.L. and Johnson, M.L. (1989) Measuring the kinetics of membrane phase transitions. J. Biochem. Biophys. Meth. 20, 1-46. [43] Mayorga, O.L., van Osdol, W.W., Lacomba, J, L.,and Freire, E. (1988) Frequency spectrum of enthalpy fluctuations associated with macromolecular transitions. Proc. Natl. Acad. Sci. U S A . 85, 95149518. [44] van Osdol, W.W., Johnson, M.L., Ye, Q. and Biltonen, R.L. (1991) Relaxation Dynamics of the gel-liquid crystalline transition of phosphatidylcholine bilayers: the effects of chain length and vesicle size. Biophys. J. 59, 775-785. [45] Ye, Q. (1992) Phase transition kinetics of multicornponent lipid membranes. Ph.D. Dissertation, Biophysics Program, University of Virginia, Charlottesville, VA. [46] Yang, C.F! and Nagle, J.F. (1988) Phase transformations in lipids follow classical kinetics with small fractional dimensionalities. Phys. Rev. A. 37, 3993-4000. [47] Ye, Q., van Osdol, W.W. and Biltonen, R.L. (1991) Gel-liquid crystalline transition of some multilamellar lipid bilayers follows classical kinetics with a fractional dimensionality of approximately two. Biophys. J. 60, 1002-1007. [48] Tsong, T.Y., Greenberg, M. and Kanehisa, M. (1977) Anesthetic action on membrane lipids. Biochemistry 16, 3115-3121. [49] Jain, M.K. and Wu, N.-Y. (1977) Effect of small molecules on the phase transition in lipid bilayers of dipalmitoyl lecithin. J. Mem. Biol. 34, 157-201. [50] Trudell, J.R., Payan, D.G., Chin, J.H. and Cohen, E.N. (1975) The antagonistic effect of an inhalation anesthetic and high pressure on the phase diagram of mixed dipalmitoyl4imyristoylphosphatidylcholine bilayers. Proc. Natl. Acad. Sci. U.S.A. 72, 210-213.

37 [51] Trudell, J.R., Payan, D.G., Chin, J.H. and Cohen, E.N. (1974) Pressure-induced elevation ofphase transition temperature in dipalmitoylphosphatidylcholine bilayers: An electron spin resonance measurement of the enthalpy of phase transition. Biochim. Biophys. Acta 373,436443. [52] Suezake, Y., Tatara, T., Kaminoh, Y., Kamaya, H. and Ueda, I. (1990) A solid-solution theory of anesthetic interaction with lipid membranes: temperature span of the main transition. Biochim. Biophys. Acta 1029, 143-148. [53] Jsrgensen, K., Ipsen, J.H., Mouritsen, O.G., Bennett, D. and Zuckermann, M.J. (1991) The effects of density fluctuations on the partitioning of foreign molecules into lipid bilayers: application to anesthetics and insecticides. Biochim. Biophys. Acta 1067,24 1-253. [54] van Osdol, W., Ye, Q., Johnson, M.L., and Biltonen, R.L. (1992) The effects of the anesthetic dibucaine on the kinetics of the gel-liquid crystalline transition of dipalmitoylphosphatidylcholine multilamellar vesicles. Biophys. J. (in press). 1551 Freire, E. and Snyder, B. (1980) Monte Carlo studies of the lateral organization of molecules in two-component lipid bilayers. Biochim. Biophys. Acta 600, 643-654. [56] Papahadjopoulos, D., Jacobsen, K., Nir, S. and Isac, T. (1973) Phase transitions in phospholipid vesicles. Fluorescence polarization and permeability measurements concerning the effect of temperature and cholesterol. Biochim. Biophys. Acta 31 1, 33fk348. [57] Cruzeiro-Hansson, L., Ipsen, J.H. and Mouritsen, O.G. (1989) Intrinsic molecules in lipid membranes change the lipid-domain interfacial area: cholesterol at domain boundaries. Biochim. Biophys. Acta 979, 166-1 76. [58] Corvera, E., Mouritsen, O.G., Singer, M.A., and Zuckermann, M.J. (1992) The permeability and the effect of acyl chain length for phospholipid bilayers containing cholesterol: theory and experiment. Biochim. Biophys. Acta I 107, 261-270. [59] Jsrgensen, K., Ipsen, J.H., Mouritsen, O.G., and M.J. Zuckermann (1993) The effect of anesthetics on the dynamic heterogeneity of lipid membranes. Chem. Phys. Lipids (in press). [60] Mountsen, O.G. and Sperotto, M.M. (1 992) Thermodynamics of lipid-protein interactions in lipid membranes: the hydrophobic matching condition. In: Thermodynamics of Cell Surface Receptors (Jackson, M., Ed.), pp. 127-181, CRC Press, Boca Raton, FL. [61] Bloom, M. and Smith, I.C.P. (1985) Manifestations of lipid-protein interactions in deuterium NMR. In: Progress in Protein-Lipid Interactions, Vol. 1 (Watts, A. and De Pont, J.J.H.H.M., Eds.), pp. 61-88, Elsevier, Amsterdam. [62] Mouritsen, O.G. and Bloom, M. (1984) Mattress model of lipid-protein interactions in membranes. Biophys. J. 46, 141-153. [63] Sperotto, M.M. and Mouritsen, O.G. (1991) Mean-field and Monte Carlo studies of the lateral distribution of proteins in membranes. Eur. Biophys. J. 19, 157-168. [64] Rodriguez-Boulan, E., and Nelson, W.J. (1989) Morphogenesis of the polarized epithelial cell phenotype. Science 245, 718-725. [65] Thompson, T.E. and Tillack, T.W. (1985) Organization of glycophingolipids in bilayers and plasma membranes of mammalian cells. Annu. Rev. Biophys. Biophys. Chem. 14, 361-386. [66] Edidin, M. and Stroynowski, I. (1991) Differences between the lateral organization of conventional and inositol phospholipid-anchored membrane proteins. A further definition of micrometer scale membrane domains. J. Cell Biol. 112, 1 143-1 150. [67] Yechiel, E., and Edidin, M. (1987) Micrometer-scale domains in fibroblast plasma membranes. J. Cell Biol. 105, 755-760. [68] Orr, J.W. and Newton, A.C. (1992) Interaction of protein kinase C with phosphatidylserine. 2. Specificity and regulation. Biochemistry 3 1, 46674673. [69] Silvius, J.R. and Gagne, J. (1984) Calcium-induced fusion and lateral phase separations in phosphatidylcholine-phosphatidylserine vesicles. Correlation by calorimetric and fusion measurements. Biochemistry 23, 3241-3247.

38 [70] Morin, RE., Diggs, D., and Freire, E. (1990) Thermal stability of membrane-reconstituted yeast cytochrome C oxidase. Biochemistry 29, 781-788. [71] Verheij, J.M., Slotboom, A.J., and de Haas, G.H. (1981) Structure and function of phospholipase A2. Rev. Physiol. Biochem. Pharmacol. 91, 91-203. [72] Pieterson, W.A., Volwerk, J.J. and de Haas, G.H. (1974) Zymogen-catalyzed hydrolysis of monomeric substrate and the presence of a recognition site for lipid-water interfaces in phospholipase A2. Biochemistry 13, 1455-1460. [73] Apitz-Castro, R., Jain, M.K. and de Haas, G.H. (1982) Origin of the latency phase during the action of phospholipase A2 on unmodified phosphatidylcholine vesicles. Biochim. Biophys. Acta 688, 349-356. [74] Biltonen, R.L., Lathrop, B., Heimburg, T. and Bell, J.D. (1990) Aspects of the activation of phospholipase A2 on lipid bilayer surfaces. In: Biochemistry, Molecular Biology, and Physiology of Phospholipase A2 and its Regulatory Factors. (Mukhejee, A.B., Ed.), pp. 85-103, Plenum, New York. [75] Romero, G., Thompson, K. and Biltonen, R.L. (1987) The activation of porcine pancreatic phospholipase A2 by dipalmitoylphosphatidylcholine large unilamellar vesicles: analysis of the state of aggregation of the activated enzyme, J. Biol. Chem. 263, 11808-11813. [76] Bell, J.D. and Biltonen, R.L. (1989) The temporal Sequence of events in the activation of phospholipase A2 by lipid vesicles: studies with the monomeric enzyme from Agkisfrodon piscivorus piscivorus. J. Biol. Chem. 264, 12194-12200. 1771 Bell, J.D. and Biltonen, R.L. (1992) Molecular details of the activation of soluble phospholipase A2 by lipid bilayer substrates: comparison of computer simulations with experimental results. J. Biol. Chem. (in press). [78] Burack, W.R., Yuan, Q. and Biltonen, R.L. (1992) Compositional phase separation in lipid bilayers and activation of phospholipase A2 (in preparation). [79] Jain, M.K., Yu, B.-Z. and Kozubek, A. (1989) Binding of phospholipase A2 to zwitterionic bilayers is promoted by lateral segregation of anionic amphiphiles. Biochim. Biophys. Acta 980, 23-32. [SO] Lathrop, B.L., Stokes, D. and Biltonen, R.L. (1992) The effect of anesthetic dibucaine on the activation of PLA2 in the gel-liquid crystalline transition region of lipid bilayer substrates (in preparation). [81] Saxton, M.J. (1990) The membrane skeleton of erythrocytes. A percolation model. Biophys. J. 57, 1167-1 177. [82] Vaz, W.C.L., Melo, E.C.C. and Thompson, T.E. (1990) Fluid phase connectivity in an isomorphous two component, two phase phosphatidylcholine bilayer. Biophys. J. 58, 237-275. [83] Vaz, W.L.C., Melo, E.C.C. and Thompson, T.E. (1989) Translational diffusion and fluid domain connectivity in a two-component, two-phase phospholipid bilayer. Biophys. J. 56, 869-876. [84] Singer, M.A. and Jain, M.K. (1980) Interaction of four local anesthetics with phospholipid bilayer membranes: permeability effects and possible mechanisms. Can. J. Biochem. 58, 8 15-821. [85] Antunes-Madeira, M.C. and Madeira, VM.C. (1985) Partitioning of lindane in synthetic and native membranes. Biochim. Biophys. Acta 820, 165-172. [86] Fraser, D.M., Louro, S.R.W., Horvith, L.I., Miller, K.W. and Watts, A. (1990) A study of the effect of general anesthetics on lipid-protein interactions in acetylcholine receptor enriched membranes from Torpedo nobiliunu using nitroxide spin-labels. Biochemistry 29, 2664-2669. [87] Arias, H.R., Sankaram, M.B., Marsh, D. and Barrantes, F.J. (1990) Effect of local anesthetics on steroid-nicotinic acetylcholine interactions in native membranes of Torpedo mumotutu electsic organ. Biochim. Biophys. Acta 1027,287-294.

39

[88] Horvath, Arias, H.R., Hankovszky, H.O., Hideg, K., Barrantes, F.J. and Marsh, D. (1990) Association of spin-labeled local anesthetics at the hydrophobic surface of acetylcholine receptor in native membranes from Torpedo mannotutu. Biochemistry 29, 8707-871 3.