sign in sign up
<<
Home , Cell membrane, Cytoplasm, Fluid mosaic model, Lipid bilayer, Membrane fluidity, Phospholipid

The of the Structure of Author(s): S. J. Singer and Garth L. Nicolson Source: Science, New Series, Vol. 175, No. 4023 (Feb. 18, 1972), pp. 720-731 Published by: American Association for the Advancement of Science Stable URL: http://www.jstor.org/stable/1733071 Accessed: 12/09/2008 12:57

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aaas.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

American Association for the Advancement of Science is collaborating with JSTOR to digitize, preserve and extend access to Science.

http://www.jstor.org experimental evidence in terms of the model; and (v) to show that the fluid mosaic model suggests new ways of The Fluid Mosaic Model of the thinkingabout membranefunctions and phenomena. Structure of Cell Membranes Thermodynamics and Cell membranes are viewed as two-dimensional solutions Membrane Structure of oriented globular and . The fluid mosaic model has evolved by a series of stages from earlier ver- sions (1-4). Thermodynamicconsidera- tions about membranesand membrane S. J. L. Singer and Garth Nicolson components initiated, and are still cen- tral to, these developments.These con- siderations derived from two decades of intensive studies of and nu- Biological membranesplay a crucial considerable detail several models of cleic structures;the thermodynamic role in almost all cellular phenomena, the gross structural organization of principles involved, however, are per- yet our understandingof the molecular membranes,in terms of the thermody- fectly general and apply to any macro- organizationof membranesis still rudi- namics of macromolecularsystems and molecular system in an aqueous en- mentary.Experience has taughtus, how- in the light of the then available ex- vironment. These principles and their ever, that in order to achieve a satisfac- perimentalevidence. From this analysis, application to membrane systems have tory understandingof how any biologi- it was concluded that a mosaic struc- been examined in detail elsewhere (1) cal system functions, the detailed ture of alternating globular proteins and are only summarizedhere. For our molecular composition and structureof and bilayer was the only present purposes, two kinds of non- that system must be known. While we membranemodel among those analyzed covalent interactions are most impor- are still a long way from such knowl- that was simultaneouslyconsistent with tant, hydrophobic (5) and hydrophilic edge about membranesin general,prog- thermodynamicrestrictions and with all (1). By hydrophobic interactions is ress at both the theoretical and experi- the experimental data available. Since meant a set of thermodynamicfactors mental levels in recentyears has brought that article was written, much new evi- that are responsible for the sequester- us to a stage where at least the gross dence has been published that strongly ing of hydrophobicor nonpolar groups aspects of the organizationof the pro- supportsand extends this mosaic model. away from water, as, for example, the teins and lipids of membranes can be In particular,the mosaic appears to be immiscibility of and discerned.There are some investigators, a fluid or dynamic one and, for many water. To be specific, it requires the however, who, impressedwith the great purposes, is best thought of as a two- expenditure of 2.6 kilocalories of free diversityof membranecompositions and dimensional oriented viscous solution. energy to transfer a mole of methane functions, do not think there are any In this article, we therefore present and from a nonpolar medium to water at useful generalizationsto be made even discuss a fluid mosaic model of mem- 25?C (5). Free energy contributionsof about the gross structure of cell mem- brane structure, and propose that it is this magnitude,summed over the many branes. We do not share that view. We applicable to most biological mem- nonpolar residues of soluble suggest that an analogy exists between branes, such as plasmalemmal and in- proteins, are no doubt of primary im- the problems of the structure of mem- tracellular membranes, including the portance in determiningthe conforma- branes and the structure of proteins. membranes of different cell , tions that protein adopt in The latter are tremendously diverse in such as mitochondriaand . (6), in which the non- composition, function, and detailed These membranes are henceforth re- polar residues are predominantly se- structure. Each kind of protein mole- ferred to as functional membranes. questered in the interior of the mole- cule is structurallyunique. Nevertheless, There may be some other membrane- cules away from contact with water. generalizations about protein structure like systems, such as , or the By hydrophilic interactions is meant a have been very useful in understanding lipoproteinmembranes of small animal set of thermodynamicfactors that are the propertiesand functions of protein , which we suggest may be rigid, responsible for the preference of ionic molecules. Similarly, valid generaliza- rather than fluid, mosaic structures,but and polar groups for an aqueous rather tions may exist about the ways in which such membrane systems are not a pri- than a nonpolar environment. For ex- the proteins and lipids are organized in mary concern of this article. ample, the free energy requiredto trans- an intact membrane.The ultimate test Our objectivesare (i) to review briefly fer a mole of zwitterionic glycine from of such generalizations, or models, is some of the thermodynamicsof macro- water to acetone is about 6.0 kcal at whether they are useful to explain old molecular, and particularlymembrane, 25?C, showing that pairs strongly experiments and suggest new ones. systems in an aqueous environment; prefer to be in water than in a non- Singer (1) has recently examined in (ii) to discuss some of the properties polar medium (1). These and related of the proteins and lipids of functional free energy terms no doubt provide the Dr. Singer is a professor of at the Uni- versity of California at San Diego, La Jolla. Dr. membranes; (iii) to describe the fluid reasons why essentially all the ionic Nicolson is a research associate at the Armand mosaic model in to residues of molecules are ob- Hammer Cancer Center of the Salk Institute for detail; (iv) analyze protein Biological Studies, La Jolla, California. some of the recent and more direct served to be in contact with water, 720 SCIENCE, VOL. 175 usually on the outer surface of the mol- Fig. 1. A phospho- (8) of erythrocytemembranes, sche- ecule, according to x-ray crystallo- bilayer: which can be removed by chelating matic cross-sectional under mild are ex- graphic studies. Similar thermodynamic view. The filled cir- agents conditions, argumentsapply to saccharide residues cles represent the amples of membrane proteins that sat- (1). It requires the expenditure of sub- ionic and polarhead isfy the criteria for peripheralproteins. stantial free energy to transfer a simple groups of the phos- On the other hand, the major portion saccharide from water to a Ita? pholipid molecules, (> 70 of the of most nonpolar which make contact percent) proteins and membranes have different characteris- solvent, such residueswill therefore withwater; the wavy be in a lower free energy state in con- lines represent the tics, which may be assigned to integral tact with water than in a less polar chains. proteins: (i) they require much more environment. drastic treatments, with reagents such There are other noncovalent inter- as , , protein dena- actions, such as hydrogen bonding and (to the maximum extent feasible) from turants,or organicsolvents, to dissociate electrostatic interactions, which also contact with water, while the ionic and them from membranes;(ii) in many in- contributeto determinemacromolecular polar groups of the proteins-along stances, they remain associated with structure.However, with respectto gross with those of the lipids and the oligosac- lipids when isolated; (iii) if completely structure, with which we are now charides-should be in contact with the freed of lipids, they are usually highly concerned, these are very likely of sec- aqueous solvent. These requirements insolubleor aggregatedin neutral aque- ondary magnitude compared to hydro- place restrictions on models of mem- ous buffers(9). phobic and hydrophilic interactions. brane structure;in particular,they ren- The distinction between peripheral der The familiar phospholipid bilayer highly unlikely the classical model and integral proteins may be useful in structureillustrates the combinedeffects of a trilaminararrangement of a con- several regards. It is assumed that only of hydrophobic and hydrophilic inter- tinuous sandwiched be- the integral proteins are critical to the actions. In this structure (Fig. 1) the tween two of protein. The structural integrity of membranes. nonpolar fatty acid chains of the phos- latter model is thermodynamicallyun- Therefore, the properties and interac- pholipids are sequesteredtogether away stable because not only are the non- tions of peripheral proteins, while in- from contact with water, thereby maxi- polar amino acid residues of the mem- terestingin their own right, may not be mizing hydrophobic interactions. Fur- brane proteins in this model perforce directly relevant to the central prob- thermore, the ionic and zwitterionic largely exposed to water but the ionic lems of membranestructure. The prop- groups are in direct contact with the and polar groups of the lipid are se- erties of cytochrome c, for example, aqueous phase at the exterior surfaces questered by a layer of protein from may not be typical of mitochondrial of the bilayer, thereby maximizing hy- contact with water. Therefore, neither membrane proteins. Furthermore, the drophilic interactions. In the case of hydrophobic nor hydrophilic interac- of peripheral and integral zwitterionicphospholipids such as phos- tions are maximized in the classical proteins and their attachment to the phatidylcholine, dipole-dipole interac- model. membrane may be very different proc- tions between ion pairs at the surface esses. This is not the appropriate oc- of the bilayer may also contribute to casion to discuss membrane biogenesis the stabilizationof the ,bilayerstructure. Some Propertiesof in any detail, but it may be significant c is a mito- In applying these thermodynamic Membrane Components that, although cytochrome principles to membranes, we recognize chondrial protein, it is synthesized on first that of the three major classes of Peripheral and integral proteins. It cytoplasmic rather than mitochondrial membrane components-proteins, lip- seems both reasonable and important ;in fact only a small fraction ids, and -the proteins to discriminatebetween two categories of the total mitochondrialprotein (per- are predominant. The ratio by weight of proteins bound to membranes,which haps only the integral proteins of the of proteins to lipids ranges from about we have termed peripheral and integral inner mitochondrial membrane?) ap- 1.5 to 4 for those functional membranes proteins (1). Peripheralproteins may be pears to be synthesized on mitochon- which have been well characterized characterizedby the following criteria. drial ribosomes (10). In any event, [compare (7)]. A substantial frac- (i) They require only mild treatments, because of the relatively unimportant tion of this protein most probablyplays such as an increasein the ionic strength membrane structural role assigned to an important role in determining the of the medium or the addition of a the peripheralproteins, they are not a structureof membranes,and the struc- chelating agent, to dissociate them mo- primaryconcern of this article. tural properties of these proteins are lecularly intact from the membrane; Properties of integral proteins. Since therefore of first-order importance. (ii) they dissociate free of lipids; and the proteins we have classified as in- Membrane proteins are considered in (iii) in the dissociated state they are tegral, according to the criteria speci- some detail in the following section. At relatively soluble in neutral aqueous fied, constitute the major fraction of this juncture, the significant point is buffers. These criteria suggest that a membraneproteins, we assume that the that if hydrophobicand hydrophilicin- peripheralprotein is held to the mem- properties to be discussed apply to the teractions are to be maximized and the brane only by rather weak noncovalent integralproteins. lowest free energy state is to be at- (perhaps mainly electrostatic) interac- 1) For several well-characterized tained for the intact membrane in an tions and is not strongly associated membrane systems, including erythro- aqueous environment, the nonpolar with . The cytochrome cyte and other plasma membranes,and amino acid residues of the proteins- c of mitochondrial membranes, which mitochondrialmembranes, the proteins along with the fatty acid chains of the can be dissociatedfree of lipids by high have been shown to be grossly hetero- -should be sequestered salt concentrations, and the protein geneous with respect to molecular 18 FEBRUARY 1972 721 weights (11). There is no convincing lar proteins are attached to the outer Two qualifications should be stressed, evidence that there exists one predom- surfaces of a lipid bilayer (16) would however, concerning the bilayer form inant type of that is not be satisfactory because, among of membrane lipids. (i) None of the specifically a structural protein; recent other reasons, it would require mem- evidence so far obtained for the bilayer reports to the contrary have been with- brane thicknesses much larger than the form permits us to say whether the drawn. We consider this heterogeneity 75 to 90 angstroms generally observed. bilayer is continuous or interrupted (1). to be more significant for a general A model in which The calorimetrically observed phase model of membrane structure than the molecules are intercalated within the transitions, for example, occur over a fact that in a few specialized instances, membrane would, however, meet these broad temperature interval, allowing the as in the case of disk membranes of restrictions. possibility that the cooperative unit in- rod outer segments (12, 13), a The phospholipids of membranes. volved in the is quite single protein species predominates. A There is now substantial evidence that small, consisting perhaps of only 100 satisfactory membrane model must be the major portion of the phospholipids lipid molecules on the average. (ii) None capable of explaining the heterogeneity is in bilayer form in a variety of intact of the experiments mentioned above is of the integral membrane proteins. membranes. For example, differential sufficiently sensitive and quantitative to 2) The proteins of a variety of intact calorimetry of intact mem- prove whether 100 percent of the phos- membranes, on the average, show ap- branes shows that they undergo a phase pholipid is in the bilayer form. It is preciable amounts of the a-helical con- transition in a temperature range very therefore not excluded that some signifi- formation, as was first shown iby Ke similar to that of aqueous dispersions of cant fraction of the phospholipid (per- (14), Wallach and Zahler (4), and the phospholipids extracted from the haps as much as 30 percent) is physi- Lenard and Singer (3). For example, membranes (16, 17). Thus the structures cally in a different state from the rest circular dichroism measurements of of the lipid in the membrane and of the of the lipid. aqueous suspensions of intact and me- lipid in isolated aqueous dispersion are Protein-lipid interactions in mem- chanically fragmented human erythro- closely similar; presumably the latter is branes. Several kinds of experiments cyte membranes (provided that we take the bilayer form. This conclusion is sup- indicate that protein-lipid interactions into account certain optical anomalies ported iby x-ray diffraction '(18) and play a direct role in a variety of of these measurements) reveal that spir-label studies (19) on similar mem- membrane functions. Many membrane- about 40 percent of the protein is in brane preparations. bound and require the right-handed a-helical conformation The bilayer character of membrane lipids, often specific phospholipids, for (15). Most soluble globular proteins lipids rules out models such as that of the expression of their activities [see whose circular dichroism spectra have Benson (20) in which the proteins and table 2 in (21)]. Furthermore, the been obtained exhibit a smaller fraction lipids form a single-phase nature of the fatty acids incorporated of a-helix in their native structures. subunit that is repeated indefinitely in into phospholipids affects the function the mem- This suggests that the integral proteins two dimensions to constitute of certain membrane-bound proteins in in intact membranes are largely globu- brane. In such a model, most of the bacterial membranes (22). lar in shape rather than spread out as lipids would be expected to have dis- On the other hand, the calorimetric monolayers. On the other hand, a tinctly different properties from those data discussed above give no significant of membrane model in which such globu- a bilayer. indication that the association of pro- teins with the phospholipids of intact membranes affects the phase transitions of the phospholipids themselves. Ex- 4. periments with C and membranes have shown that the en- zymic release of 70 percent of the phosphorylated amines from intact erythrocyte membranes profoundly perturbs the physical state of the resid- ual fatty acid chains, but has no detect- able effect (as measured by circular dichroism spectra) on the average con- formation of the membrane proteins (2). Such results therefore suggest that the phospholipids and proteins of membranes do not interact strongly; in Fig. 2. The lipid-globular protein mosaic model of membrane structure: schematic fact, they appear to be largely inde- cross-sectional view. The phospholipids are depicted as in Fig. 1, and are arranged as pendent. and heads in contact with water. a discontinuous bilayer with their ionic polar Some This paradox, that different types of lipid may be structurally differentiated from the bulk of the lipid (see text), but this is not shown in the figure. The integral proteins, with the heavy lines repre- experiments suggest strong protein-lipid explicitly the and senting the folded polypeptide chains, are shown as globular molecules partially em- interactions on one hand, weak bedded in, and partially protruding from, the membrane. The protruding parts have or no interactions on the other, can be on their surfaces the ionic residues (- and +) of the protein, while the nonpolar resolved in a manner consistent with are in the embedded the molecules are am- residues largely parts; accordingly, protein all the data if it is proposed that, while The to which the are embedded and, in phipathic. degree integral proteins particular, the of the whether they span the entire membrane thickness depend on the size and structure of largest portion phospholipid the molecules. The arrow marks the plane of to be expected in freeze-etching is in bilayer form and not strongly experiments (see text). [From Lenard and Singer (3) and Singer (1)] coupled to proteins in the membrane, 722 SCIENCE, VOL. 175 a small fraction of the lipid is more adopt an amphipathicstructure, can be protein or in the association of two or tightly coupled to protein. With any integral proteins of membranes;in this more integral protein subunits to form one membrane protein, the tightly manner, the heterogeneity of the pro- a specific aggregate within the mem- coupled lipid might be specific; that is, teins of most functionalmembranes can brane. These features can be accom- the interaction might require that the be rationalized. modated in Fig. 2 without any changes phospholipidcontain specific fatty acid The same considerationsmay also ex- in the basic structure. chains or particularpolar head groups. plain why some proteins are membrane- The phospholipids of the mosaic There is at present, however, no satis- bound and others are freely soluble in structureare predominantlyarranged as factory direct evidence for such a dis- the . The difference may be an interruptedbilayer, with their ionic tinctive lipid fraction. This problem is that either the amino acid sequence of and polar head groups in contact with considered again in connection with a the particularprotein allows it to adopt the aqueous phase. As has been dis- discussionof the experimentsof Wilson an amphipathic structure or, on the cussed, however, a small portion of the and Fox (23). contrary,to adopt a structurein which lipid may be more intimately associated the distributionof ionic groupsis nearly with the integral proteins. This feature spherically symmetrical, in the lowest is not explicitly indicated in Fig. 2. The Fluid Mosaic Model free energy state of the system. If the thickness of a mosaic membranewould ionic distribution on the protein sur- vary along the surface from that across Mosaic structure of the proteins and face were symmetrical, the protein a phospholipid bilayer region to that lipids of membranes. The thermody- would be capable of interactingstrongly across a protein region, with an average namic considerationsand experimental with water all over its exterior surface, value that could be expected to corre- results so far discussed fit in with the that is, it would be a monodispersesol- spond reasonablywell to experimentally idea of a mosaic structure for mem- uble protein. measured membrane thicknesses. branes (1-3, 24) in which globularmol- The mosaic structure can be readily of the mosaic: lipid or pro- ecules of the integral proteins (perhaps diversified in several ways. Although tein? In the cross section of the mosaic in particularinstances attached to oli- the nature of this diversificationis a structurerepresented in Fig. 2, it is not gosaccharides to form , matter of speculation,it is importantto indicatedwhether it is the proteinor the or interactingstrongly with specific lip- recognizethat the mosaic structureneed phospholipidthat providesthe matrix of ids to form )alternate with not be restrictedby the schematic rep- the mosaic. In other words, which com- sections of phospholipid bilayer in the resentation in Fig. 2. Protein-protein ponent is the mortar, which the bricks? cross section of the membrane(Fig. 2). interactionsthat are not explicitly con- This question must be answered when The globularprotein molecules are pos- sidered in Fig. 2 may be important in the third dimensionof the mosaic struc- tulated to be amphipathic(3, 4) as are determiningthe propertiesof the mem- ture is specified. Trhesetwo types of the phospholipids. That is, they are brane. Such interactions may result mosaic structure may be expected to structurallyasymmetric, with one highly either in the specific binding of a have very differentstructural and func- polar end and one nonpolar end. The peripheral protein to the exterior ex- tional properties, and the question is highly polar region is one in which the posed surface of a particular integral therefore a critical one. It is our hy- ionic amino acid residues and any co- valently bound saccharide residues are clustered, and which is in contact with the aqueous phase in the intact mem- brane; the nonpolarregion is devoid of ionic and saccharide residues, contains many of the nonpolar residues, and is embedded in the hydrophobic interior of the membrane. The amphipathic structure adopted by a particular in- tegral protein (or lipoprotein), and therefore the extent to which it is embedded in the membrane, are under thermodynamic control; that is, they are determined by the amino acid se- quence and covalent structure of the protein, and by its interactionswith its molecularenvironment, so that the free energy of the system as a whole is at a minimum.An integralprotein molecule with the appropriatesize and structure, or a suitable aggregateof integral pro- teins (below) may transversethe entire membrane (3); that is, they have re- gions in contact with the aqueous sol- Fig. 3. The lipid-globularprotein mosaic model with a lipid matrix (the fluid mosaic vent on both sides of the membrane. model); schematic three-dimensional and cross-sectional views. The solid bodies with surfaces It is clear from these considerations stippled represent the globular integral proteins, which at long range are randomly distributed in the plane of the membrane. At short some that range, may form differentproteins, if they have the specific aggregates, as shown. In cross section and in other details, the legend of appropriate amino acid sequence to Fig. 2 applies.

18 FEBRUARY 1972 723 pothesis that functional cell membranes area of the membrane surface. In able evidence. Some more recent re- have a long-range mosaic structure with other words, we suggest that long-range sults, however, bear even more directly the lipids constituting the matrix, as is random arrangements in membranes are on the problem, and only this evidence shown in Fig. 3. Supporting evidence is the norm; wherever nonrandom distri- is discussed below. discussed later. At this point, let us butions are found, mechanisms must consider some of the consequences of exist which are responsible for them. this hypothesis. 2) It has been shown that, under Some Recent Experimental Evidence 1) There should generally be no long- physiological conditions, the lipids of range order in a mosaic membrane with functional cell membranes are in a Evidence for proteins embedded in a lipid matrix. By long range, we mean fluid rather than a crystalline state. membranes. One proposal of the fluid over distances of the order of a few (This is not true of myelin, however.) mosaic model is that an integral pro- tenths of a micrometer and greater. This evidence comes from a variety of tein is a globular molecule having a Suppose we have a membrane prepara- sources, such as spin-labeling experi- significant fraction of its volume em- tion containing many different protein ments (25), x-ray diffraction studies bedded in the membrane. The results species, and suppose further that 10,000 (18), and differential calorimetry (16, of recent freeze-etching experiments molecules of protein A are present in 17). If a membrane consisted of integral with membranes strongly suggest that a the membrane of a single cell or or- proteins dispersed in a fluid lipid matrix, substantial amount of protein is deeply ganelle. How is protein A distributed the membrane would in effect be a two- embedded in many functional mem- over the membrane surface? If the dimensional liquid-like solution of mon- branes. In this technique (26) a frozen membrane proteins formed the matrix omeric or aggregated integral proteins specimen is fractured with a microtome of the mosaic, defined by specific con- (or lipoproteins) dissolved in the lipid knife; some of the frozen water is sub- tacts between the molecules of different bilayer. The mosaic structure would be limed (etched) from the fractured sur- integral proteins, protein A might be a dynamic rather than a static one. The face if desired; the surface is then distributed in a highly ordered, two- integral proteins would be expected to shadow cast with metal, and the surface dimensional array on the surface. On undergo translational within replica is examined in the mi- the other hand, if lipid formed the the membrane, at rates determined in croscope. By this method the topog- matrix of the mosaic, there would be no part by the effective viscosity of the raphy of the cleaved surface is re- long-range interactions intrinsic to the lipid, unless they were tied down by vealed. A characteristic feature of the membrane influencing the distribution some specific interactions intrinsic or exposed surface of most functional of A molecules, and they should there- extrinsic to the membrane. However, membranes examined by this technique, fore be distributed in an aperiodic - because of their amphipathic structures, including plasmalemmal, vacuolar, nu- dom arrangement on the membrane the integral proteins would maintain clear, , mitochondrial, and surface. their molecular orientation and their bacterial membranes (27, 28), is a The absence of long-range order degree of intercalation in the membrane mosaic-like structure consisting of a should not be taken to imply an ab- while undergoing translational diffusion smooth matrix interrupted by a large sence of short-range order in the mem- in the plane of the membrane (as dis- number of particles. These particles brane. It is very likely that such short- cussed below). have a fairly characteristic uniform range order does exist, as, for example, In contrast, if the matrix of the mo- size for a particular membrane, for among at least some components of the saic were constituted of integral pro- example, about 85-A diameter for eryth- electron transport chain in the mito- teins, the long-range structure of the rocyte membranes. Such surfaces re- chondrial inner membrane. Such short- membrane would be essentially static. sult from the cleavage of a membrane range order is probably mediated by Large energies of activation would be along its interior hydrophobic face specific protein (and perhaps protein- required for a protein component to (29). This interior face (Fig. 2) corre- lipid) interactions leading to the forma- diffuse in the plane of the membrane sponds to the plane indicated by the tion of stoichiometrically defined ag- from one region to a distant one be- arrow. If cleavage were to occur gregates within the membrane. How- cause of the many noncovalent bonds smoothly between the two layers of ever, in a mosaic membrane with a between the proteins that would have phospholipid in the bilayer regions, but lipid matrix, the long-range distribu- to be simultaneously broken for ex- were to circumvent the protein mole- tion of such aggregates would be ex- change to take place. Therefore, a cules penetrating the mid-plane of the pected to be random over the entire mosaic membrane with a protein ma- membrane, then the alternating smooth surface of the membrane. trix should make for a relatively rigid and particulate regions observed on the The objection may immediately be structure with essentially no transla- freeze-etch surfaces can be readily ex- raised that long-range order clearly tional diffusion of its protein compo- plained by a mosaic structure for the exists in certain cases where differen- nents within the membrane. membrane (Fig. 2), provided that the tiated structures (for example, ) From the discussion in this and the particles can be shown to be protein are found within a membrane. We sug- previous section, it is clear that the in nature. That the particles are indeed gest, in such special cases, either that fluid mosaic model suggests a set of protein has been suggested by recent short-range specific interactions among structural properties for functional experiments (30). integral proteins result in the formation membranes at least some of which can Another consequence of the mosaic of an unusually large two-dimensional be tested experimentally. In an earlier model, suggested from its inception aggregate or that some agent extrinsic article (1), a large body of experimen- (3), is that certain integral proteins pos- to the membrane (either inside or out- tal evidence was examined for its rele- sessing the appropriate size and struc- side the cell) interacts multiply with vance to models of membrane structure. ture may span the entire thickness of specific integral proteins to produce a It was concluded that a mosaic struc- the membrane and be exposed at both clustering of those proteins in a limited ture was most consistent with the avail- membrane surfaces. Chemical evidence

724 SCIENCE, VOL. 175 that a trans-membraneprotein, whose the treated (sensitized) cells were lysed Rho(D) site (36) on the mem- molecular weight is about 100,000, is at an air-water interface, so that the brane. Since the clusters were distrib- present in large amounts in the human cell membranes were spread out flat. uted in a random array, we conclude erythrocyte membrane has been ob- The flattened membranes, after being that the Rho(D) antigen, which exhibits tained by two independent methods- picked up on an properties of an integral protein (37), one involving of normal grid, were treated with the specific "in- is molecularly dispersed and is distrib- compared to everted membranes (31), direct stain," -conjugatedgoat uted in a random,two-dimensional array and the other specific chemical labeling specific for human y-globu- on the human erythrocyte membrane. of the membraneproteins !(32). lin. Thus, whereverthe human anti-Rho Similar experimentswere carried out Distribution of components in the (D) molecules were bound to the Rho with the H-2 alloantigenic sites on plane of the membrane. A prediction (D) antigen on the membrane surface, mouse erythrocyte membranes. In this of the fluid mosaic model is that the the ferritin-labeledgoat antibodies be- case (Fig. 5) the clusters -of ferritin two-dimensionallong-range distribution came specifically attached. In other molecules of the indirect stain were not of any integral protein in the plane of words, the human y-globulin isolated, as in the case of the Rho(D) the membrane is essentially random. now functioned as an antigen for the antigen,but insteadoccurred in patches. To test this prediction, we have devel- goat antibodies (Fig. 4). The ferritin The patchy distribution of the H-2 oped and applied electron microscopic was distributedin discrete clusters,each histocompatibilityalloantigenic sites had techniques to visualize the distribution containing two to eight ferritin mole- earlier been observed by different tech- of specific membrane antigens over clues within a circle of radius about niques (38), but the two-dimensional large areas of their membrane surfaces 300 A. The numblersof such clusters distributionof the patches could not be (33) and have so far studied the dis- per unit area of the membrane surface ascertained. In our experiments, the tribution *of the Rhd(D) antigen on corresponded to the number of 125I- patches contained variable numbers of human erythrocytemembranes (34), and labeled human anti-Rho(D) molecules clusters, and were arranged in an ir- of H-2 histocompatibility alloantigens bound per unit area. This indicates that regular two-dimensional array on the on mouse erythrocytemembranes (35). each ferritin cluster was bound to a membrane surface. The histocompati- In the case of the Rho(D) antigen, single anti-Rho(D)molecule, and a clus- bility antigen appears to be glycopro- for example, cells of 0, Rh-positive ter represents the number of goat tein in nature (39). The long-rangedis- type were reacted with a saturating antibody molecules bound to a single trilbutionof both the Rho(D) and H-2 amount of 125I-labeledpurified human human y-globulin molecule. Each clus- histocompatibilityantigens on their re- antibody to Rho(D) [anti-Rlho(D)],and ter therefore corresponds to a single spective membrane surfaces, therefore,

Fig. 4 (left). The outer membrane surface of an Rh-positive human erythrocyte sensitized with human anti-Rho(D) and stained with ferritin-conjugatedgoat antibody to human y-globulin. The cells were first labeled to saturation with purified lIL-labeled human anti- body to Rho(D) and then lysed at an air-water interface. The erythrocyte membrane ghosts, flattened by surface forces (inset, low magnification) were picked up on a coated, electron microscope grid and indirectly stained with ferritin-conjugatedgoat anti- bodies to human y-globulin. The ferritin appears bound to the membrane in discrete clusters of two to eight ferritin-conjugates;each cluster is circumscribedby a circle of radius 300 A. The number of such clusters per cell (9300) is equal within experimental error to the number of "2I-labeled human antibody to Rho(D) molecules bound per cell (10,200). Each cluster -therefore corresponds to an individual Rho(D) antigenic site. Scale is 0.1 tum;inset scale is i ,tm. [From Nicolson, Masouredis, and Singer (34)] Fig. 5 (right). The outer membrane surface on a mouse erythrocyte (H-2b) sensitized with alloantibodies against H-2b histocompatibil- ity antigens and stained with ferritin-conjugatedantibodies against 7S mouse r-globulin. The procedures are the same as listed in the legend to Fig. 4. The ferritin-antibodyclusters are present in randomly spaced "patches"of variable size on the membrane sur- face. Scale is 0.1 ,um. [From Nicolson, Hyman, and Singer (35)]

18 FEBRUARY 1972 725 are in accord with the prediction of the trated two-dimensionalfluid solution of explained by a diffusion mechanism for fluid mosaic model that the integral identical protein molecules will appear, the intermixingprocess, as follows. The proteins of membranes are randomly when dried, to be arranged in an or- antibodiesto the human arranged in two dimensions. dered array, particularly when optical were no doubt directed to a heteroge- The particles on the inner membrane tricks are used to enhance the apparent neous set of antigens, whereas the anti- faces revealed by freeze-etchingexperi- order (43). What is really a fluid phase bodies ito the mouse cell were directed ments, which (as discussed above) are may therefore artifactuallybe made 'to specifically to the histocompatibility probablyprotein in nature, are generally appear as a crystalline solid. This ap- alloantigen.However, the histocompati- also relatively randomly distributed in pears to be the situation with the reti- bility antigens occur as large aggregates two dimensions. nal receptor disk membranes. in the membrane (Fig. 5), and might Evidence that proteins are in a A major contribution to membrane therefore be expected to diffuse more fluid state in intact membranes. An im- studies has been made by Frye and slowly than a complex mixtureof largely portant series of experiments has been Edidin (44), who investigatedthe mem- unaggregated human antigens in the carried out (12, 40-44) with receptor brane properties of some membrane. Thus, at appropriateinter- disk membranesfrom the retina of the heterokaryons.Human and mouse cells mediate times after cell fusion, signifi- frog. This membranesystem is unusual in culture were induced to fuse with cant numbers of (M1/2-HI) but not of in that it contains as its predominant, one another, with Sendai as the (Ml-HI/2) fused cells might appear, to if not only, protein component the pig- fusing agent. The distributionof human be converted at longer times to cells ment .In electron microscopy and mouse antigeniccomponents of the with completelyintermixed components. of the negativelystained surfaces of the fused cell membranes was then deter- A rough estimate may be made of dried membranes, a somewhat tightly mined by immunofluorescence,with the the average effective diffusion constant packed and ordered array of par- use of rabbit antibodies directed to the required of the membrane components ticles (about 40 A) was observed.These whole human cells, mouse antibodies to account for the kinetics of intermix- particles are the individual rhodopsin directed against the H-2 alloantigen on ing in the Frye-Edidin experiments. the molecules. Although the earlier studies mouse cell membranes,and, as in- Taking the average distance of migra- suggested that there was a long-range direct stains, goat antiserum to rabbit tion, x, to have been about 5 micro- order in the distributionof the particles y-glolbulinand goat antiserumto mouse meters in a time, t, of 40 minutes gives (40), more recent x-ray diffractiondata y-globulin labeled with two different an apparentdiffusion constant, D=x2/ (42) on pellets of wet, receptor disk fluorescent dyes. Shortly after cell fu- 2t, of 5 X 10-11 cm2/sec. For com- membranes showed that only a few sion, the mouse and human antigenic parison, the diffusionconstant of hemo- orders of reflection were observed cor- components were largely segregated in glob,in in aqueous solutions is about respondingto the spacings of the rho- different halves of the fused cell mem- 7 X 10-7 cm2/sec. The apparent effec- dopsin molecules in the plane of the branes; but after about 40 minutes at tive viscosity of the membrane fluid membrane. This indicated that a non- 37?C the components were essentially phase is therefore about 103 to 104 crystalline, aperiodic arrangement of completely intermixed. Inhibitors of times that of water. the particles existed in the plane of the protein synthesis, of adenosine triphos- The Frye-Edidin experimentscan be membrane. Furthermore, the tempera- phate (ATP) formation, and of gluta- rationalized by the fluid mosaic model ture dependence of the characteristics mine-dependentsynthetic pathways, ap- of membranestructure as being the re- before of the x-ray diffraction maxima were plied or after cell fusion, had no sult of the free diffusionand intermixing effect on consistent with the suggestion that the the rate of this intermixing of the lipids and the proteins (or lipo- particles were in a planar liquid-like process, but lowering the temperature proteins) within the fluid lipid matrix. state in the intact membrane.,Additional below 15?C sharply decreased it. Some experiments, however, appear support for the existence of this liquid- Frye and Edidin (44) suggest that to suggest that the lipids of membranes like state was the observation that the the intermixing of membrane compo- are not readily interchangeablewithin absorptionof a foreign protein (bovine nents is due to diffusion of these com- the membrane and are therefore not albumin)to the membranecould ponents within the membrane, rather free to diffuse independently. For ex- definitely alter the x-ray spacings due than to their removal and reinsertion, ample, Wilson and Fox (23) have to the rhodopsin particles; that is, the or to the synthesis and insertion of studied the induction of /l-galactoside distributionof the rhodopsin molecules new copies of these components, into and 8/-glucoside transport systems in in the plane of the membranewas rad- the heterokaryonmembrane. An unex- mutants of Escherichia coli that cannot ically altered by the weak binding of plained findingof these experimentswas synthesizeunsaturated fatty acids. Such the albumin. This alteration would not the fairly frequent occurrence, at early fatty acids can be lincorporatedinto be expected if a rigid lattice structure and intermediatetimes after cell fusion, phospholipids,however, if they are sup- of Ithe rhodopsin molecules, or aggre- of heterokaryon membranes in which plied in the growth medium.When cells gates, were present in the plane of the the human antigenic components were were grown in particularfatty acid sup- membrane, uniformly distributed over the mem- plements and induced for the synthesis These studies are particularly note- brane surface but the mouse compo- of the transport systems, the effect of worthy because they involved a mem- nents were still largely segregated to temperatureon the transport rate was brane which, by conventional electron about half the membrane (Ml/2-H1 characteristicof that fatty acid. If, then, microscopictechniques, appears to show cells). On the other hand, the reverse the cells were first grown in medium long-range periodicity over its surface. situation, with the mouse antigenic containingoleic acid and then shifted to Other specialized membraneshave also components uniformly spread out over growth in a medium supplementedwith exhibited orderedelectron micrographic the membrane and the human compo- linoleic acid during a brief period of images of their surfaces [compare(43)]. nents segregated (M1-H1/2), was only induction of either of the transportsys- However, it is likely that a very concen- rarely observed.This result can now be tems, the effect of temperatureon trans- 726 SCIENCE, VOL. 175 port was said to be characteristic of cells charides on membranes in the electron surfaces must occur at only negligibly grown continually in the linoleic acid microscope (33). For example, the fer- slow rates. This conclusion probably medium. In other words, although most ritin conjugate of concanavalin A, a applies to membrane proteins other of the phopholipids of the membrane protein agglutinin that binds specifically than glycoproteins; for example, the contained oleic acid chains, these did to terminal a-D-glucopyranosyl or a-D- Na,K-dependent and Mg-dependent not appear to exchange with the newly mannopyranosyl residues (46), attaches adenosine triphosphatase activities of synthesized small amounts of phospholi- specifically to the outer surface of eryth- erythrocyte membranes are exclusively pids containing linoleic acid chains. rocyte membranes and not at all to localized to the inner cytoplasmic sur- These experiments, however, do not the inner cytoplasmic surface (33). A faces (51). Individual molecules of spin- necessarily contradict the thesis that similar, completely asymmetric distri- labeled zwitterionic and anionic phos- most of the phospholipids of membranes bution of ferritin conjugates of ricin (a pholipids also exhibit very slow inside- are freely diffusible and, hence, ex- protein agglutinin) on the membranes outside transitions in synthetic vesicles changeable. For example, each of the of rabbit erythrocytes is shown in Fig. of phospholipid bilayers (52). The very two transport systems might be or- 6. Ricin binds specifically to terminal slow or negligible rates of such transi- ganized in the membrane as a specific f/-D-galactopyranosyl and sterically re- tions can be explained by the aggregate containing intercal- lated residues (47). Such asym- model and the thermodynamic argu- ated and strongly bound phospholipid metry has now been observed with ments already discussed. If the integral components. If such lipoprotein aggre- several ferritin-conjugated agglutinins proteins (including the glycoproteins) gates had first to be assembled in order and a number of different mammalian in intact membranes have, like the phos- to be incorporated into the bulk lipid cell plasma membranes (48). These find- pholipids, an amphipathic structure, a matrix of the membrane, the results of ings extend earlier results obtained by large free energy of activation would Wilson and Fox would be anticipated. different methods (49). be required to rotate the ionic and polar In particular, the small fraction of The foregoing observations bear on regions of the proteins through the the membrane phospholipid that was many problems, including cell-cell inter- hydrophobic interior of the membrane strongly bound, and perhaps segregated actions and membrane biogenesis (50). to the other side. in such aggregates from the bulk of the In the context of this article, however, To accommodate the fluid mosaic membrane lipid, might not exchange the absence of oligosaccharides on in- model to these conclusions concerning rapidly with the bulk lipid. The Wilson- ner membrane surfaces indicates that asymmetry, we specify that, while the Fox experiments therefore do not re- rotational transitions of the glycopro- two-dimensional translational diffusion quire that the major part of the mem- teins of erythrocyte and other plasma of the integral proteins and the phos- brane phospholipid be static, but only membranes from the outer to the inner pholipids of membranes occurs freely, that a small fraction of the lipids be structurally differentiated from the rest. The structural differentiation of some of the membrane lipid by strong bind- ing to integral proteins is a possibility that was discussed above. The observations of Wilson and Fox, that there is a significant coupling of lipid and protein incorporation into membranes, appear to be a special case. The experiments of Mindich (45) dem- onstrate that more generally lipid and protein incorporation into bacterial membranes can occur independently, and 'thatquite wide variations in the ratio of lipids and proteins in the membrane can be produced in vivo, as might be expected from the fluid mosaic model of membrane structure. The asymmetry of membranes. A substantial amount of evidence has ac- cumulated showing that the two sur- faces of membranes are not identical in composition or structure. One aspect of this asymmetry is the distribution of oligosaccharides on the two surfaces of membranes. There exist proteins, Fig. 6. The inner (i) and outer (o) membrane surfaces of a rabbit erythrocyte mem- called lectins or plant agglutinins, which brane that has been stained with ferritin-conjugatedricin. In preparing membrane speci- 4 a cell with membrane bind to specific sugar residues, and, as mens such as are shown in Figs. and 5, occasionally lyses inner and outer surfaces of the membrane are In this a can cause the of rupture such that both exposed. result, agglutination case the mounted membrane was stained with ferritin conjugated to ricin, a plant agglu- cells the residues on their bearing sugar tinin that specifically binds to terminal p-D-galactopyranosyl and sterically related surfaces. By conjugating several such terminal sugar residues in oligosaccharides. The ferritin-agglutininis found on the outer agglutinins to ferritin, we have been able membrane surface only. The scale is equivalent to 0.1 ,m; the insert scale is equiva- to visualize the distribution of oligosac- lent to 1 Am. 727 18 FEBRUARY 1972 the rotational diffusion of these com- sional solution, thereby allowing new ture, may be proposed. Consider first ponents is generally restricted to axes thermodynamic interactions among the the proteolysis experiments with nor- perpendicular to the plane of the mem- altered components to take effect. This mal cells. Suppose that the integral gly- brane; that is, in general, molecular general mechanism may play an im- coproteins in the normal cell mem- tumbling does not occur at significant portant role in various membrane-me- brane are molecularly dispersed in the rates within the membrane. The asym- diated cellular phenomena that occur fluid mosaic structure. It is likely that metry of the membrane introduces on a time scale of minutes or longer. mild proteolysis would preferentially another factor into the problem of Much more rapidly occurring phenom- release a small amount of glycopeptides translational diffusion of membrane ena, such as nerve impulse transmission, and other polar peptides from these components. In the experiments of Frye would find the mosaic structure to be a proteins because these are the most and Edidin (44) only those membrane static one, insofar as translational diffu- exposed portions of the integral pro- antigens exposed at the outer surface of sion of the membrane components is teins at the outer surface of the mem- the membrane were labeled by fluores- concerned. In order to illustrate the brane (Figs. 2 and 3). The remaining cent antibodies, and the conclusion that concepts involved, we discuss two spe- portions of these proteins may still these particular antigens were mobile in cific membrane phenomena. contain a large fraction of their original the plane of the membranes therefore, Malignant transformation of cells and chains after the limited strictly speaking, applies only to those the "exposure of cryptic sites." Normal proteolysis, but the release of some of components accessible at the outer sur- mammalian cells grown in the more polar structures would make face. Whether components confined to culture generally exhibit "contact in- the remaining portions more hydro- the inner surfaces also intermix and hibition"; that is, they divide until they phobic. As these more hydrophobic diffuse should be separately established. form a confluent monolayer and they glycoproteins diffused in the membrane, Thus, recent evidence obtained with then stop dividing. Cells that have be- they might then aggregate in the plane many experimental methods and differ- come transformed to by of the membrane. The result would be ent kinds of functional membrane sys- oncogenic viruses or by chemical car- a clustering of the agglutinin-binding tems is entirely consistent with the pre- cinogens lose the property of contact sites on the -treated cell sur- dictions of the fluid mosaic model of inhibition; that is, they overgrow the face, as compared to the normal un- membrane structure and provides strong monolayer. For some time, this experi- treated surface. Such clustering (with support for the model. It seems amply mental finding has been thought to re- no increase, or perhaps even a decrease justified, therefore, to speculate about flect the difference between the normal in the total numbers of sites because how a fluid mosaic structure might and the malignant states in vivo, and of ) could enhance the agglu- carry out various membrane functions, to be due to differences in the surface tination of the treated cells, as com- and to suggest specific mechanisms for properties of normal and malignant pared to that of normal cells, because various functions that can be subjected cells. Much excitement and investiga- it would increase the probability of to experimental tests. tive activity therefore attended the dem- agglutinin bridges forming between the onstration (53, 54) that malignant trans- surfaces of two cells. formation is closely correlated with a In malignant transformation, distinct The Fluid Mosaic Model and greatly increased capacity for the trans- chemical changes in the and formed cells to be of cell membrane Membrane Functions agglutinated by sev- the glycoproteins the eral saccharide-binding plant aggluti- are known to occur (56), and the en- The hypothesis that a membrane is nins. Furthermore, mild treatment of hanced agglutinability of the transform- an oriented, two-dimensional, viscous normal cells with proteolytic enzymes ed cells may be much more complicated solution of amphipathic proteins (or can render them also more readily ag- than is the case in the proteolysis of lipoproteins) and lipids in instantaneous glutinable by these protein agglutinins. normal cells. If, however, the two phe- thermodynamic equilibrum, leads to Burger (54) has suggested, therefore, nomena do have a basic feature in com- many specific predictions about the that the agglutinin-binding sites are pres- mon, it could be a similar clustering of mechanisms of membrane functions. ent on the membrane surfaces of nor- saccharide-binding sites on the trans- Rather than catalog a large number of mal cells but are "icryptic" (Fig. 7A) formed and the enzyme-treated normal these, we suggest some directions that (that is, they are shielded by some other cells. In malignant transformation, such such speculations may usefully take. membrane components from effectively clustering could be the result of the Among these problems are nerve im- participating in the agglutination proc- chemical changes in the membrane pulse transmission, transport through ess), and that proteolytic digestion of mentioned above; or some virus-induced membranes, and the effects of specific normal cells or the processes of malig- product (57) may be incorporated drugs and on membranes (1). nant transformation "exposes" these into the cell membrane and serve as a The fluidity of the mosaic structure, cryptic sites on the membrane surface. nucleus for the aggregation of the ag- which introduces a new factor into such In some cases, quantitative binding glutinin-binding glycoproteins within the speculations, is emphasized here. This studies have indeed indicated that no membrane. new factor may be stated in general significant change in the numbers of These suggestions can be tested ex- form as follows. The physical or chem- agglutinin-binding sites on the mem- perimentally by the use of ferritin-con- ical perturbation of a membrane may brane accompanies either mild pro- jugated agglutinins (33) as already dis- affect or alter a particular membrane teolysis of normal cells or malignant cussed (Fig. 6). The prediction is that component or set of components; a re- transformation (55). with normal cells subjected to mild distribution of membrane components An alternative explanation of these proteolysis, and also with malignant can then occur by translational diffu- phenomena (Fig. 7B), based on the transformed cells, the total number of sion through the viscous two-dimen- fluid mosaic model of membrane struc- ferritin-agglutinin particles specifically 728 SCIENCE, VOL. 175 bound to the outer surfaces of the cells and his co-workers (63) have pro- be quite different. The basic theory of might not be greatly differentfrom those posed an extension to membranes of Changeuex et al. (63) might still be of normal cells, but larger clusters of the Monod-Wyman-Changeuxallosteric formally applicable, but with impor- ferritin particles would be found. model of protein cooperative phenom- tant changes in physical significance.It Cooperative phenomena in mem- ena, using as a model of membrane is possible, for example, that a particu- branes. By a cooperative phenomenon structure an infinite two-dimensional lar integral protein can exist in either we mean an effect which is initiated aggregate of identical lipoprotein sub- of two conformational states, one of at one site on a complex structureand units [as, for example, the model de- which is favored by ligand binding; in transmittedto another remote site by scribed by Benson (20)]. In this theo- its normal unbound conformation the some structural coupling between the retical treatment,the individualsubunits integral protein is monomolecularly two sites. A number of important are capable of existing in either of two dispersed within the membrane, but in membrane phenomena may fall into conformational states, one of which the conformation promoted by ligand this category. However, before enum- has a much larger binding affinity for binding, its aggregation is thermody- erating them, we should first discrimi- a specific ligand than does the other. namically favored. The binding of a nate between two types of cooperative The binding of a single ligand molecule ligand molecule at one integral protein effects that may occur. These can be to any one subunit then triggers the site, followed by diffusion of the non- termed trans and cis. Trans effects refer cooperative conversion of many of the liganded protein molecules to it, might to cooperative (allosteric) changes that subunits to the ligand-bound confor- then lead to an aggregationand simul- have been postulatedto operate at some mation, in order to maximize the inter- taneous change in conformationof the localized region on the membrane sur- actions among the subunits. aggregated protein within the mem- face, to transmitan effect from one side This theory as presented relies on brane. This mechanism could result in of the membraneto the other (58). For the membranemodel used. If, however, a long-range cis-type cooperative phe- example,fan integral protein may exist in the membraneis not a two-dimensional nomenon, if the eventual aggregate the membraneas an aggregateof two (or aggregate of lipoprotein subunits, but size was very large and if its presence more) subunits,one of which is expc,sed is instead a fluid mosaic of proteins produced local perturbations in the to the aqueoussolution at the outer sur- and lipids, the physical situation would propertiesof the membrane. However, face of the membrane, and the other is exposed to the cytoplasm at the inner surface. The specific binding of a drug or molecule to the active site of the outward-orientedsubunit may induce a conformationalrearrangement within the aggregate,and therebychange some functional property of the aggre- A gate or of its inward-orientedsubunit. By cis effects, on the other hand, we refer to cooperative changes that may be producedover the entire membrane, or at least large areas of it, as a conse- quence of some event or events occur- ring at only one or a few localized points on the membrane surface. For example, the killing effects of certain bacteriocinson (59), the of the cortical granules of egg cells upon fertilization of eggs by (60), and the interaction of growth hormone with erythrocyte membranes (61) are cases which may involve transmissionand amplificationof local- B ized events over the entire surface of a membrane. These phenomena may not all occur by the same or related mechanisms,but in at least two experi- mental studies, that involving the inter- action of E1 with intact Esche- w v Fig. 7. Two differentmechanisms to explainthe findingsthat either malignantlytrans- richia coli cells (62), and that of human formed cells or normal cells that are subjectedto mild proteolysisbecome much growth hormone and isolated human more readilyagglutinable by severalplant agglutinins.(A) The mechanismof Burger membranes (61), there is (54): agglutinin-bindingsites that are presenton the surfacesof normalcells, but are erythrocyte are or the processesof malignant substantialevidence that long-rangecis- obstructed("cryptic sites"), exposedby proteolysis transformation.(B) The redistributionmechanism (see text): the agglutininsites on type cooperative effects intrinsic to the normalcell surfacesare largely monomolecularlydispersed in the fluid mosaic struc- membranesare involved. ture, but on proteolysisor malignanttransformation, they diffuse and aggregatein The question we now addressis, How clusters.The probabilityof agglutinationof two such modifiedcells is enhancedby sites. might such cis effects work? Changeux the clusteringof binding 729 18 FEBRUARY 1972 the transition would occur at a rate the membrane, an effect that may be ments with a wide variety of techniques and over a time period determined by involved in the phenomenon of anti- and several different membrane sys- the rate of diffusion of the molecules genic modulation (66). There are other tems are described, all of which are of the integral protein in the fluid specific examples as well. consistent with, and add much detail mosaic membrane. This time period is It may well be that a number of to, the fluid mosaic model. It therefore likely to be relativelylong, of the order critical metabolic functions performed seems appropriate to suggest possible of minutes (44), as already mentioned. by cell membranes may require the mechanisms for various membrane On the other hand, if cis-type cooper- translational mobility of some impor- functions and membrane-mediated ative effects occurred in a lipoprotein tant integral proteins. This could be phenomena in the light of the model. subunit model according to the mecha- the ultimate significance of the long- As examples, experimentally testable nism postulated by Changeux et al. standing observation (67) that the mechanisms are suggested for cell sur- (63), one would expect the coopera- membranelipids of poikilothermicorga- face changes in malignant transforma- tive change to be much faster. Con- nisms contain a larger fraction of un- tion, and for cooperative effects ex- formation changes in the soluble allo- saturated fatty acids the lower their hibited in the interactionsof membranes steric protein aspartyltranscarbamylase, temperature of growth. Appropriate with some specific ligands. for example, have half-times of the enzymes apparentlycarry out the nec- Note added in proof: Since this ar- order of 10 milliseconds (64). It is essary biochemical adjustment (68) ticle was written, we have obtained therefore of some interest that in the that keeps the membrane lipids in a electron microscopicevidence (69) that studies of the interaction of colicin fluid state at the particulartemperature the concanavalin A binding sites on E1 and E. coli the fluorescencechanges of growth; if these enzymes are not the membranes of SV40 virus-trans- that marked the apparent cis-type co- functional, for example, because of formed mouse (3T3 cells) operative transitions in the cell mem- mutations, the -to grow at are more clusteredthan the sites on the brane occurred over intervals of one the lower temperature (65)-must be membranesof normalcells, as predicted to several minutes (62). If this sug- suppliedwith the unsaturatedfatty acid by the hypothesis represented in Fig. gested mechanism for the colicin effect exogenously. While it has been sug- 7B. There has also appeareda study by is valid, one would predict that (i) gested before that the maintenance of Taylor et al. (70) showing the re- freeze-etching experiments on the coli- lipid fluidity may be important to markable effects produced on lympho- cin-treated bacteria (28) might reveal carrier mechanisms operating across a cytes by the addition of antibodies di- an aggregation of normally dispersed functional membrane,it is also possible rected to their surface immunoglobulin particles at the inner membrane face, that the real purpose of fluidity is to molecules. The antibodies induce a re- or (ii) changes in , permit some critical integral proteins distribution and of these such as would be produced by suitable to retain their translationalmobility in surface immunoglobulins,so that within changes in temperatureor by different the plane of the membrane, as an about 30 minutes at 37?C the surface compositions of membrane phospho- obligatory step in their function. immunoglobulinsare completely swept lipids (65), might markedly affect the out of the membrane.These effects do kinetics of the fluorescence changes not occur, however,if the bivalent anti- that are observed on addition of the Summary bodies are replaced by their univalent colicin to the bacteria. Fab fragments or if the antibody ex- In this discussionof membranefunc- A fluid mosaic model is presented periments are carried out at 0?C in- tions, some detailed mechanisms to for the gross organizationand structure stead of 37?C. These and related results account for two membranephenomena of the proteins and lipids of biological strongly indicate that the bivalent anti- have been presented. It may well turn membranes. The model is consistent bodies produce an aggregation of the out that these mechanisms are incor- with the restrictions imposed by ther- surface immunoglobulin molecules in rect. Our object has been not so much modynamics. In this model, the pro- the plane of the membrane,which can to argue for these specific mechanisms, teins that are integral to the membrane occur only if the immunoglobulinmole- as to illustrate that the fluid mosaic are a heterogeneous set of globular cules are free to diffuse in the mem- model of membranestructure can sug- molecules, each arrangedin an amphi- brane. This aggregationthen appearsto gest novel ways of thinking about pathic structure, that is, with the ionic trigger off the pinocytosis of the mem- membrane functions-ways that are and highly polar groups protruding brane components by some unknown amenable to experimental tests. Other from the membrane into the aqueous mechanism. Such membrane transfor- membrane phenomena may be influ- phase, and the nonpolar groups largely mations may be of crucial importance enced by similardiffusional mechanisms: buried in the hydrophobic interior of in the induction of an antibody re- for example, cell-cell and cell-sub- the membrane.These globularmolecules sponse to an antigen, as well as in strate interactions,where the apposition are partially embedded in a matrix of other processes of cell differentiation. of intense local electric fields to a cell phospholipid.The bulk of the phosplho- membrane may affect the distribution lipid is organized as a discontinuous, References and Notes of fluid bilayer, although a small fraction 1. S. J. Singer, in Structure and Function of electrically charged integral proteins Biological Membranes, L. I. Rothfield, Ed. within the membranes; or the specific of the lipid may interact specifically (Academic Press, New York, 1971), p. 145. of multivalent to cell with the membrane The fluid 2. M. Glaser, H. Simpkins, S. J. Singer, M. binding antibody proteins. Sheetz, S. I. Chan, Proc. Nat. Acad. Sci. U.S. surface antigens, where the simultane- mosaic structure is therefore formally 65, 721 (1970). 3. J. Lenard and S. J. Singer, ibid. 56, 1828 ous binding of one antibody molecule analogous to a two-dimensional ori- (1966). to several molecules of the antigen ented solution of integral proteins (or 4. D. F. H. Wallach and P. H. Zahler, ibid., p. induce of the dis- lipoproteins) in the viscous 1552. may rearrangements phospho- 5. W. Kauzmann, Advan. Protein Chem. 14, 1 tribution of the antigen in the plane of lipid bilayer solvent. Recent experi- (1959). 730 SCIENCE, VOL. 175 6. C. Tanford, J. Amer. Chem. Soc. 84, 4240 Branton [Biochim. Biophys, Acta 205, 125 50. G. L. Nicolson and S. J. Singer, in prepara- (1962). (1970)] find no evidence by freeze-etching for tion. 7. E. D. Korn, Annu. Rev. Biochem. 38, 263 the existence of the stalked-knobs on mito- 51. V. T. Marchesi and G. E. Palade, J. Cell (1969). chondrial inner membranes that are observed Biol. 35, 385 (1967). 8. V. T. Marchesi and E. Steers, Jr., Science with negatively stained preparations. 52. R. K. Kornberg and H. M. McConnell, Bio- 159, 203 (1968). 28. M. E. Bayer and C. C. Remsen, J. Bacteriol. chemistry 10, 1111 (1971). 9. S. H. Richardson, H. 0. Hultin, D. E. Green, 101, 304 (1970). 53. M. M. Burger and A. R. Goldberg, Proc. Nat. Proc. Nat. Acad. Sci. U.S. 50, 821 (1963). 29. P. Pinto da Silva and D. Branton, J. Cell Acad. Sci. U.S. 57, 359 (1967); M. Inbar and 10. M. Ashwell and T. S. Work, Annu. Rev. Bio- Biol. 45, 598 (1970); T. W. Tillack and V. T. L. Sachs, ibid. 63, 1418 (1969). chem. 39, 251 (1970). Marchesi, ibid., p. 649. 54. M. M. Burger, ibid. 62, 994 (1969). 11. D. Haldar, K. Freeman, T. S. Work, Nature 30. P. Pinto da Silva, S. D. Douglas, D. Branton, 55. B. Sela, H. Lis, N. Sharon, L. Sachs, Bio- 211, 9 (1966); E. D. Kiehn and J. J. Holland, Abstracts of the 10th Meeting of the American chim. Biophys. Acta, in press; B. Ozanne and Proc. Nat. Acad. Sci. U.S. 61, 1370 (1968); Society for , San Diego, Calif., J. Sambrook, Nature 232, 156 (1971). D. E. Green, N. F. Haard, G. Lenaz, H. I. November 1970 (1970), p. 159; T. W. Tillack, 56. S. Hakamori and W. T. Murakami, Proc. Nat. Silman, ibid. 60, 277 (1968); S. A. Rosenberg R. E. Scott, V. T. Marchesi, ibid., p. 213. Acad. Sci. U.S. 59, 254 (1968); P. T. Mora, and G. Guidotti, in Red Cell Membrane, G. 31. T. L. Steck, G. Fairbanks, D. F. H. Wallach, R. 0. Brady, R. M. Bradley, V. M. McFar- A. Jamieson and T. J. Greenwalt, Eds. (Lip- 10, 2617 (1971). land, ibid. 63, 1290 (1969); C. A. Buck, M. pincott, Philadelphia, 1969), p. 93; J. Lenard, 32. M. S. Bretscher, Nature 231, 225 (1971); J. C. Glick, L. Warren, Biochemistry 9, 4567 Biochemistry 9, 1129 (1970). Mol. Biol. 59, 351 (1971). (1970); H. C. Wu, E. Meezan, P. H. Black, 12. J. K. Blasie, C. R. Worthington, M. M. 33. G. L. Nicolson and S. J. Singer, Proc. Nat. P. W. Robbins, ibid. 8, 2509 (1969). Dewey, J. Mol. Biol. 39, 407 (1969). Acad. Sci. U.S. 68, 942 (1971). 57. T. L. Benjamin and M. M. Burger, Proc. Nat. 13. D. Bownds and A. C. Gaide-Huguenin, Na- 34. G. L. Nicolson, S. P. Masouredis, S. J. Singer, Acad. Sci. U.S. 67, 929 (1970). ture 225, 870 (1970). ibid., p. 1416. 58. T. R. Podleski and J.-P. Changeux, in Funda- 14. B. Ke, Arch. Biochem. Biophys. 112, 554 35. G. L. Nicolson, R. Hyman, S. J. Singer, J. mental Concepts in Drug-Receptor Interac- (1965). Cell Biol. 50, 905 (1971). tions, D. J. Triggle, J. F. Danielli, J. F. 15. M. Glaser and S. J. Singer, Biochemistry 10, 36. R. E. Lee and J. D. Feldman, ibid. 23, 396 Moran, Eds. (Academic Press, New York, 1780 (1971). (1964). 1969), p. 93. 16. J. M. Steim, M. E. Tourtellotte, J. C. Reinert, 37. F. A. Green, Immunochemistry 4, 247 (1967); 59. M. Nomura, Proc. Nat. Acdd. Sci. U.S. 52, R. N. McElhaney, R. L. Rader, Proc. Nat. J. Biol. Chem. 243, 5519 (1968). 1514 (1964). Acad. Sci. U.S. 63, 104 (1969). 38. W. C. Davis and L. Silverman, Transplanta- 60. D. Epel, B. C. Pressman, S. Elsaesser, A. M. 17. D. L. Melchoir, H. J. Morowitz, J. M. Sturte- tion 6, 536 (1968); T. Aoki, U. Hammerling, Weaver, in The : Gene-Enzyme In- vant, T. Y. Tsong, Biochim. Biophys. Acta E. de Harven, E. A. Boyse, L. J. Old, J. Exp. teractions, G. N. Padilla, G. L. Whitson, I. 219, 114 (1970). Med. 130, 979 (1969). L. Camerson, Eds. (Academic Press, New 18. D. M. Engelman, J. Mol. Biol. 47, 115 (1970); 39. A. Shimada and S. G. Nathenson, Biochemis- York, 1969), p. 279. M. H. F. Wilkins, A. E. Blaurock, D. M. try 8, 4048 (1969); T. Muramatsu and S. G. 61. M. Sonenberg, Biochem. Biophys. Res. Com- Engelman, Nature 230, 72 (1971). Nathenson, ibid. 9, 4875 (1970). mun. 36, 450 (1969); Proc. Nat. Acad. Sci. 19. M. E. Tourtellotte, D. Branton, A. Keith, 40. J. K. Blasie, M. M. Dewey, A. E. Blaurock, U.S. 68, 1051 (1971). Proc. Nat. Acad. Sci. U.S. 66, 909 (1970). C. R. Worthington, J. Mol. Biol. 14, 143 (1965). 62. W. A. Cramer and S. K. Phillips, J. Bacteriol. 20. A. A. Benson, J. Amer. Chem. Soc. 43, 41. M. M. Dewey, P. K. Davis, J. K. Blasie, L. 104, 819 (1970). 265 (1966). Barr, ibid. 39, 395 (1969). 63. J. P. Changeux, J. Thi6ry, Y. Tung, C. Kittel, 21. D. Triggle, Recent Progr. Surface Sci. 3, 273 42. J. K. Blasie and C. R. Worthington, ibid., Proc. Nat. Acad. Sci. U.S. 57, 335 (1967). (1970). p. 417. 64. J. Eckfeldt, G. G. Hammes, S. C. Mohr, C. 22. H. U. Schairer and P. Overath, J. Mol. Biol. 43. R. C. Warren and R. M. Hicks, Nature 227, W. Wu, Biochemistry 9, 3353 (1970). 44, 209 (1969). 280 (1970). 65. D. F. Silbert and P. R. Vagelos, Proc. Nat. 23. G. Wilson and C. F. Fox, ibid. 55, 49 (1971). 44. C. D. Frye and M. Edidin, J. Cell Sci. 7, 313 Acad. Sci. U.S. 58, 1579 (1967). 24. J. Lenard and S. J. Singer, Science 159, 738 (1970). 66. E. A. Boyse and L. J. Old, Annu. Rev. Genet. (1968); G. Vanderkooi and D. E. Green, 45. L. Mindich, J. Mol. Biol, 49, 415, 433 (1971). 3, 269 (1969). Proc. Nat. Acad. Sci. U.S. 66, 615 (1970). 46. R. D. Poretz and I. J. Goldstein, Biochemis- 67. E. F. Terroine, C. Hofferer, P. Roehrig, Bull. 25. W. L. Hubbell and H M. McConnell, Proc. try 9, 2870 (1970). Soc. Chim. Biol. 12, 657 (1930); G. Frankel Nat. Acad. Sci. U.S. 61, 12 (1968); A. D. 47. R. G. Drysdale, P. R. Herrick, D. Franks, and A. S. Hopf, Biochem. J. 34, 1085 (1940). Kieth, A. S. Waggoner, 0. H. Griffith, ibid., Vox Sang. 15, 194 (1968). 68. M. Sinensky, J. Bacteriol. 106, 449 (1971). p. 819. 48. G. L. Nicolson and S. J. Singer, in prepara- 69. G. L. Nicolson, Nature 233, 244 (1971). 26. R. L. Steere, J. Biophys. Biochem. Cytol. 3, tion. 70. R. B. Taylor, W. P. H. Duffus, M. C. Raff, 45 (1957); H. Moor, K. Miihlethaler, H. Wald- 49. E. H. Eylar, M. A. Madoff, 0. V. Brody, J. S. dePetris, ibid., p. 225. ner, A. Frey-Wyssling, ibid. 10, 1 (1961). L. Oncley, J. Biol. Chem. 237, 1962 (1962); 71. The original studies reported in this article 27. D. Branton, Annu. Rev. Plant Physiol. 20, 209 E. L. Benedetti and P. Emmelot, J. Cell Sci. were supported by grant GM 15971 from the (1969). J. M. Wrigglesworth, L. Packer, D. 2, 499 (1967). National Institutes of Health (to S.J.S.).

NEWS AND COMMENT study also indicates that the Soviet Union is almost certainlypressing ahead -cautiously but intently-with a manned lunar programthat may be ex- pected to put cosmonauts on the moon The Soviet Space Program: Effort in the mid-1970's and possibly as early as 1973. A related conclusion, perhaps Said to Peak U.S. Level the most surprising of the 670-page Surpass study, is that the Russians may end up spending the equivalent of $49 billion to land men on the moon, far more than the cost of the Apollo program. A new and authoritativestudy of the deaths of three cosmonauts last year. Whether or not the Soviets actually Soviet space program indicates that, The study,* produced for the Senate carry through with their evident inten- while American space efforts continue Committee on Aeronautical and Space tions, the study goes on, "it is not pos- winding down toward the last Apollo Sciences by analysts in three divisions sible to establishthat the Russians have flight this year, the overall Soviet space of the Library of Congress, concludes invested smaller total resources in lunar programremains "a strong and growing that the current level of Soviet space explorationthan the United States"even enterprise,"its ambitionsunhindered by activity exceeds that of the United though the Soviet effort "has not pro- budgetary strain and undimmed by the States at its peak in 1966. The space duced the visible result in this regard * " Soviet Space Programs, 1966-70" Report of the Committee on Aeronautical and Space Sciences, which the United States has achieved." prepared by the Science Policy Research Division, Foreign Affairs Division, and the European Law These and other findingsstand in direct Division, Library of Congress; available from the Government Printing Office, Washington, D.C. 20402, $3; stock number 5271-0263. contradiction of assertions by Soviet 18 FEBRUARY 1972 731