Upsala J Med Sci 82: 167-181, 1977 The Plasma Membrane Consists of Two Po la r-No n po la r-Po lar Leaf lets Interpretation of Some Data, Mainly from the Erythrocyte ERIK CERVEN From the Institute of Medical and Physiological Chemistry, Biomedical Center, University of Uppsala, Uppsala, Sweden ABSTRACT The implications of a double polar-nonpolar-polar leaflet construction of the plasma membrane are investigated. Experimental data from transmission electron microscopic and enzymologic characterization of plasma membranes are advantageously interpreted in these terms compared to interpretation in terms of lipid bilayer. X-ray diffraction and electron spin resonance studies do not differentiate between the present and previous models for the structure of plasma membranes but electron spin resonance data that fail to indicate a statistical distribution of spin labels also fail to support the fluid mosaic model for cell mem- branes. Results from experiments involving vectorial di- gestion and labelling of plasma membranes as well as freeze fracture electron microscopic data are compatible with the present model. The molecular composition of the human erythrocyte membrane is investigated whereby the band 111 protein and glycophorin are suggested to be the structural proteins of the outer leaflet and the spectrins those of the inner leaflet. INTRODUCTION The erythrocyte membrane is the most thoroughly investigated of biological membranes. Several approaches have been used in elucidating its structure, ranging from early estimations of lipid content (47) t o recent X-ray photo-electron spectroscopy (69). The vast amount of information available on particularly the human erythrocyte membrane provides a basis for testing the recently developed concept that the plasma membrane con- sists of two polar-nonpolar-polar leaflets (15). This concept was originally developed from considera- tions of structural and functional similarities and dissimilarities between various biological mem- branes. In particular it required knowledge about the nucleotide metabolism associated with intact cells (3, 4, 5, 6 , 29, 90) and elaboration (16) of the idea (58) that actin-containing filaments attached to the plasma membrane (52, 74) are analogous to the coupling factors (cf. 58) of mitochondria and chloroplasts. It is the aim of this work t o examine data that are difficult to reconcile with the fluid mosaic model for the structure of cell membranes (95) and that tend to support the concept that such membranes consist of two parallel polar-nonpo- lar-polar leaflets. Primarily, data dealing with the human erythrocyte membrane will be discussed. Some previous models f o r the structure of the human erythrocyte membrane The lipid bilayer concept for the structure of biolog- ical membranes (see ref. 104 for a short review on membrane models) (Fig. 1) was originally de- veloped from considerations of the surface area of intact erythrocytes compared with that occupied by the lipids extracted from erythrocytes when spread as a monolayer on an aqueous surface (47). Accord- PROTEIN MSPHOLIRO Fig. 1 . Illustration of the fluid mosaic model for the structure of cell membranes according to which intra- membranous particles revealed by freeze-fracturing rep- resent globular proteins surrounded by unperturbed lipid bilayer (ref. 95). 12-772856 Upsala J M e d S c i 8 2 168 E . Cerve'n ELECTRON BEAM ELECTRON OPTICAL SYSTEM 4 IMAGE c r / r l trilaminar strkture Fig. 2. Formation of a trilaminar image seen in the trans- mission electron microscope. ing to this model (47, 88, 94, 9 9 , a trilaminar ap- pearance of the erythrocyte membrane under the transmission electron microscope is attributed to lack of affinity of stains to the nonpolar residues be- tween polar residues of the bilayer (88). Fracture planes generated by the freeze fracture procedure for electron microscopic preparation are attributed to lack of cooperativity of van der Waals forces between methyl groups of contrapositioned lipids in the bilayer (25). Proteins spanning across two leaflets of the erythrocyte membrane (cf. 95) are claimed to be detected by the occurrence of intra- membranous particles after freeze fracturing. The stability of pure lipid bilayers ( I 04) and the apparent lack of interdependence of protein and lipid conformation in biological membranes as in- terpreted from measurements of circular dichroism (94) have contributed significantly to the develop- ment of this concept (94, 95). However, lipid in bilayers is presumably equilibrated between mem- branes of various cellular systems by way of a small aqueous pool (104). It is therefore difficult to ex- plain the fact that membranes from different organ- elles are characterized by different lipid composi- tion (23, 70) and by characteristic enzymatic activi- ties and proteins. This is because the fluid mosaic model for cell membranes implies that both lipid and protein move about, lacking close and specific contact in the planes of membranes (39, 95) that may in addition be capable of fusion (cf. 8). With this model it is also difficult to imagine any genetic control mechanism responsible for species-specific lipid composition of biological membranes (cf. 70, 120). Very complex mechanisms including re- strained mobility of integral proteins imposed by specifically associated fibrous proteins have been Upsala J M e d Sci 82 developed (74) in order to overcome some of these difficulties. The unit membrane hypothesis (88), according to which all biological membranes are represented by lipid bilayers with various amounts of loosely and peripherally associated protein, was partly based on electron microscopic observations of KMn0,- or Os0,-stained and dehydrated membranes. How- ever, the thickness of the erythrocyte membrane, as determined by this methodology, is uncertain be- cause of extensive losses of both protein and lipid when osmium and dehydration are used for electron microscopic preparation (67). In dehydrated prep- arations, cross-linked with glutaraldehyde and stained with low concentrations of osmium, thereby avoiding extensive losses of protein, no trilaminar structure can be seen and the membrane thickness is about 160 A (67). Great variations of membrane thickness as measured by independent techniques (see Table I) indicate that the degree of hydration and the amounts of associated water-soluble pro- teins are important factors. Table 1. The thickness of the (human) erythrocyte membrane determined by various methods For comparison, the thickness of a lipid bilayer may be 45-50 A determined by X-ray diffraction (27, 104) and 45f5 8, determined by electron microscopy (72, 88). See ref. 71 for a more exhaustive constellation of data Method Thickness Ref. Birefringence 5 0 0 0 8 , 71 Leptoscope (rabbit, rat steer) 160-220 A 114 Argon ion etching combined with iron signals determined by electron spectroscopy 1 000 A 69 Energy transfer 65-100A 82 Electron microscopy Glutaraldehyde and high con- Excluded volume 600 A 44 centrations of osmium tetroxide 70 A 67 centrations of osmiurn- tetroxide 160 A 67 Embedding in water-soluble polymers and staining with uranylacetate and lead citrate 70-110A 52, 81 The nonpolar core estimated from lipid content assuming even distribution in one Glutaraldehyde and low con- surface layer 27 A 33 Double polar-nonpolar-polar leaflet construction of the plasma membrane 169 0 POLAR RESIDUES NONPOLAR RESIDUES 0 AQUEOUS PHASE PROTEIN @ OIL P H A Y PROTEIN 0 K - H B l X \ ( GLYCOPFlOTElN F i g . 3 . The structure of a complete biological membrane according to ref. 15. The indexed capitals P (polar), N (nonpolar) and Z (intramembranous) serve to denote the location of different components. For example, actin is a P, component of the plasma membrane, NDP-kinase and adenylate-kinase are presumably I, components of the mitochondrial and the plasma membranes (ref. 15). Definition of a complete biological membrane The nucleotide metabolism associated with the surface of many intact cells is similar to that as- sociated with intact mitochondria (4, 6 , 15, 29). Furthermore, actin-containing filaments, the sub- unit structure of which is reminiscent of that of the coupling factors of the inner mitochondrial mem- brane ( 5 8 ) , are closely associated with the inner leaflet of the plasma membrane of many cells (52, 74). The similarity of the two mitochondrial mem- branes and the inner and outer leaflets of the plasma membrane is further emphasized by the asymmetric distribution of carbohydrates in both instances (10, 74, 75, 102). In addition, covalently bound DNA is included as a marker for the inner leaflets in both these in- stances (cf. 15, 45, 50). By these criteria, the two mitochondrial membranes and the two leaflets of the plasma membrane may be designated as com- plete biological membranes (15, see Fig. 3). Since the outer leaflet of the plasma membrane contains identity markers and receptors of importance for cellular interactions in multicellular systems, there is reason to regard one complete biological mem- brane as a functional unit composed of one leaflet carrying endogenous information (DNA) and/or energy transducing activity and one leaflet mediat- ing informational exchange with the surrounding medium (exogenous information). These functions may be more or less developed in particular in- stances. For example, endogenous information may be more important than the informational exchange function (exogenous information) in the complete biological membrane represented by the inner nuclear membrane + the outer nuclear membrane combined with the endoplasmatic membranes of animal cells. Structure of a single leaflet Nonpolar residues of amino acids and lipids tend to be shielded from hydrogen binding water-soluble compounds (104). According t o this fundamental thermodynamic principle (the hydrophobic effect), a monolayer of phospholipids in contraposition to nonpolar amino acid residues of a peptide in a P-pleated sheet-configuration is stable, as il- lustrated in Fig. 4 A . This structure is maximally stabilized in an aqueous environment if every sec- ond amino acid residue of the peptide is polar. However, clusters of 15-17 nonpolar amino acids are stabilized in the form of an a-helix embedded in the nonpolar core (see Fig. 4B) because of the hydrophobic effect (104). The peptide may bind another layer of phospholipids after it has traversed the membrane, as illustrated in Fig. 4 C . Similar arrangements of phospholipid, protein and other biological compounds may be imagined to exist throughout the three-dimensional membrane. This structure would be extremely stable owing to the contribution of cooperative nonpolar bonds, hydrogen bonds and salt bridges, as illustrated in F+6RIOCIPID A 0 c F i g . 4 . Transsection of a monolayer of phospholipid and a peptide in ,&pleated sheet configuration (A and C) and a peptide segment of 15-17 exclusively nonpolar amino acid residues stabilized in a-helix form in the nonpolar core ( B ) . Upsala J Mrd Sci 82 170 E . Cerve‘n phospholipid 00 corbohydrate glycolipid A,A protein - cholesterol Q o charges F i g . 5 . Illustration of the possible contribution of lipid, protein and sugar to the structure of a single leaflet. Broad lines denote compounds located close to and thin lines far from the reader. This structure is composed of a nonpolar core ( I ) , two planes of hydrogen bonding ( 2 ) and two planes of salt bridges (3). The binding of aqueous phase and oil phase proteins to this type of structure is illustrated in Fig. 3. Fig. 5. The possible contribution of different com- pounds t o these three types of binding is listed in Table 11. The association of water-soluble and in- tegral proteins as well as pore formation in one leaflet has been described previously (15). Sugar residues are assumed (cf. 15) to profoundly influence a membrane on account of their nonpolar surfaces and their property to structunze water (cf. 35) thus presumably forming part of the planes of hydrogen bonding. The presence of sugar in bio- logical membranes presumably make them more permeable to polar solutes by influencing the hydrophobic effect. This is illustrated tentatively by the decreased surface tension in lecithin monolayers brought about by sialic acid in combi- nation with metal ions (26). Antigens and receptors residing in glycosyl residues would thus be shielded from detection until the membrane structure is dis- turbed by proteolytic or lipolytic cleavage. Thus treatment of fat cells with phospholipase C in- creases the insulin-binding capacity (24). Some experimental data in support of this type of structure (see Fig. 5) may be accrued from the lit- erature. For example the low extractability of camel erythrocyte phospholipid with ether/methanol ( 1 10) in contrast to that of human erythrocyte phos- pholipid of similar composition and of similar amount/surface area (30, 61) may suggest that pro- tein is important in binding the phospholipids to the camel erythrocyte membrane which is charac- terized by a high protein to lipid ratio. If the thermo- dynamic principles denoted the hydrophobic effect are strictly applied to glycophorin (the sequenced main sialoglycoprotein of the human erythrocyte (62)) then the polar thr-74, ser-92 and under some conditions weakly acidic tyr-93 amino acid residues are preferably excluded from the nonpolar segment thought t o extend through the lipid bilayer (cf. 104). The remaining segment, composed of 17 nonpolar amino acid residues and stabilized in an a-helix form due to the nonpolar environment (104) would Table 11. Examples of atoms and groups in different biological compounds possibly contributing to the nonpolar core and the planes of hydrogen bonding and of salt-bridges in a biological membrane Compound Nonpolar core bonding bridges Plane of hydrogen Plane of salt- Cholesterol Phospholipid Sphingolipid Glycolipid and carbohydrate Peptides and protein water metal ions -CH,, -CHB, -C=C- -OH -CHz-, -CH,, -C=C- =O, -C=C-, -NHz -inositol -CHz-, -CH,, -C=C- =O, -OH, -NH, -C=C- -CH,, -CH,, -C=C-, =0, -OH, -NH, -C=C- nonpolar surfaces of carbohydrate Nonpolar residues and atoms, a-helix P-pleated sheet =0, -NH, polar residues, Interphase peptide of unknown configuration Polar “pockets” -OH P-0-, -choline+ , -serine- -NH: -P-0-, -choline+ -fucosate-, -sialate- Polar and ionic residues -OH +, ++ Upsala J Med Sci82 Double polar-nonpolar-polar leaflet construction of the plasma membrane 17 1 be 25.5 A long (1.5 &residue, cf. 9) in good agree- ment with the postulated width of the nonpolar core according to the present model. The abun- dance of basic proteins in some biological mem- branes (e.g. myelin, ref. 34) is interesting because the positively charged residues of arginine and lysine are separated by 5 and 4 atoms respectively of similar size from a plane of hydrogen bonding (see Fig. 5) while a negatively charged phosphorous oxygen of phospholipid is separated from such a plane by 4-5 atoms, indicating the possibility of favourable conditions for electrostatic interactions between these residues. However, ionic solutes are probably important contributors to the planes of salt bridges (see Fig. 5). Ca++ ions have been reported to bind to the carboxyl groups of protein in the human erythrocyte membrane and these binding sites are shielded by amino groups (42, 43). In accordance with these findings, the carboxyl groups of glutamic and aspartic acid residues would be separated from a plane of hydrogen bonding by only 2 and 1 atom(s) respectively and shielded from the surrounding aqueous medium by the amino groups of phospholipid and protein (see Fig. 5 and Table 11). Analytical freeze-cleaving has re- vealed that cholesterol may be enriched three-fold in the outer leaflet compared with the inner leaflet of the erythrocyte membrane (40). This finding is not compatible with the existence of a transmem- branous nonpolar lipid core which would facilitate the rapid equilibrium of the nonpolar cholesterol molecule between the two leaflets in the absence of specific binding to protein and/or lipid. Structural similarities between the present model and the bilayer model, of importance for experi- mental determination of physical parameters can be pointed out. These include the orientation of the axis of lipid perpendicular to the plane of the mem- brane and increased mobility of methyl groups in comparison with methylene groups. The latter are probably being more rigidly held by the planes of hydrogen bonding and fixed by adjacent nonpolar amino acid residues. If ions are excluded from the planes of hydrogen bonding and if cholesterol is present in the membrane (conferring on the lipids a preferred extended configuration, 18) then the re- gion of comparatively low electron density would be 28-33 A (see Fig. 5, the nonpolar amino acid resi- dues contribute 2-5 A). The presence of adsorbed ions in the planes of salt bridges would confer on these regions a higher electron density than that of the surrounding buffer. The planes of salt bridges containing electron-dense phosphorous and metal ions are probable reflection planes, which when detected by wide-angle X-ray analysis give rise to diffraction patterns. For example the Bragg spacing related to lipid in the bovine erythrocyte membrane as determined by wide-angle X-ray diffraction was 43 8, while that of extracted phospholipid was 51.5 8, (27). This discrepancy has led t o the postulation of “tilted” phospholipid in the assumed lipid bilayer phase of that particular biological membrane (27). However, according to the present model for the plasma membrane the centres of the reflecting planes are between 28 and 53 8, (most probably 4 0 4 3 8,) apart (see Fig. 5) in better agreement with the 43 8, spacing reported from the bovine erythro- cyte membrane. Thus, physical data which tend to support the lipid bilayer model, are also applicable t o the present model for biological membranes. However, biological membranes are regarded’ as delicate structures with optimized functions which have been acquired during the course of evolution. They are characterized by cooperative effects (98) and vectorial non-random motion in the plane of the membrane (7, 17), phenomena which are difficult to reconcile with the fluid mosaic model for cell mem- branes. The present model is compatible with: 1. Translational rearrangements in the plane of a membrane coupled or not coupled to translational rearrangements in an adjacent membrane. 2. Vectorial movement of protein and lipid in “channels” (in the plane of a membrane) formed by the asymmetric apposition of protein, carbohydrate and lipid. 3. Rearrangements along an axis perpendicular to the plane of a membrane permitting transmem- branous diffusion of macromolecules in their water-soluble form across one or two membranes (see figure 6 in ref. 15). 4. Adsorption of water-soluble enzymes, e.g. pro- teases and phospholipases with or without rear- rangements and digestion of protein and lipid. 5. Shrinking of the membrane surface by folding, lipid bilayer formation and/or by exclusion of aqueous phase proteins. 6. Expansion of the membrane surface by integra- tion of aqueous phase proteins in the plane of the membrane, /3-pleated sheet formation or by unfold- ing of the entire membrane. 7. Cooperative effects mediated by conforma- Upsalu J M e d Sci 82 172 E . Cerven tional changes of lipid and structural protein and/or membrane-associated water (cf. 37) or by modified permeability properties of either of the two mem- branes. 8. Receptor-effector transducing mechanisms (e.g. the adenylcyclase system) across two mem- branes by ( a ) cooperative effects (see above), ( b ) proteins spatially linked in the intramembranous space (see Fig. 3), (c) penetration of the first mes- senger through one leaflet combined with allosteric regulation of enzymes in the other leaflet. 9. Dissociability of regulatory and catalytic sub- units of hormone-sensitive enzyme complexes lo- cated in different membranes. 10. Specific o r preferred interactions between protein and lipid, conferring on biological mem- branes individuality in each organelle and each cell line derived from different species. Thus, biological membranes may contain a significant amount of information and may be of semiconservative nature since lipid and proteins are partly complementary constituents of the membranes. Fusion The demonstration that fusion may occur (8), for instance between cells and extracellular vesicles, does not contradict the present hypothesis for the structure of plasma membranes. On the contrary, some experimental data from fusion experiments may be interpreted advantageously by assuming that the plasma membrane is two membranes. This is because the lipid vesicles may fuse either with the outer leaflet only, or with regions of the plasma membrane where the two leaflets are already fused. For example fusion with the outer leaflet would re- sult in loading of the intramembranous space with compounds enclosed in the bilayer vesicles and leakage as has been reported (8) of these com- pounds through a “semiporous” outer leaflet. Fu- sion of human lymphocytes with mono-layered liposomes containing a fluorescent probe detectable only when diluted, resulted in fluorescence that was inhibited by 0.1 % Triton X-100, thus suggesting either that the vesicles were adsorbed to the cell surface (116) o r that the fluorescent probe was di- luted in the entire cell (116) o r in the intramem- branous space after fusion of the vesicles with the outer leaflet. Liposome-cell interactions involve adsorption of positively charged vesicles to the negatively charged cell surface (8, 65), endocytosis (8), and possible exchange of lipid between the cell surface and the liposomes (8). Complex interactions are presumably also involved in the fusion of human erythrocytes with lipid LP-X vesicles accumulating in serum during some hepato-biliary diseases since the fusion process (detected by enlarged surface area) does not occur in the absence of serum (1 11). The dilution of the intramembranous particles con- taining glycoprotein (see below) in these fusion ex- periments (111) may indicate that the intramem- branous particles contain reserve material for form- ing membrane area. In agreement with this inter- pretation, antigens residing in the band 3 protein (see below) were detectable close to the intramem- branous particles but not elsewhere (84), probably reflecting cryptic location at these other sites. Fu- sion of cells mediated by liposomes and virus mem- branes has sometimes been used as an argument for the occurrence of unperturbed lipid bilayer in cell membranes. However, if such were the case, the same argument would, contrary to what is generally observed, favour extensive fusion of cells in the absence of liposomes. Fusion of intracellular vesicles with the plasma membrane (exocytosis) may in addition involve either displacement of the inner membrane or re- lease of trapped compounds in the intramembran- ous space, followed by diffusion through the outer leaflet. Present model applied to the human erythrocyte membrane According t o the present model, the two leaflets of the erythrocyte membrane which stain with osmium-tetroxide represent two nonpolar layers containing cholesterol and other lipids with carbon double bonds that induce precipitation of electron- dense osmium compounds (117). In addition, amino groups of protein (12) and polar heads of phos- pholipid (88) may contribute to osmium stain- ing. Glutaraldehyde is believed to preserve a trilaminar appearance of the erythrocyte membrane of intact cells by cross-linking the proteins of the inner leaflet (cf. 67, 100). The inner leaflet of human erythrocytes is prefer- entially visualized by glutaraldehyde fixation fol- lowed by embedding in water-soluble polymers and staining with uranyl-acetate and lead-citrate (5 1 ) . The interleaflet space of the plasma membrane in such preparations where extensive extraction of lipid is avoided may be more than 70 A (51, 80) Upsala J Med Sci 82 Double polar-nonpolar-polar leaflet construction of the plasma membrane 173 which is indicative of a polar character. This in- terpretation is supported by the facts that the inter- leaflet space in such preparations is usually more than 100 8, in tight junctions where the outer membrane leaflets of two adjacent cells adhere (80) and these preparations when stained with acidified silicotungstic acid reveal substantial gaps between tissue cells (81) probably corresponding to the outer leaflet containing carbohydrate and the intramem- branous spaces. The profound difference between the appearance of the plasma membrane and the intracellular mem- branes under the electron microscope is well- documented by several preparation techniques (5 1, 80, 96). It is also amply demonstrated by conven- tional preparation of vesicles derived from yeast cell plasma membranes and mitochondria (cf. 45). A thin electron-lucent rim in one leaflet of the plas- ma membrane can be seen in some electron microscopic images prepared by embedding in water-soluble polymers (51, 80, compare Fig. 5). Divalent cations may contribute to the electron density of both leaflets of the human erythrocyte membrane by association with negatively charged residues (42, 43). The main contribution to a trilaminar image of the erythrocyte membrane would, however, be the ad- ditive electron density of components located at different depths of the transverse section (see Fig. 2). This superposition effect probably diminishes the importance of the specific affinity of stains to different chemical compounds and increases the importance of close spatial location of the compo- nents (cf. 15). The fracture plane of the erythrocyte membrane is presumably created by shear forces, generated during freezing, between contracting (18, 104) nonpolar layers and the expanding aqueous phase. The outer leaflet is probably anchored in the ex- tracellular ice by glycoprotein, while the inner leaflet is anchored in the intracellular ice by water- soluble protein (see Fig. 3) leading to predominant intramembranous splitting of the plasma mem- brane . The lipid bilayer model may be included in this system as a special structure adoptable by a single leaflet or by fusion of two leaflets (see figure 4 in ref. 15). In addition one leaflet may be formed by proteins or by protein and lipid in close association. Perturbation of a delicate structure such as that illustrated in Fig. 5 could be expected to induce a lipid bilayer structure in one leaflet and consequent multiple fracture planes of the plasma membrane. This interpretation is particularly attractive in freeze-etch preparations where the cell surface may be identified (figure 10a and c in ref. 32). The erythrocyte membrane is u delicate structure The lipids of intact human erythrocytes are digested neither with pancreatic phospholipase A, (89), nor with phospholipase C from Bucillus cereus (89). However, lipids of the corresponding permeable ghosts are extensively digested by these and other lipolytic enzymes (46, 89, 112, 120). Hydrolysis of 2 0 4 8 % of the total phospholipid of intact human erythrocytes by phospholipase A, from Nuju nuja orland sphingomyelinase from Staphylococcus au- reus results in ultrastructural alterations of both the inner and outer leaflets of the membrane (112). Similarly, digestion of intact rat erythrocytes with sphingomyelinase from Staphylococcus uureus and phospholipase C from Bacillus cereus under non- lytic conditions profoundly affects the appearance of both leaflets, leading to a significantly lower den- sity of the intramembranous particles (46). Lipo- lytic enzymes from different sources digest phos- pholipids of intact erythrocytes from different species in a specific manner reflecting the protec- tion afforded by adjacent molecules in the mem- brane (110, 112, 120). The degree of hydrolysis and the morphological effects of lipolytic enzymes upon erythrocytes influenced by the sequence of diges- tion ( I 12), the intracellular ATP concentration (46), the presence or absence of extracellular albumin (49), and protecting effects of non-digestible lipid and protein ( 1 10, 112, 120). The complexity of these interactions is difficult to reconcile with the fluid mosaic model for cell membranes, according to which most of the erythrocyte surface is formed by unperturbed lipid bilayer. Molecular composition of the two leaflets of the human erythrocyte membrane The major proteins of the human erythrocyte mem- brane have been characterized by their electro- phoretic migration in SDS-polyacrylamide gels (102) whereby mainly 7 protein-staining (bands I- VII) and 3 carbohydrate-staining (the sialoglycopro- tein, glycophorin) bands have been distinguished. By various techniques, including protease digestion and the labelling of intact erythrocytes and resealed ghosts with radioactive markers, only band I11 and Upsain .I Med Sci82 174 E . Cerve'n glycophorin of the main proteins have been shown to be exposed at the outer surface (102). These proteins, both of which are hydrophobically as- sociated with lipid, constitute about 30% (102) of the total protein mass and the remaining 70% are not to a similar extent detectable from the outside (102). Some of these remaining 70 % detectable in leaky ghosts (102) o r inside-out vesicles are susceptible to crosslinking agents (100) and may form a continuum in the intact membrane (12, 102). Two of them, the band I and I1 proteins, constitute about 30% of the total protein mass (102) and are jointly designated spectrin. It has been estimated that the amount of spectrin is sufficient to cover the inside of the erythrocyte (12). Immunological methods have established that spectrin is located on the cyto- plasmic side of the membrane (76). The solubility of these proteins in 80 % ethanol (20) and their capabil- ity for both polar and nonpolar interactions with phospholipid (53) suggest that they may be as- sociated with lipid in siru. They can be extracted from human ghosts by incubation with cholate (22), EDTA (87) or neutral and alkaline buffers of low ionic strength (87). These procedures lead to ve- siculation of ghosts, solubilization of up to 90 % of the membrane proteins (87) and/or alteration of the trilaminar structure obtained by osmium-staining the pelleted material obtained from high-speed centrifugation (22,87). Extraction with water under gentle stirring seems to specifically elute spectrin leaving a residue of atypical trilaminar membranes loosely associated with osmiophilic droplets which may represent phospholipid (87). In this connection it is noteworthy that phospholipid-glycoprotein aggregates ( 1 15) as well as proteins (36) and lipid bilayers (72) may give rise to trilaminar images in the electron microscope. The fact that osmium staining and dehydration leads to significant losses of protein and lipid (67) is also pertinent. Spectrin is an elongated (fibrous) protein in aqueous solution (20) but fibrous networks have never been observed by freeze-etching to cover the inside of erythro- cytes or ghosts (108) although other fibrous proteins are detectable with this technique (106). Part of the spectrin is firmly associated with camel erythrocyte ghosts and is experimentally defined as an integral (not extractable by EDTA o r slight modifications of ionic strength) protein in that system (30). Camel erythrocyte ghosts contain five times more integral protein per unit surface area than d o human ghosts and a higher density of intramembranous particles which are also larger (30). Extraction of spectrin from camel erythrocyte ghosts leads to conforma- tional changes in the membrane as evidenced by decreased surface area and increased density observed in phase contrast microscope (85). Ve- siculation of human ghosts by incubation with pro- tamine is accompanied by the appearance of multi- ple fracture planes some of which form in the inner leaflet (the author's interpretation of figure IOU and c in ref. 32) and the shedding of protein-free lipid vesicles (32). The conclusion that lipid bilayer is formed from structures such as that illustrated in Fig. 5 during vesiculation is (contrary to the fluid mosaic model) compatible with the increased fluid- ity of lipids (and decreased mobility of protein) in the ghost membrane as measured by electron spin resonance (1) under conditions which lead to ve- siculation (cf. 32). Other evidence (cf. 101, 102) indicates that the band V (actin-like polypeptide) and the band VI (glyceraldehyde 3-phosphate dehydrogenase, GAPDH) proteins are components of the inner leaflet. The band IV and VII proteins are however more difficult to locate. The band I, 11, V, VI and VII proteins can be extracted from ghosts by incu- bation with cholate (22) indicating that this proce- dure preferentially solubilizes the inner leaflet. Af- ter extraction, the remaining thin (22) trilaminar structure is predominantly composed of the band 111 and IV proteins, the sialoglycoproteins and some lipid (22). This cholate-extracted cholesterol and lipid-depleted structure should be expected to exhibit staining properties markedly different from those of the native membrane. Incubation in 8 M urea also seems to elute preferably the proteins of the inner leaflet leaving mainly the band I11 protein associated with about 50 % of the lipid (54). On the other hand, Triton X-100 and Tween 20 preferably solubilize the proteins of the outer leaflet (22, 59). The contours of the ghosts can be seen with the aid of the phase contrast microscope during elution with cholate or Triton X-100 (22). Ghost-like structures can also be seen after these procedures by electron-microscopic examination (22). These structures are globular after extraction with Triton X-100 and trilaminar after extraction with cholate (22). These data justify the suggestion that spectrin is the structural (48) protein of the inner leaflet while the band I11 protein and glycophorin are Upsala J Med Sci 82 Double polar-nonpolar-polar leaflet construction of the plasma membrane 175 the structural proteins of the outer leaflet of the human erythrocyte membrane. (The hydrophobic segment of glycophorin is only about 25.5 A (see above) which is not sufficient for spanning two leaflets. See also page 176 for an evaluation of ex- perimental data concerning the localization of pro- teins in native membranes.) If spectrin is not as- sociated with lipid in situ, then a highly ordered network of hydrogen bonds and salt bridges must be assumed to exist in the inner leaflet in order to account for the low permeability of the erythrocyte membrane towards various polar solutes. Since spectrin may be partly extracted from human ghosts (full extraction requiring sonication) with appar- ently maintained structure (31), it is reasonable to assume that other proteins as well contribute to the formation of the inner leaflet. The extracellular activity of GAPDH and other glycolytic enzymes which is firmly associated with intact cells (3, 6, 90) may reflect the “semiporous” properties of the outer leaflet. It is notable in this connection that phosphate added t o Hela cells and erythrocytes may first enter the 1,3- diphosphoglycerate and ATP pools (77) indicating that phosphate transport is actually mediated by GAPDH, which presumably holds a transmem- branous position in the inner leaflet. The outer leaflet of the human erythrocyte membrane may be regarded as the diffusion barrier inferred from studies of glucose and galactose transport (86), while the carrier mechanism for these solutes is situated in the inner leaflet and is destroyed con- comitantly with the proteolytic cleavage of spectrin by proteolytic enzymes included in resealed ghosts (66). This is also supported by the finding that GAPDH (and aldolase) most of which is bound to the inner (101) surface of the ghost membrane is more available to its substrate than are a number of cytoplasmic enzymes (e.g. lactate dehydrogenase and pyruvate kinase) under various conditions (28) indicating that this enzyme (and aldolase) constitute or is located exterior to a diffusion barrier (presum- ably the inner leaflet). These interpretations are corroborated by the demonstration that the en- zymatic properties of GAPDH are not significantly different in the membrane-associated and s o h - bilized states (91). The Na+- and K+-stimulated ATPase is presuma- bly located in the inner leaflet (cf. 63). Ouabain and potassium are regarded as being capable of influenc- ing the activity of this enzyme by binding to the outer surface of the inner leaflet after passage through the outer leaflet. The intramembranous particles Conditions which labilize the spectrin (the band I and I1 proteins) tend to facilitate aggregation of the intramembranous particles (3 1). Cross-linking of spectrin with glutaraldehyde inhibits particle ag- gregation (31) (as well as solubilization of these proteins by sonication in the absence of divalent metal ions (31)). These and other findings (84) sug- gest that spectrin is involved in creating o r forming the intramembranous particles seen by freeze fracture (31, 84). There is also much evidence correlating the intramembranous particles with glycophorin, the main sialoglycoprotein of the hu- man erythrocyte membrane (107). However, re- cently it was shown that mutant erythrocytes which lack this protein still contain intramembranous particles (13). Ferritin-conjugated antibodies di- rected against antigens residing in glycophorin as well as in the band 111 protein bind to the surface of human erythrocytes in a pattern that is reminiscent of the pattern formed by the intramembranous particles (84, 107). Thus spectrin, glycophorin, the band I11 protein and other proteins and lipid (60) as well, may be involved in forming the intra- membranous particles which are tetrameric or heterogeneous when visualized by rotatory replica- tion (64). Several difficulties attend the correlation of morphological observations on freeze-fractured specimens with other parameters. These include variations due to the choice of freeze fracture equipment and methodology (60), the proportion of cross fractures to tangential fractures (60) and altered fracture planes (99). The findings that fixa- tion (79) cooling before freezing ( I 18) and the pres- ence of dimethylsulphoxide or glycerol (57) affect the particle frequency of various biological mem- branes indicate that particles are impermanent structures. They d o not necessarily represent pro- teins of high molecular weight since they are found in protease-digested membranes lacking high mo- lecular weight proteins (1 19) and in artificial mem- branes containing hydrophobic peptides (92). The fluid mosaic model for cell membranes pre- dicts that the density of the intramembranous parti- cles representing intramembranous protein (94, 95) should increase when lipid but not protein is re- moved (21, 60) from the membrane by digestion Upsulu J Med S C I 82 176 E . Cerven with phospholipase C. This is because the smooth fracture planes are considered to represent the no,npolar region of the membrane built up of lipid bilayer (95) in which the integral proteins are freely (39) diffusible (95) and appear after freeze-etching as intramembranous particles. Contrary to this pre- diction, a decreased density of the particles follow- ing digestion with phospholipase C was reported (60). Ghosts digested maximally with phospholipase C very rarely split tangentially (60). When they do, smooth fracture faces are occasionally (60) but not always (1 12) produced. If the membrane-associated particles of ghosts are permanent intramembranous structures, it is unlikely that their higher density would counteract the tangential splitting of the plasma membrane, since aggreagated particles have been visualized repeatedly (84, 107). Exogenously applied proteins of high molecular weight (e.g. actin, M W : nx43000; ferritin, M W : 450 000; the tetramer of phytohemagglutinin, MW: 126000; the tetramer of Concanavalin A, M W : 110000) bound to the cell surface may be visualized by the freeze fracture procedure (83, 84, 106, 107). If it is assumed that endogenous globular and fi- brous peripheral proteins of high molecular weight (e.g. actin, M W : nx43000; spectrin, M W : 200 000-240 000, and the tetramer of GAPDH, M W : 140000) are also visualized by this procedure, then they must be confined to the intramembranous particles. This is because both the intracellular and extracellular surfaces are comparatively smooth (see ref. 83). These considerations are compatible with the view that proteins experimentally defined as peripheral proteins (94, 95) partly form the inner leaflet of intact erythrocytes and minimally perturbed ghosts. The significance of water-soluble proteins in forming membrane area is also adduced from the observation that extensively washed and resealed ghosts are smaller than slightly washed and resealed ghosts (55). It follows from these considerations that the outer leaflet cannot be a lipid bilayer, since such structures are cleaved smoothly by freeze-fracture, and no fracture planes are seen within the outer leaflet under ordinary conditions. Current status of previous reports Proteins spanning two leaflets of the plasma mem- brane (95) were included in the original version of the present model for biological membranes with some reservations (15). However, other features of the model such as the postulated complex interac- tions between various residues forming the planes of hydrogen bonding and salt bridges and also the transmembranous diffusion of water-soluble protein must be taken into account. Such phenomena may partly compensate for the apparent vectorial label- ling and digestion of intact cells, resealed ghosts, resealed inside-out and right side-out vesicles and leaky ghosts (cf. 102). The interdependency of pro- tein and lipid conformations in erythrocyte mem- branes is documented by electron-spin resonance studies (93, 103, 113). The complex pattern of diges- tion of intact erythrocytes with lipolytic enzymes (110, 112, 120) is tentatively explained by polar interactions between protein and lipid. In this con- nection is may be added that iodination of intact mitochondria with the lactoperoxidase method re- sults in labelling of one component apparently re- lated to the inner membrane (1 1) and the same pro- cedure applied to intact erythrocytes may cause labelling of hemoglobin (109). Thus, the existence of proteins spanning two polar-nonpolar-polar leaflets with those properties required by the pres- ent model for the structure of plasma membranes and other biological membranes is not corroborated by the available evidence. The globular structures preserved by some electronmicroscopic preparatory techniques (67, 68, 96, 97) may be interpreted as foldings of one leaflet around globular water-soluble structures (15). Other arrangements are equally possible, the characterization of which depends on further de- finement of the preparatory techniques (97). General considerations By analogy with the inhibitory effect of proteolytic enzymes on glucose and galactose carrier function in erythrocytes (66), the function of giant axons is similarly disturbed by proteolytic action on the in- ner leaflet of the axolemma (105). A thermodynamic analysis of nerve cell depolarization across, two membranous leaflets separated by a polar space has been presented and found to describe satisfactorily the experimental data ( 5 6 ) . The interpretation of data from nerve cell axons is complicated by the occurrence of serial membranes in addition to the two leaflets of the plasma membrane. However, according to the double leaflet model for biological membranes only the inner leaflet is involved in energy transducing activity (15). Therefore, the periaxonal space deduced from experimental devia- Upsulu J Med Sci 82 Double polar-nonpolar-polar leaflet construction of the plasma membrane 177 tions from theoretically expected depolarization currents ( 2 ) may partly represent the intramem- branous space. The possibility of a sodium gradient primarily across the inner leaflet (cf. 15) is an attrac- tive explanation why the generalizing of the concept of sodium-dependent transport of certain metabo- lites across cellular membranes (cf. 19) fails ex- perimentally (73). The main criticism that can be directed against the lipid bilayer concept for biological membranes involves the fact that no other structure which ex- plains the experimental data has seriously been con- sidered. The controlled references in the experi- ments have of necessity been lipid bilayers or pro- tein in aqeous media. For example the fact that fracture planes are generated in the nonpolar core of lipid bilayer ( 2 5 ) seems t o suggest that non- cooperative hydrophobic interactions are always responsible for the fracture planes in biological membranes. Estimation of the amount of a a-helix, /3-structure and random coil in biological mem- branes rests on the use of water-soluble reference peptides without considerations of additional in- teractions possibly occurring in biological mem- branes. Interpretation of X-ray diffraction data in- volves a comparison with electronmicroscopic pictures that are considered to represent bilayers. Interpretation of electron spin resonance data sometimes rests on the assumption that the lipid is statistically distributed in the native membrane which is assumed to be a lipid bilayer. However, electron spin resonance studies should be in- terpreted with care, since a statistical distribution of spin labels and lipid is not to be expected a priori (93, 113). ACKNOWLEDGEMENT I thank Dr Gunnar Agren and Dr Gunnar Ronquist for generous support and for reading the manuscript. This work was supported by grants from the Swedish Research Council (Project B77-13X-228-13B), the University of Uppsala and a personal grant from Uddeholms AB. REFERENCES 1 . Adams, D., Markes, M. E., Leivo, W. J. & Carra- way, K . L.: Electron spin resonance analysis of ir- reversible changes induced by calcium perturbation of erythrocyte membranes. Biochim Biophys Acta 426: 3 8 4 5 , 1976. 2. Adelman, W. J., Jr, Palti, Y. & Senft, J. P.: Potas- sium ion accumulation in a periaxonal space and its effect on the measurement of membrane potassium ion conductance. J Membr Biol13: 387-410, 1973. 3 . Agren, G., Ponten, J., Ronquist, G. & Westermark, B.: Formation of extracellular adenosine triphos- phate by normal and neoplastic human cells in culture. J Cell Physiol77: 331-335, 1971. 4. - Nucleoside diphosphate kinase at the cell surface of neoplastic human cells in culture. J Cell Physiol 5 . Agren, G. & Ronquist, G.: On the availability of certain metabolites at the outer surface of normal and malignant cells for the membranous d e novo synthesis of ATP and other nucleotides. Upsala J Med SciSO: 1 4 , 1975. 6. - Formation of extracellular adenosine triphos- 83: 91-101, 1974. 7 8 9 10. 11. 12. 13. 14. 15. 16. phate by tumour cells. Acta Physiol Scand 75: 124-128, 1%9. Albrecht-Buehler, G. & Chen, L. B.: Local inhibi- tion of centripetal particle transport where LETS protein patterns appear on 3T3 cells. Nature 266:454456, 1977. Batzri, S. & Korn, E . D.: Interaction of phospholipid vesicles with cells. Endocytosis and fusion as alternate mechanisms for the uptake of lipid-soluble and water-soluble molecules. J Cell Biol66: 621-634, 1975. Bohinski, R. C.: Modem Concepts in Biochemistry. Allyn & Bacon Inc., Boston, London, Sydney and Toronto, 1976. Bosmann, H. B., Myers, M. W., Dehond, D., Ball, R. & Case, K. R.: Mitochondria1 autonomy. Sialic acid residues on the surface of isolated rat cerebral cortex and liver mitochondria. J Cell Biol Brdiczka, D. & Schumacher, D.: Iodination of peripheral mitochondria1 membrane proteins in cor- relation to the functional state of the ADP/ATP car- rier. Biochem Biophys Res Com 73: 823-832. Bretscher, M. S.: Membrane structure: Some gen- eral principles. Membranes are asymmetric lipid bilayers in which cytoplasmicdly synthesized pro- teins are dissolved. Science 181: 622-629, 1973. Bachi, Th., Whiting, K., Tanner, M. J. A., Metax- as, M. N. & Anstee, D. J . : Freeze-fracture electron microscopy of human erythrocytes lacking the major membrane sialoglycoprotein. Biochim Biophys Acta 464: 635-639, 1977. Cabantchik, 2. I., Balshin, M., Breuer, W . , Markus, H. & Rothstein, A.: A comparison of intact human red blood cells and resealed and leaky ghosts with respect to their interactions with surface labelling agents and proteolytic enzymes. Biochim Biophys Acta382: 621-633, 1975. CervCn, E.: The fluid double polar-nonpolar-polar leaflet model for biological membranes. Upsala J Med Sci 81: 193-200, 1976. CervCn, E.: The migrating thermodynamic quantum hypothesis for cytoplasmic streaming, sodium pump- ing and other cell biological phenomena, deduced from biofunctional considerations of the ultra- structure of brush border microvilli. Upsala J Med Sci81: 201-212, 1976. 5.5: 147-160, 1972. Upsala J M e d Sci82 178 E . Cervkn 17. CervCn, E.: Characterization of a rate-limiting factor involved in ectosialyltransferase activity of Ehrlich ascites cells. Submitted for publication 1977. 18. Chapman, D.: Phase transitions and fluidity characteristics of lipids and cell membranes. Quarterly Reviews of Biophysics 8, 2: 185-235, 1975. 19. Christensen, H. N.: Biological Transport, second edition. W. A. Benjamin, Inc., London, Amsterdam, Don Mills, Ontario, Sydney, Tokyo, 1975. 20. Clarke, M.: Isolation and characterization of a water-soluble protein from bovine erythrocyte membranes. Biochem Biophys Res Com 45: 1063- 1070, 1971. 21. Coleman, R., Finean, J. B., Knutton, S. & Limbrick, A. R.: A structural study of the modification of erythrocyte ghosts by phospholipase C. Biochim Biophys Acta219: 81-92, 1970. 22. Coleman, R., Holdsworth, G. & Finean, J. B.: De- tergent extraction of erythrocyte ghosts. Compari- son of residues after cholate and Triton X-100 treat- ments. Biochim Biophys Acta436: 38-44, 1976. 23. Comte, I . , Maisterrena, B. & Gautheron, D. C.: Lipid composition and protein profiles of outer and inner membranes from pig heart mitochondria. Comparison with microsomes. Biochim Biophys Acta419: 271-284, 1976. 24. Cuatrecasas, P.: Insulin receptor of liver and fat cell membranes. Fed Proc32, 8: 1838-1846, 1973. 25. Deamer, D. & Branton, D.: Fracture planes in an ice-bilayer model membrane system. Science 158: 655-657, 1967. 26. Deman, J . , van Vaerenbergh, P. M. & Joos, P.: Evidence for sialic acid-lecithin binding at the plasma membrane in the presence of cations. Europ J Cancer 7: 3 17-323, 197 I . 27. Dresdner, G. W., Cid-Dresdner, H., Meza, L . & Cardemil, E.: Wide-angle X-ray-diffraction studies of the bovine-erythrocyte membrane. Treatment with sodium chloride, deoxycholate and lipid sol- vents. Can J Biochem52: 1044-1052, 1974. 28. Duchon, G . & Collier, H. B.: Enzyme activities of human erythrocyte ghosts: Effects of various treat- ments. J Membr Biol6: 138-157, 1971. 29. Dunkley, C . R . , Manery, J. F. & Dryden, E. F.: The conversion of ATP to IMP by muscle surface en- zymes. J Cell Physiol68: 241-247, 1966. 30. Eitan, A., Aloni, B. & Livne, A.: Unique properties of the camel erythrocyte membrane. 11. Organization of membrane proteins. Biochim Biophys Acta 426: 647-658, 1976. 3 I . Elgsaeter, A. & Branton, D.: Intramembrane particle aggregation in erythrocyte ghosts. I. The effects of protein removal. J Cell Biol63: 1018-1030, 1974. 32. Elgsaeter, A., Shotton, D. M. & Branton, D.: In- tramembrane particle aggregation in erythrocyte ghosts. 11. The influence of spectrin aggregation. Biochim Biophys Acta426: 101-122, 1976. 3 3 . Engelman, D. M.: Surface area per lipid molecule in the intact membrane of the human red cell. Nature 34. Eylar, E. H.: Amino acid sequence of the basic pro- 223: 1279-1280, 1969. tein of the myelin membrane. Proc Nat Acad Sci 3 5 . Feeney, R. E.: A biological antifreeze. A glycopro- tein in the blood of polar fishes lowers the freezing temperature. Am Sci62: 712-719, 1974. 36. Fernhdez-Mor&n, H.: Membrane ultrastructure in nerve cells. I n The Neurosciences. A Study Program (ed: G. C. Quarton, T. Melnechuk and F . 0. Schmitt), pp. 28 1-304. The Rockefeller University Press, New York, 1967. 37. Finch, E. D. & Schneider, A. S.: Mobility of Water Bound to Biological Membranes. A proton NMR relaxation study. Biochim Biophys Acta 406: 1 4 6 154, 1975. 38. Finean, J. B., Coleman, R., Green, W. G. & Limbrick, A. R.: Low-angle X-ray diffraction and electron-microscope studies of isolated cell mem- branes. J Cell Sci 1: 287-296, 1966. 39. Finegold, L.: Cell membrane fluidity: Molecular modeling of particle aggregations seen in electron microscopy. Biochim Biophys Acta 448: 393-398, 1976. 40. Fisher, K . A.: Analysis of membrane halves: Cholesterol. Proc Nat Acad Sci USA 73: 173-177, 1976. 41. Fleischer S . , Fleischer, B . & Stoeckenius, W.: Fine structure of whole and fragmented mitochondria af- ter lipid depletion. Fed Proc 24: 296, 1965. 42. Forster, J . & Manery, J. F.: Calcium binding by human erythrocyte membranes. Biochem J 124: 563- 571, 1971. 43. Forstner, J. & Manery, J. F.: Calcium binding by human erythrocyte membranes. Significance of car- boxyl, amino and thiol groups. Biochem J 125: USA 67: 1425-1431, 1970. 343-352, 1971. 44. Frey, C. A. & Bryan, W. P.: The blue dextran ex- cluded volume of the human erythrocyte membrane. Biochim Biophys Acta356: 156-163, 1974. 45. Fuhrmann, G. F., Wehrli, E. &Boehm, C.: Prepara- tion and identification of yeast plasma membrane vesicles. Biochim Biophys Acta363: 295-310, 1974. 46. Gazitt, Y . , Loyter, A., Reichler, Y . & Ohad, I.: Correlation between changes in the membrane or- ganization and susceptibility to phospholipase C at- tack induced by ATP depletion of rat erythrocytes. Biochim Biophys Acta419: 479-492, 1976. 47. Gorter, E. & Grendel, R.: On biomolecular layers of lipid on the chromocytes of the blood. J Exp Med 41:439, 1925. 48. Green, D. E. & Perdue, J. F.: Correlation of mito- chondrial structure and function. Ann N.Y. Acad Sci 137: 667484, 1966. 49. Gul, S. & Smith, A . D.: Haemolysis of washed hu- man red cells by the combined action of Nuju naju phospholipase A, and albumin. Biochim Biophys Acta288: 237-240, 1972. 50. Hall, M. R., Meinke, W., Goldstein, D. A. & Lerner, R.: Synthesis of cytoplasmic membrane-associated DNA in lymphocyte nucleus. Nat New Biol 51. Heckman, C . A. & Barnett, R. J.: GACH: A water- miscible, lipid-retaining embedding polymer for 234: 227-229, 1971. Upsala J M e d Sci82 Double polar-nonpolar-polar leaflet construction of the plasma membrane 179 electron microscopy. J Ultrastruct Res 42: 156-179, 1973. 52. Hynes, R. 0.: Cell surface proteins and malignant transformation. Biochim Biophys Acta 458: 73-107, 1976. 53. Juliano, R. L., Kimelberg, H. K. & Papahad- jopoulos, D.: Synergistic effects of a membrane pro- tein (spectrin) and Ca2+ on the Na+ permeability of phospholipid vesicles. Biochim Biophys Acta 241:894-905, 1971. 54. Juliano, R. L. & Rothstein, A.: Properties of an erythrocyte membrane lipoprotein fraction. Biochim Biophys Acta249: 227-235, 1971. 55. Johnson, R. M.: The kinetics of resealing of washed erythrocyte ghosts. J Membr Biol 22: 231-253, 1975. 56. Katchalsky, A.: Membrane thermodynamics. In The Neurosciences. A Study Program (ed.: G. C. Quarton, T. Melnechuk and F. 0. Schmitt), pp. 326-343. The Rockefller University Press, New York, 1967. 57. Kirk, R. G. & Tosteson, D. C.: Cation transport and membrane morphology. J Membr Biol 12: 273-285, 1973. 58. Laki, K.: Actin as an energy transducer. J Theor Biol44: 117-130, 1974. 59. Liljas, L. L., Lundahl, P. & HjertCn, S.: The major sialoglycoprotein of the human erythrocyte mem- brane. Release with a non-ionic detergent and purifi- cation. Biochim Biophys Acta426: 526-534, 1976. 60. Limbrick, A. R. & Knutton, S.: The perturbation of the human erythrocyte membrane by phospholipase C. J Cell Sci 19: 341-355, 1975. 61. Livne, A. & Kuiper, P. J.: Unique properties of the camel erythrocyte membrane. Biochim Biophys Acta318: 4 1 4 9 , 1973. 62. Marchesi, V. T., Furthmayr, H. & Tomita, M.: The red cell membrane. I n Ann Rev Biochem, Vol. 45 (ed. E. E. Snell), pp. 667-698. Ann Rev Inc, Palo Alto, 1976. 63. Marchesi, V. T. & Palade, G. E.: The localization of Mg++-Na+-K+-activated adenosine triphosphatase on red cell ghost membranes. J Cell Biol35: 385- 404,1%7. 64. Margaritis, L. H., Elgsaeter, A. & Branton, D.: Ro- tary replication for freeze-etching. J Cell Biol 65. Martin, F. & MacDonald, R.: Liposomes can mimic virus membranes. Nature252: 161-163, 1974. 66. Masiak, S . J. & Lefevre, P. G.: Glucose transport inhibition by proteolytic degradation of the human erythrocyte membrane inner surface. Biochim Bio- phys Acta465: 371-377, 1977. 67. McMillan, P. N. & Luftig, R. B.: Preservation of erythrocyte ghost ultrastructure achieved by various fixatives. Proc Nat Acad Sci USA 70:3060-3064, 1973. 68. - Preservation of membrane ultrastructure with aldehyde or imidate fixatives. J Ultrastruct Res 52: 243-260, 1975. 69. Meisenheimer, R. G., Fischer, 3. W. & Rehfeld, S. J.: Thallium in human erythrocyte membranes: An 72: 47-56, 1977. X-ray photoelectron spectroscopy study. Biochem Biophys Res Com 68: 994-999, 1976. 70. Meymaris, E.: Chemistry and physiology of local anaesthesia. Chemical anatomy of the nerve mem- brane. Br J Anaesth47: 164472, 1975. 71. Mitchison, J. M.: A polarized light analysis of the human red cell ghost. J Exptl Biol30: 397432,1953. 72. Miyamoto, V. K. & Stoeckenius, W.: Preparation and characteristics of lipid vesicles. J Membr Biol 73. Naftalin, R. J. & Holman, G. D.: The effects of removaI of sodium ions from the mucosal solution on sugar absorption by rabbit ileum. Biochim Biophys Acta419: 385-390, 1976. 74. Nicolson, G. L.: Transmembrane control of the re- ceptors of normal and tumour cells. I. Cytoplasmic influence over cell surface components. Biochim Biophys Acta457: 57-108, 1976. 75. Nicolson, G., Lacorbiitre, M. & Delmonte, P.: Outer membrane terminal saccharides of bovine liver nuclei and mitochondria. Exp Cell Res 71: 4 6 8 4 7 3 , 1972. 76. Nicolson, G. L., Marchesi, V. T. & Singer, S . J.: The localization of spectrin on the inner surface of human red blood cell membranes by femtin- conjugated antibodies. J Cell Biol51: 265-272, 1971. 77. Niehaus, W. G. Jr & Hammerstedt, R. H.: Mode of orthophosphate uptake and ATP labeling by Mam- malian cells. Biochim Biophys Acta 443: 515-524, 1976. 78. Oschman, J. L. & Wall, B. J.: Calcium binding to intestinal membranes. J Cell Biol55: 58-73, 1972. 79. Parish, G. R.: Changes of particle frequency in freeze-etched erythrocyte membranes after fixation. J Microsc 104: 245-256, 1975. 80. Pease, D. C.: Glycol methacrylate copolymerized with glutaraldehyde and urea as an embedment re- taining lipids. J Ultrastruct Res45: 124-148, 1973. 81. Pease, D. C. & Peterson, R. G.: Polymerizable glutaraldehyde-urea mixtures as polar water- containing embedding media. J Ultrastruct Res 41: 133-159, 1972. 82. Peters, R . : The effect of rehydration on the thickness of erythrocyte membranes as observed by energy transfer. Biochim Biophys Acta318: 4 6 9 4 7 3 , 1973. 83. Pinto da Silva, P. & Branton, D.: Membrane splitting in freeze-etching covalently bound femtin as a membrane marker. J Cell Biol45: 598-605, 1970. 84. Pinto da Silva, P. & Nicolson, G. L.: Freeze-etch localization of concanavalin A receptors to the membrane intercalated particles of human erythro- cyte ghost membranes. Biochim Biophys Acta 363:311-319, 1974. 85. Ralston, G. B.: Proteins of the camel erythrocyte membrane. Biochim Biophys Acta401: 83-94, 1975. 86. Regen, D. M. & Tarpley, H. L.: Anomalous trans- port kinetics and the glucose carrier hypothesis. Biochim Biophys Acta339: 218-233, 1974. 87. Reynolds, J. A. & Trayer, H.: Solubility of mem- brane proteins in aqueous media. J Cell Biol 88. Robertson, J. D.: Molecular structure of biological 4:252-269, 1971. 246:1337-7342, 1971. Upsala J Med Sci 82 180 E . CervPn membranes. I n Handbook of Molecular Cytology (ed. A. Lima-de-Faria), pp. 1404-1443. North- Holland Publishing Company, Amsterdam & London 1972. 89. Roelofsen, B., Zwaal, R. F. A., Comfurius, P., Woodward, C. B. & Van Deenen, L. L. M.: Action of pure phospholipase A, and phospholipase C on human erythrocytes and ghosts. Biochim Biophys Acta241: 925-929, 1971. 90. Ronquist, G.: Studies on the formation of adenosine triphosphate by the cell membrane of human eryth- rocytes and of tumour cells. Acta Universitatis Up- saliensis, Abstr Upps Dissert Medicine, No. 54: 1 , 1 968. 91. Schrier, S. L . , Junga, I. &Johnson, M.: Comparison of the characteristics of the immobilized and solu- bilized glyceraldehyde phosphate dehydrogenase of human erythrocyte membranes. Life Sci 17: 735- 738, 1975. 92. Segrest, J. P., Gulik-Krzywicki, T. & Sardet, C.: Association of the membrane-penetrating poly- peptide segment of the human erythrocyte MN- glycoprotein with phospholipid bilayers. I. Forma- tion of freeze-etch intramembranous particles. Proc Nat Acad Sci USA 71: 3294-3298, 1974. 93. Shiga, T., Suda, T. & Maeda, N.: Spin label studies on the human erythrocyte membrane. Two sites and two phases for fatty acid spin labels. Biochim Bio- phys Acta466: 231-244, 1977. 94. Singer, S. J.: The molecular organization of biologi- cal membranes. I n Structure and Function of Biolog- ical Membranes (ed. L. I . Rothfield), pp. 145-222. Academic Press, New York and London, 1971. 95. Singer, S. J. & Nicolson, G. L.: The fluid mosaic model of the structure of cell membranes. Science 175:720-731, 1972. 96. Sjostrand, F. S.: A comparison of plasma mem- brane, cytomembranes, and mitochondria1 mem- brane elements with respect to ultrastructural fea- tures. J Ultrastruct Res 9: 561-580, 1963. 97. Sjostrand, F. S.: The problems of preserving molecular structure of cellular components in con- nection with electron microscopic analysis. J Ultrastruct Res55: 271-280, 1976. 98. Sonnenberg, M.: Interaction of human growth hormone and human erythrocyte membranes studied by intrinsic fluorescence. Proc Nat Acad Sci USA 68: 1051-1055, 1972. 99. Speth, V., Wallach, D. F. H., Weidekamm, E. & Kniifermann, H.: Micromorphologic consequences following perturbation of erythrocyte membranes by trypsin, phospholipase A, lysolecithin, sodium dodecyl sylfate and saponin. A correlated freeze- etching and biochemical study. Biochim Biophys Acta255: 386-394, 1972. 100. Steck, T. L.: Cross-linking the major proteins of the isolated erythrocyte membrane. J Mol Biol 66: 295-305, 1972. 101. Steck, T. L.: The organization of proteins in human erythrocyte membranes. I n Membrane Research (ed. C. F. Fox), pp. 71-93. Academic Press, New York and London, 1972. 102. Steck, T. L.: The organization of proteins in the human red blood cell membrane. A review. J Cell Biol62: 1-19, 1974. 103. Tanaka, K.-I. & Ohnishi, S . - I . : Heterogeneity in the fluidity of intact erythrocyte membrane and its homogenization upon hemolysis. Biochim Biophys Acta426: 218-231, 1976. 104. Tanford, C.: The Hydrophobic Effect: Formation of Micelles and Biological Membranes. John Wiley & Sons, New York, London, Sydney & Toronto, 1973. 105. Tasaki, I . : Nerve Excitation. A Macromolecular Approach. Ch. C. Thomas, Publisher, Springfield, Illinois, USA, 1968. 106. Tillack, T. W. & Marchesi, V. T.: Demonstration of the outer surface of freeze-etched red blood cell membranes. J Cell Biol45: 649-653, 1970. 107. Tillack, T. W., Scott, R. E. & Marchesi, V. T.: The structure of erythrocyte membranes studied by freeze-etching. 11. Localization of receptors for phytohemagglutinin and influenza virus to the in- tramembranous particles. J Exptl Med 135: 1209- 1227, 1972. 108. Tilney, L. G. & Detmers, P.: Actin in erythrocyte ghosts and its association with spectrin. Evidence for a nonfilamentous form of these two molecules in situ. J Cell Biol66: 508-520, 1975. 109. Tsai, C.-M., Huang, C.-C. & Canellakis, E. S.: Iodi- nation of cell membranes. I . Optimal conditions for the iodination of exposed membrane components. Biochim Biophys Acta332: 47-58, 1973. 110. Turner, J. C., Anderson, H. M. & Gandal, C. P.: Comparative liberation of bound phosphatides from red cells of man, ox and camel. Proc SOC Exptl Biol Med 99: 547-550, 1958. 1 1 1 . Verkleij, A. J., Nauta, I. L. D., Werre, J. M., Man- dersloot, J . G., Reinders, B., Ververgaert, P. H. J . Th. & de Gier, J.: The fusion of abnormal plasma lipoprotein (LP-X) and the erythrocyte membrane in patients with cholestasis studied by electron- microscopy. Biochim Biophys Acta 436: 366-376, 1976. 112. Verkleij, A. J . , Zwaal, R. F. A., Roelofsen, B., Comfurius, P., Kastelizn, D. & van Deenen, L. L. M.: The asymmetric distribution of phospho- lipids in the human red cell membrane. A combined study using phospholipases and freeze-etch electron microscopy. Biochim Biophys Acta 323: 178-193, 1973. 13. Verma, S. P. & Wallach, D. F. H.: Evidence for constrained lipid mobility in the erythrocyte ghost. A spin label study. Biochim Biophys Acta 382: 73-82, 1975. 14. Waugh, D. F.: The ultrastructure of the envelope of mammalian erythrocytes. Ann N.Y. Acad Sci 50: 835-853, 1950. 15. Weinbaum, G., Fischman, D. A. & Okuda, S.: Membrane modifications in nutritionally induced filamentous Escherichia coli B. J Cell Biol 116. Weinstein, J. N., Yoshikami, S., Henkart, P., Blumenthal, R. & Hagins, W. A.: Liposome-cell interaction: Transfer and intracellular release of a 45: 493-508, 1970. Upsala J Med Sci 82 Double polar-nonpolar-polar leaflet construction of the plasma membrane 181 trapped fluorescent marker. Science 195: 489-492, 1977. 117. White, D. L., Andrews, S. B., Faller, J. W. & Barr- nett, R. J.: The chemical nature of osmium tetroxide fixation and staining of membranes by X-ray photo- electron spectroscopy. Biochim Biophys Acta 436: 577-592, 1976. 118. Wunderlich, F . , Wallach, D. F. H., Speth, V . & Fischer, H.: Differential effects of temperature on the nuclear and plasma membranes of lymphoid cells. A study by freeze-etch electron microscopy. Biochim Biophys Acta373: 34-43, 1974. 119. Yamanaka, N . & Deamer, D. W.: Protease digestion of membranes. Ultrastructural and biochemical ef- fects. Biochim Biophys Acta426: 132-147, 1976. 120. Zwaal, R . F. A., Roelofsen, B. & Colley, C. M.: Localization of red cell membrane constituents. Biochim Biophys Acta300: 159-182, 1973. Received May 16,1977 Address for reprints: Erik CervCn Instute of Medical and Physiological Chemistry Biomedical Center Box 515 University of Uppsala S-75123, Uppsala Sweden Upsulu J Med Sci 82