During the last 10–15 years, a large number of proteinshave been recognized as playing central roles duringendo-exocytosis phenomena and more generally duringvesicle traffic: clathrin, NSF, ARF, dynamin, caveolin,annexins and GTP-binding proteins, have been implicatedin the processes of membrane vesicularization (also calledbudding) or during fission, targeting and fusion ofvesicles. A few reconstitution experiments with a limitednumber of proteins have reproduced, if not quantitatively,at least qualitatively some of the important steps [1, 2] andfrom these observations, generalized models were pro-posed. However, the main emphasis of earlier research hasbeen to show which proteins are required, rather than todissect the molecular mechanisms involved. Membranefolding which is a prerequisite for vesicle formation, isoften assumed to be explained by the binding of clathrinor of coat proteins. Yet, endocytosis can exist withoutclathrin or without clathrin polymerization [3, 4]. Further-more, increasing the amount of polymerized clathrin doesnot systematically increase the number of coated pits atthe plasma membrane [5]. In fact, it has not been proven
yet whether clathrin polymerizes and then pinches off themembrane to form the buds or if polymerization takesplace around a pre-formed bud. An alternative explanationproposed for membrane budding during endocytosis is theformation of local invaginations named caveola formed byself association of the hydrophobic protein caveolin withspecific lipids [6]. To what extent is ATP necessary forbudding is still debated [7]. In summary, the physicalorigin of the local membrane curvature is still unexplainedin spite of the correlation between the presence of specificproteins and the formation of coated pits. Concerningexocytosis, it is believed that the translocation of lipidswhich a priori is necessary after the fusion of micro-vesicles to the relatively flat plasma membrane can takeplace spontaneously, yet it is known that lipid flip-floprequires several hours which is quite incompatible withthe time scale of a continuous endocytotic process.
In general, the role of lipids has been largely over-looked in these cell biology investigations. Only recently,in experiments with reconstituted vesicles to which coatproteins were attached, has the importance of lipids beenrealized. In fact, even pure liposomes enable one to mimicsome fundamental steps of biological membrane traffic.To those who believe that the representation of a biologi-cal membrane by a mere lipid bilayer is irrelevant, itshould be recalled that membrane budding, fission andfusion refer primarily to the status of the lipid bilayer.Thus, a detailed analysis of the physical constraintsassociated with folding and fusion of a pure lipid bilayeris of primary importance. Of course, investigating lipo-some behavior would be sterile for biology if one does not
* Correspondence and reprints: [email protected]: PC, phosphatidylcholine; PS, phosphatidylserine;PE, phosphatidylethanolamine; L-PC, lyso-phosphatidylcholine;PG, phosphatidylglycerol; GUV, giant unilamellar vesicle; LUV,large unilamellar vesicle; SUV, sonicated unilamellar vesicle;NMR, nuclear magnetic resonance; MDR, multi drug resistanceprotein
keep in mind the numerous specificities of biologicalmembranes. Not only the lipid composition of all eukar-yotic membranes is complicated, i.e., involves mixtures ofmany different lipids, the proportion of which seems to bestrictly regulated, but in addition the plasma membrane ofeukaryotic cells has an asymmetrical transmembrane dis-tribution of phospholipids which is difficult, if not impos-sible, to reproduce with model membrane [8]. Anotherimportant characteristic of in vivo vesicularization is theactual size of the vesicles involved in membrane traffic.Sufficient attention has not been paid to this point inprevious investigations on shape change of liposomes byphysicists attempting to mimic endo- and/or exo-cytosis ingiant vesicles.
The major issue that I will address in this article is thefollowing: is membrane bending and unfolding duringendo-exocytosis caused by the transmembrane redistribu-tion of lipids by the aminophospholipid translocase whichis a ubiquitous lipid transporter present in the plasmamembrane of eukaryotic cells? I shall first try to shedsome light on physical constraints which cannot beignored when trying to understand how membranes in-vaginate. In view of our knowledge on lipid-proteininteractions this will bring up some suggestions on pos-sible mechanisms involved at the early stage of endocy-tosis and the late stage of exocytosis, in particular theputative role of ‘fl ippases’ and ‘scramblases’ . The possibleinvolvement of the aminophospholipid translocase waspresented several years ago by myself as a workinghypothesis [9, 10]. Recent studies have given experimen-tal support to this idea [11–14].
2. Membrane budding
Let us consider first a protein-free giant unilamellarvesicle containing one or several types of phospholipids.Unlike a soap bubble, the surface tension of a liposome inthe absence of osmotic pressure, is very low and themembrane undergoes visible surface undulations. How-ever, it would be an erroneous conclusion to infer that abilayer is easy to deform. The budding of a synaptic-likemicrovesicle out of the surface of a giant liposomeimposes locally a high curvature which for a bilayer mustbe associated with a difference in the area of the inner andouter monolayers. If lipids were free to diffuse from oneside of a membrane to the other, invaginations andbudding would happen as a manifestation of thermalfluctuations as it happens with surfactant films. But in alipid bilayer, in general it costs a lot of energy for aphospholipid to traverse the membrane: t1/2 of spontane-ous phosphatydylcholine flip-flop is of the order of severalhours or even days at physiological temperatures [15].Furthermore, the surface compressibility of a lipid mono-layer is low, so thermal fluctuations only give rise tosurface undulations.
In a biological membrane, local deformations can beachieved by applying external forces for example by thecontraction of cytoskeleton proteins. In a liposome aprotrusion forming a long tether can be formed by suckingthe membrane with a micropipette or by pulling it withoptical tweezers. Alternatively, the shape can be changedin a more subtle manner by progressive modification ofthe ratio between inner and outer areas: Ai and Ao. Forexample, if the two opposing monolayers react differentlyto their environment or are selectively modified by addi-tion or depletion of lipids, they eventually have differentareas (∆A = Ai-Ao 7 0), then within the framework of theso-called bilayer couple hypothesis [16], the membranebends.
As indicated originally by Helfrich, the bending energyof a liposome can be written in the following way [17]:
Ec =kc
2 �A� C1 + C2 − 2 C0 �
2 dA
Here, the Gaussian curvature term is omitted. As long asvesicle fission (or fusion) has not taken place, we dealwith vesicles of spherical topology and the Gaussiancurvature term can be omitted. kc has the dimension of anenergy and is of the order of a few times the thermalenergy (kBT). The integral extends over the whole area Aof the lipid vesicle. C1 and C2 are the local curvatures (seefigure 1). In practice for a closed vesicle the two leafletshave a slightly different area. Under the condition that thetwo leaflets are everywhere separated by the same spacingh which is of the order of 6 nm, ∆A is related to the localcurvature by the following relation which is correct toorder h/R [18]:
DA = h � dA� C1 + C2 �
C0 is the spontaneous curvatjure that can be defined as thecurvature that the membrane would take in a relaxed state,i.e., not constrained by the necessity to form a closedliposome in order to avoid exposing the hydrophobic lipidchains to water. C0 7 0 can be due to an asymmetry in thelipid composition or lipid environment or simply to adifference in the number of lipids present at each interface.In Helfrich’s formulation C0 includes both the local andnon-local tendency to bend. A more explicit formalismwas proposed recently in the laboratory of Wortis [18]. Inthis model, the elastic energy has two separate terms,hence the difference in area of the two leaflets is ac-counted for by a term corresponding to a new elasticstretching energy:
Ec =kc
2 �A
� C1 + C2 − 2 C0 �2 dA + j� p/2 Ah2
� � DA − DAo �2
∆A is the actual difference in area, whereas ∆Ao is thedifference in area that would exist in a relaxed state. j is
498 Devaux
a new elastic modulus having also the dimension of anenergy. Then C0, the spontaneous curvature, reflects solelythe local curvature constraints due for example to localbinding of proteins or ions or to the chemical nature of thephospholipids which are usually asymmetrically distri-buted in a biological membrane [8, 10]. In principle, theelastic modulus κ is different from kc. However, it has thesame order of magnitude. For example, in the case ofhomogeneous phosphatidylcholine vesicles Miao et al.[18] have reported that the ratio α = j/kc ≈ 1. Thus, for the
qualitative approach of the present review, I shall continueto use Helfrich’s formulation throughout, and considerthat C0 reflects either local spontaneous curvature or thearea difference term. For a rigorous and more quantitativedevelopment of this section the separation between localand non-local spontaneous curvature would be necessary.
If a liposome is formed by gentle swelling of a lipidfilm in water, in general both leaflets are identical incomposition and density of lipids, hence: C0 ≈ 0. How-ever, the necessity to close the bilayer in order to avoidexposing hydrophobic surfaces forces the bilayer to bendand the total energy is minimize for C1 and C2 ≈ 0, i.e., foran average vesicle radius, < R >, very large. In practice,giant vesicles are formed with a radius of several microns,that is exceeding by far the thickness of the bilayer andcorresponding to a low average curvature.
Theoreticians have predicted shapes of liposomes fordifferent values of C0 by minimization of Helfrich’sformula or of formula derived from this formula [18–21].Phase diagrams of vesicle shapes were drawn from thesecomputations and typical shapes corresponding to thebudding of a small vesicle from a ‘mother vesicle’ oflarger size can be recognized. Figure 2 shows some
Figure 2. Schematic ‘phase diagram’ of vesicle shapes. This‘phase’ diagram shows the shapes of lowest bending energy forgiven reduced volume v and equilibrium differential area ∆ao.∆ao in this diagram is a dimension-less or scaled area differencewhich is effectively proportional to ∆Ao defined in the text. Withthe exception of the shapes indicated with a dashed contour, theshapes calculated have an axis of rotational symmetry along thevertical (reproduced from [32]).
Figure 1. Principal radii of curvature of a bilayer of thickness h.Locally any continuous surface can be approximated by anhyperboloid (or a paraboloid) which can be written:
Z = − 12� x2
R1+ y2
R2�. z is the direction of the normal to the
surface; the axes x and y are chosen so as to obtain thesymmetrical quadratic form of the surface equation. This defines
the two principal axes. By definition C1 = 1R1
and C2 = 1R2
are
the radii of curvature: they correspond to local parameters. Seethe text for the definition of the spontaneous curvature C0.
Is lipid translocation involved during endo- and exocytosis? 499
predicted shapes. Such shape changes in giant vesicleswere observed by several groups who found ways tomanipulate ∆A.
The contraction or expansion of one of the two leafletsof a liposome or of a biological membrane, which I shallcall a C0 modification, can be achieved by chemicalmodifications in situ resulting, for example, from phos-pholipase or sphingomyelinase degradation [7, 22–24].Phosphorylation of phosphonositides on the cytosolicinterface of an erythrocyte has been reported to beinvolved in shape changes in the case of red cells [25]. C0modification can be triggered by ions and/or proteinsinteracting with specific lipids of one of the leaflets,causing the formation of rigid domains or ‘ rafts’ with acondensed area and therefore leading to asymmetricalmembranes [6].
Alternatively, C0 can be modified by the insertion ordepletion of lipids from one leaflet. A non-physiologicalway to enrich the lipid composition of one bilayer leafletconsist in adding amphiphilic molecules such as lyso-phosphatidylcholine (L-PC) to pre-formed giant lipo-somes or to erythrocytes. L-PC molecules penetrate theouter leaflet where they remain and expand selectively theexternal surface since the flip-flop rate of L-PC is ex-tremely slow even in phosphatidylcholine liposomes [26].In vivo a net translocation of phospholipids can beachieved by the ‘phospholipid flippases’ which are able tocatalyze the translocation of phospholipids or to transportphospholipids at the expenses of ATP consumption.Modulation of the transmembrane asymmetry of certainlipids, such as phosphatidic acid or phosphatidylglycerol,can be achieved in liposomes if a transmembrane pHgradient is applied [27]. These various techniques whichenables one to modulate the spontaneous curvature of amembrane by the formation of asymmetrical bilayersinduce membrane invaginations.
Addition of less than 1% of lipids to the external leafletsuffices to modify a GUV with a discoid or obloid shapeinto a eight-shape vesicle formed by a small sphericalvesicle connected to a larger one; similarly a very smalllipid transfer of lipids from the inner to the outer leaflettriggers the formation of a budded vesicle with a typicaldiameter of the order of a few µm (see figure 3). In someinstances the small vesicle is separated from the largervesicle by a thin tether difficult to detect unless fluorescentlipids have been incorporated [28]. But in general, thesingle budded vesicle does not shed off. If more than 1%of L-PC is added, the overall shape of the GUV can takemore complicated forms with several connected spheres(figure 4) . If the external leaflet of a liposome is depletedof a fraction of its lipids, for example by removing L-PCfrom the outer leaflet with bovine serum albumin, a singleinvagination is formed which resembles membrane in-vaginations during endocytosis [29]. In Sackmann’s labo-ratory the same sequences of shape change were observedwith GUVs when the temperature was varied, the inter-
pretation being that even a very small differential thermalexpansion of the two monolayers can induce a change ofC0 [30, 31].
3. The scaling effect
How much asymmetrical has the membrane to be inorder to induce a detectable shape change or moreprecisely to generate budding of a small lipid vesicle outof a larger one that I shall call the ‘mother vesicle’? Inother words, if Co modification is caused by a transfer oflipids, what proportion of lipids has to be transported fromone leaflet to the other to obtain vesicularization?
The result from quantitative experiments with GUVs,discussed in the above paragraphs, is rather astonishing:0.1% of lipids in excess on the external leaflet suffice tomodify a GUV with a typical size of 20–50 µm. Theore-ticians had even predicted that 0.01% asymmetry wouldbe sufficient [20]. Simple geometrical arguments allowone to predict a scaling effect: namely the fraction oflipids that has to be reoriented or added selectivelydepends on the vesicle size. The scaling parameter isessentially the ratio h/Ro where h is the thickness of themembrane and Ro is the radius of the vesicle. Obviously,if the thickness h is close to the radius of the vesicle Ro,then membrane folding is difficult. If N is the averagenumber of lipids per monolayer, the asymmetry in lipiddistribution between both leaflets (δ N/N) must be of the
Figure 3. Shape transformation of a unilamellar giant vesiclecontaining egg lecithin and egg phosphatidylglycerol with amolar ratio of 99:1. The shape change is triggered by raising theexternal pH from 6 to 9. The time scale for total shapetransformation is approximately 10 s. The bar corresponds to10 µm. Vesicles are observed with an optical microscope withNomarski configuration (reproduced from [28]).
500 Devaux
order of h/Ro for ‘vesicularization’ . In the case of a GUVwith a typical diameter of about 50 µm, the experimentsshow that the ‘small’ vesicles which are generated by lipidasymmetry have a size of 5–10 µm. Since the bilayerthickness is around 6 nm, a very small asymmetry, below1%, is sufficient in principle. In the case of a 200 nm LUV
and if one neglects temporarily the lateral compressibility,pure geometrical considerations show that the excess oflipids on the outer layer of a budded vesicle must bebetween 5 and 10%, meaning that a proportion of phos-pholipids of the order of 5% may have to be translocated.In the latter case, a significant surface tension is generatedby the mismatch between both layers and the processrequires more energy than for giant vesicles.
Indeed, surface asymmetry not only leads to bendingbut also to surface tension T. By definition: T = Ka, whereK is the lateral compressibility coefficient and α thesurface strain which is the relative area variation of oneleaflet due to lateral compressibility. In the initial state, thesurface tension of both leaflets is practically zero. If oneadds new molecules to the outer monolayer, because of thecoupling with the inner monolayer which resists to thisexpansion, a surface tension is generated. Eventually atorque is exerted on the membrane. But it is important tonote that bending does not relax completely the tension.Because of the conditions of closure (fixed volume), themembrane cannot bend so as to completely relax thetension. In fact the curvature adopted eventually by themembrane equilibrates the difference in surface con-straints between the two leaflets, so that the torque iscanceled but not the surface tension.
In practice, the addition of 0.1% lipids on the externalleaflet of a GUV generates a negligible surface tension.However, an asymmetry of 5–10% between the twoleaflets of a GUV or of a LUV generates a significantsurface tension. One manifestation of the tension is theinhibition of surface undulations which are partially ortotally inhibited by addition of lyso-PC to their externalsurface of vesicles [29]. Recently the symmetrical surfacetension generated in liposomes by the asymmetrical addi-tion of L-PC was demonstrated by Traïkia using NMR31P-chemical shift as a way to detect the molecularpacking at the level of the phospholipid head groups ofeach monolayer in an experiment involving magic anglespinning of lipid vesicles [26]. The surface tension whichis equivalent to an increased lipid packing, manifests itselfalso by an increased resistance to transmembrane lipiddiffusion. Thus, while the formation of a 5 µm vesiclebudding out of a 50 µm giant unilamellar vesicle can beachieved in a few minutes at room temperature when themembrane (containing a PC/PG mixture) is submitted to apH gradient [29], the same overall shape with 100 nmLUVs required 60 mn of incubation at 60 °C [32]. Thelatter observation reported by the Cullis’group is inaccordance with previously published theoretical predic-tions [33, 34].
Finally, I would like to emphasize as a conclusion ofthis section that the formation of one or several mi-crovesicles with a diameter of about 20 nm is alwaysaccompanied by a surface tension. Because of the conti-nuity between a budded vesicle and the rest of theliposome this tension must be the same in the whole
Figure 4. Shape transformation of an egg phosphatidylcholineunilamellar vesicle containing fluorescent dextran (0.1 mM)after L-PC addition in proximity of the vesicle with a micropi-pette. A. t = 0. B. t = 4 min. C. t = 7 min. The bar correspondsto 10 µm (reproduced from [28]).
Is lipid translocation involved during endo- and exocytosis? 501
vesicle membrane, even in a GUV. In other words,although local fluctuations of the surface density in a GUVcould give birth to a microvesicle of the size encounteredin vivo during endocytosis or vesicle shedding, thisvesicle would promptly disappear because of the instabil-ity of a local surface tension. Thus a stable budded vesicleconnected to the rest of the plasma membrane is onlypossible if the whole plasma membrane is constrained.
4. Is microvesicle formation in liposomes and redcells relevant to endocytosis?
Typical diameters of the vesicles involved inendocytosis-exocytosis, and more generally vesicle trafficwithin eukaryotic cells, are around 200 nm. They resultfrom local membrane curvature, either of the plasmamembrane itself or of organelles such as the ER or theGolgi system, the typical size of which is in the range ofone or several microns. Although lipid vesicle experi-ments demonstrate a close relationship between vesicleshape and lipid transmembrane distribution and althoughthey show that endocytic-like invaginations may be trig-gered in GUVs by the redistribution of a small proportionof lipids, they do not satisfactorily mimic endocytosis.Indeed, endocytosis implies the formation of one orseveral microvesicles from the invaginations of a cellmembrane which has a typical radius of several µm. Thisis not what is observed in giant unilamellar vesicles whenfor example 1% of L-PC is added externally. We havementioned above that the formation of very small spheri-cal vesicles requires a lipid asymmetry of the order of5–10% and because these vesicles (or invaginations)remain at least temporarily connected to the rest of themembrane, the same lipid asymmetry must exist all alongthe cell surface and will cause an important surfacetension. The so-called large unilamellar vesicles (LUVs)with a diameter of 100–200 nm can sustain surface tensionwithout collapse and form budded vesicles or protrusionswith a diameter of about 50 nm or less. On the other hand,a giant vesicle of the size of an eukaryotic cell will notform at its surface a series of small vesicles with adiameter which is ≈ 10–3 times the average diameter ofthis ‘mother vesicle’ without a complete shape changeunless some scaffolding, analogous to the cytsoskeletonmeshwork in a biological cell prevents the transformationof this large vesicle. Figure 4 shows that if ∆A isincreased by the addition of a large amount of L-PC intothe external leaflet of a GUV, the shape undergoes adramatic change that does not correspond to a mere‘decoration’ of a large obloid vesicle by a series ofmicrovesicles. Of course, if the giant vesicle is initially aperfect sphere and if the internal volume remains approxi-mately constant, there is no possibility of any other shape.In the latter case, the additional surface on the outer layerwill generate a tension that will inhibit surface undulations
and eventually the surface of the sphere may becomecovered by microvesicles.
Figure 5 shows an experiment where a large number ofrelatively small vesicles appear at the external surface ofa giant vesicle. This figure was obtained with multilamel-lar vesicles (MLVs) on the surface of which a large excessof L-PC was added. The inner membranes probablyprevent the liposome from an overall deformation orcollapse. The large excess of lipids on the external leafletof the external membrane can only form series of smallvesicles which eventually shed off, thereby pealing theexternal bilayer of this multilamellar vesicle. But even inthis example the size of the small vesicles (≈ 1 µm) wasmuch larger than the size of endo-exo-cytic vesicles. Theabove example is reminiscent of the situation obtainedwith red blood cells to which a large excess of L-PC hasbeen added. In the latter case, very small membraneprotrusions cover the cell surface which seems to shrinkand form a uniform sphere (spherocyte) if viewed with anoptical microscope. In reality, the erythrocyte surface iscovered by small vesicles with a radius below resolutionof the optical microscope. The cell resists because of thecytoskeleton and eventually lyses if more L-PC is added(> 5%).
Figure 6a-f shows in an (over)simplified way the shapetransformations that can undergo an obloid GUV(figure 6a), by adding to the external surface moleculessuch as L-PC [28, 29] or by increasing area to volumeratio by heating [30, 31] Shapes in figure 6b, c are thecanonical axial symmetrical shapes calculated by theore-ticians (see figure 2). The radius R of the spherical‘budded’ vesicle(s) is large. Such transformation does notresemble in vivo budding. Shapes in figure 6d, e corre-spond to the formation of relatively small vesicles. Theyrequire first a transformation of the ‘mother vesicle’ into asphere. Again, overall this is not what is seen with realcells. In the case of figure 6f, which represents a multila-mellar vesicle (MLV), the obloid shape is maintained bythe inner bilayers. In vivo the cell shape is maintained bythe cytoskeleton and by the presence of organelles.
I infer that the activity of a lipid pump transportinglipids from the outer monolayer to the inner monolayer (ora selective fraction of the outer leaflet lipids) can induceinvaginations of the plasma membrane and form en-docytic vesicles of the size observed in vivo, provided thecell surface is reinforced by a cytoskeleton. This cyto-skeleton and tubulin meshwork is represented schemati-cally in figure 6g,h
5. Experimental observations in favor of the role oflipid translocation during endocytosis
The aminophospholipid translocase is present in theplasma membrane of all animal cells [10, 35–37]; recentresults suggest that it also exists in the plasma membrane
502 Devaux
Figure 5. Shape transformation observed by phase contrast microscopy of a multilamellar egg phosphatidylcholine vesicle a afteraddition of L-PC with a micropipette. A. t = 0. B. t = 8 s. C. t = 30 s. D. t = 1 min. E. 2 min. The bar corresponds to 20 µm. Notethat the scale is different in E (reproduced from [29]).
Is lipid translocation involved during endo- and exocytosis? 503
of plant cells [38]. This protein transports phosphati-dylserine (PS) and to a lesser extent phosphatidylethanol-amine (PE) from the outer to the inner leaflet where theseaminophospholipids are predominantly located. CytosolicATP is necessary and must be hydrolyzed for this transportto take place. By controlling the level of cytosolic ATP,which in vivo is normally of the order of 2–3 mM, it ispossible to modify the equilibrium distribution of the
aminophospholipids and in turn to modify erythrocyteshapes by shifting a fraction of the lipids from the inner tothe outer leaflet or conversely. Indeed, unpaired lipidtransport is accompanied by echinocyte formation whileartificial ATP enrichment leads to stomatocytes. When thelevel of ATP is of the order of 5–6 mM endocytic vesiclesform at the cytosolic interface of erythrocytes. They arecharacterized by the formation of a few relatively large
Figure 6. Schematic representa-tion of examples of shape transfor-mations of on obloid giant vesiclewhen a difference in area ∆A be-tween inner and outer leaflet isimposed. When the external area isincreased by 5–10% very smallmicrovesicles will bud at the sur-face of the ‘mother vesicle’ only ifthe latter has a spherical shape (d,e) or if the overall shape of the‘mother vesicle’ is protected byinner lipid vesicles (f) or by acytoskeleton meshwork (g, h). h.The external surface is depleted.See text for more comments.
504 Devaux
vesicles and can be quantified by monitoring the acetyl-choline esterase activity which normally takes place onthe outer surface of red cells [12, 39]. The decrease inesterase activity is an indication of the amount of surface
becoming non-accessible. If phosphatidylcholine (PC) isadded to the outer surface of red cells this vesicularizationprocess is affected because the plasma membrane wouldrather tend to fold in the opposite direction (see figure 7
Figure 7. Modulation of endocytosis activity in resealederythrocyte ghosts as determined by the acetycholinesteraseactivity. The reduction of the esterase activity which isnormally exposed on the external surface of the plasmamembrane is an indication of the formation of endocyticvesicles. A. Effect of ATP concentration in the resealedghosts. ATP stimulates the aminophospholipid translocase. B.Inhibition of endocytosis which normally takes place with5 mM ATP by addition of phosphatidylcholine to the outerleaflet. C. Inhibition followed by stimulation after addition ofphosphatidylserine. This lipid is initially incorporated in theouter leaflet and therefore slows down the inward folding ofthe plasma membrane. The translocation to towards the innerleaflet by the aminophospholipid translocase reverses thesituation and favors membrane invagination. The exogenouslipids added have a short � chain which permits their rapidincorporation (from [12]).
Is lipid translocation involved during endo- and exocytosis? 505
and [12]). Actually, the difference in the shape changescenarios which are triggered by the addition of variousphospholipids to the external surface of erythrocytes werenoticed already in 1984 by Seigneuret and Devaux [40]and used later by Daleke and Huestis to monitor theactivity of the aminophospholipid translocase [41].
Erythrocytes cannot be considered as the best system tostudy endocytosis since they do not naturally endocytose.It is interesting to point out that concomitantly theaminophospholipid translocase activity (or ‘fl ippase’ ac-tivity) is rather weak in erythrocytes compared to itsactivity in cells having a high endocytic activity. Cribierand collaborators have compared the translocase activityin some of the cells of the red cell lineage: erythroblasts,reticulocytes and erythrocytes and found that the highestactivity was in the cells provided with the highest en-docytic activity, namely in K562 cells which are trans-formed cells derived from human erythroblasts [11].Similarly, Pomorski et al. [42] have compared the trans-locase activity in human erythrocytes and in humanfibroblasts. It was found that the initial rate of PStranslocation was two orders of magnitude larger infibroblasts than in erythrocytes. In 1995, Farge has mea-sured endocytosis in K562 cells by measuring the uptakeof spin-labeled PC added externally and which can onlybe internalized by endocytosis [13]. He has found that theconcomitant addition of PS stimulates PC internalizationwhich suggests an enhanced endocytosis activity. This canbe explained by the fact that PS is a substrate of thetranslocase and that increasing PS should allow morerapid turnover. The interpretation of Farge’s experimentsraised some questions because of the instability of theshort � chain bearing the nitroxide: these slightly water-soluble lipids were found to be good substrates for theendogenous phospholipase A2 present in such cells as theK562 cells. Nevertheless, the increase in initial rate of PCuptake is probably associated with an increased endocy-tosis activity. In 1999, a different technique was used tomeasure endocytosis in K562 cells [14]. The internaliza-tion of fluoresceinated membrane proteins was eitherwatched directly or monitored by the decrease of fluores-cence which is quenched by the acidity of early endocyticvesicles. It was found that the addition of 4% PS on theexternal side of the plasma membrane resulted in afive-fold increase in fluorescence quenching in 10 min at37 °C. Similarly, PE addition stimulated endocytosis butlyso-PS, which is not transported by the aminophospho-lipid translocase [10], inhibited the residual endocytosis(figure 8) .
In a recent investigation aimed at trying to find theaminophospholipid translocase protein, Marx et al. [43]tested the translocation of PS in endocytosis-deficientyeast cells and found very little specific transport of theaminophospholipid analogues and essentially no ATPdependence. These results could suggest that the cellslacking endocytosis activity have also no aminophospho-
lipid transporter. However, it is surprising that these cellsseem to have normal lipid asymmetry.
Actually, the doubt concerning the proper identificationof the aminophospholipid translocase is the most severelimitation in any experimental proof of its direct implica-tion in endocytosis. Although the sequence of an ATPasewith a molecular mass of 170 kDa has been published byTang et al. [44] and identified as the aminophospholipid
Figure 8. Enhancement of bulk protein endocytosis in K562cells after addition to the external surface of phosphatidylserine(PS) with a short � chain (C5). Bulk membrane proteins werebiotinylated and coupled with FITC streptavidin. The timecourse of internalization of the labeled proteins was monitoredby following the fluorescence decrease of fluorescein which isquenched in the acidic environment of early endocytic vesicles.A. Protein internalization is observed as a function of percentageof PS added. B. Initial bulk flow rate of endocytosis as a functionof the percentage of PS added to the estimate plasma membranephospholipids (reproduced from [14]).
506 Devaux
translocase, there is still a controversy in the literature [43,45]. One will probably have to wait until this protein canbe expressed. Recently, a preliminary report was givenabout a successful complementation of an aminophospho-lipid translocase activity in translocase deficient yeastcells (DRS2) using RNA obtained from plants [38].
6. Exocytosis and lipid scramblase
The same geometrical argument that led us to concludethat the formation of microvesicles by bending of theplasma membrane requires the translocation of lipids fromthe outer to the inner leaflet, forces us to conclude thatfusion of small vesicles to the plasma membrane, i.e.,exocytosis, requires the opposing transfer of lipids (seefigure 9). A priori the two processes (endocytosis andexocytosis) are not exactly symmetrical. Indeed a singlesmall vesicle, with a high surface tension and an asym-metrical repartition of phospholipids can fuse with theplasma membrane without perturbing it because the lattercan be considered as a sink of infinite size. The smallperturbation brought by fusion of one vesicle is a localperturbation that will eventually be canceled out by theslow spontaneous flip-flop of lipids. However, in a livingcell exocytosis is a continuous process that permits thewhole plasma membrane to be renewed within a shortperiod, and fusion of synaptic vesicles or granules is asynchronized event which concerns a large number ofvesicles. It is therefore apparent that independently fromthe process that enables the two bilayers to come in closecontact and to merge into a single bilayer, it would beuseful for the efficiency of this process that some mecha-nism accelerates lipid flip-flop in order to facilitate therandomization of the transmembrane lipid distribution.Additionally, if this lipid scrambling process brings to the
outer leaflet some aminophospholipids (PS and PE), itshould facilitate later the formation of inward invagina-tions (i.e., endocytosis) by the aminophospholipid trans-locase. In this regard, it is quite interesting to note thatmost cells do have in their plasma membrane a proteincalled lipid ‘scramblase’ which is activated by cytosoliccalcium. This scramblase plays an important role inspecialized cells such as platelets which expose PS ontheir outer leaflet when they are stimulated, but it is alsofunctional in all cells of the blood circulation duringapoptosis [46].
In principle this ‘scramblase’ has been purified andcloned [47]. There is also a rare and very severe diseasecalled Scott syndrome which was thought to correspond toa lack of scramblase; at least it is associated with animpairment of calcium triggered lipid redistribution inplatelets. However the reconstituted vesicles with the‘purified’ scramblase have a very low scrambling effi-ciency [48] and this protein is present in the blood cells ofthe Scott patients [49]. So the identification of the proteinmay yet have to be confirmed.
7. Discussion and conclusions
It has been demonstrated in giant unilamellar vesiclesthat a small area increase of one of the two opposingmonolayers triggers budding or invaginations. In variouseukaryotic cells, a positive correlation between amino-phospholipid translocase activity and the endocytic activ-ity was demonstrated. There is also strong evidence thatthe stimulation of phospholipid translocation from theouter to the inner monolayer of the plasma membrane inK562 cells accelerates endocytosis while the addition ofphospholipids that are not transported has an inhibitoryeffect. Therefore, one can infer that the active transport of
Figure 9. Schematic model show-ing the translocation of phospholip-ids from the outer to the innerleaflet during endocytosis and con-versely from the cytosolic to theexternal leaflet during exocytosis.This lipid traffic could be caused(or facilitated) by the ATP-dependent aminophospholipidtranslocase and the calcium depen-dent scramblase respectively. Notethat even if these proteins were notinvolved, i.e., were not the drivingforces; this translocation of lipidshas to take place because of thesmall size of endocytic vesicles.
Is lipid translocation involved during endo- and exocytosis? 507
lipids may serve as a driving force for membrane foldingduring endocytosis. It is an ATP driven process which actson the whole membrane by generating a non-localizedsurface tension; this tension is at least partly relaxed bythe formation of membrane budding and invaginations.The cytoskeleton is necessary to prevent whole celldeformation but organelles which are present in eukaryo-tic cells may help also in a way comparable to innervesicles in MLVs (figures 5, 6). The localization of theseinvaginations has to be random in a large unilamellarvesicle which has an homogeneous surface, but in abiological membrane which contains heterogeneous do-mains, it is quite conceivable that rigid domains orlocalized region with a non-zero spontaneous curvatureimposed by the binding of peripheral proteins such asclathrin, or associated with the clustering of caveolin,serve as a point of emergence. Note the model ofdomain-induced budding of fluid membranes proposed byLipowsky is based on a different physical property,namely the competition between the minimization of thebending energy and the energy of the line tension betweentwo domains [50]. The applicability of the latter model tobiological membranes is still very speculative.
The scramblase hypothesis which suggests a role of thecalcium triggered scramblase during exocytosis is morespeculative at this stage than the hypothesis about the roleof the aminophospholipid translocase during endocytosis.It clearly needs experimental confirmation. One aspectthat I have not discussed yet is the synchronizationbetween endocytosis and exocytosis. Clearly in this modelthe absence of exocytosis should eventually stop endocy-tosis because there is a need of substrates for the lipidpump, namely if no PS and/or PE returns to the exoplas-mic leaflet, the pump cannot continue to work. Thereforethe system is safe. If, on the other hand, the pump had nolipid selectivity, it would continue to pump the whole cellplasma membrane regardless of what happens within thecell. This would eventually be a suicide activity. In theframework of the aminophospholipid translocase modelan evolutionary advantage of lipid heterogeneity becomestherefore apparent.
There are reports of protein catalyzed outward move-ment for example by MRP proteins that would return PCto the outer leaflet in erythrocytes [51]. Although the datain the literature do not allow one to compare quantitativelythe activity of the aminophospholipid translocase and thatof the MRP protein, the very fact that if the level of ATPis raised in the erythrocyte cytosol, the membrane bendsinwardly suggests that, if there is competition between thetwo transports, the inward movement wins. Thus, againthis is a safe situation. There is one example in nature ofcells shedding off pieces of their own plasma membrane.Indeed, stimulated platelets send in the blood circulationmicrovesicles formed by plasma membrane budding.However, platelets do not recover from this process.Whether this vesicle shedding is due to a calcium trig-
gered pump transporting an excess of lipids outwardly, orto the activity of cytoskeleton proteins, is a question notsettled yet [52].
In the framework of this model, clathrin binding couldbe necessary to determine which part of the membranewill invaginate rather than clathrin binding being thedriving force. On the other hand the coat proteins identi-fied as proteins involved in budding of the Golgi system[1] may have a similar effect than the addition of L-PCadded externally to a GUV. Because there is now compel-ling evidence that lipid flip-flop in inner membranes ismuch more rapid than in the plasma membrane which isrich in cholesterol [53], it is unlikely than the samemechanism could be used in the ER or Golgi membranesto generate microvesicles. Of course several mechanismsof budding may exist within a cell.
As a final conclusion, the models proposed here havemany implications that can be verified. Reconstitution ingiant vesicles of the purified proteins involved in lipidtraffic appears to be the most significant type of experi-ments to perform in the future to test the above hypothesis.
Acknowledgments
The author thank Drs. E. Farge, H.G. Döbereiner and S. Cribierfor helpful discussions. Work supported by grants from theCentre National de la Recherche Scientifique (UPR9052) and theUniversité Paris 7-Denis Diderot.
References
[1] Spang A., Matsuoka K., Hamamoto S., Schekman R., Orci L.,Coatomer, ARF1p and nucleotide are required to bud coat proteincomplex I-coated vesicles from large synthetic liposomes, Proc.Natl. Acad. Sci. USA 98 (1998) 11199–11204.
[2] Bremser M., Nickel W., Schweikert M., Ravazzola M., AmherdtM., Hughes C.A., Söllner T.H., Rothman J.E., Wieland F.T.,Coupling of coat assembly and vesicle budding to packaging ofputative cargo receptors, Cell 96 (1999) 495–506.
[3] Sandvig K., van Deurs B., Endocytosis without clathrin, TrendsCell Biol. 4 (1994) 275–277.
[4] Cupers P., Veithen A., Kiss A., Baudhuin P., Courtoy P.J., Clathrinis not required for bulk phase endocytosis in rat fetal fibroblasts, J.Cell Biol. 127 (1994) 725–735.
[5] Miller K., Shipman M., Trowbridge I.S., Hopkins C.R., Transfer-rin receptors promote the formation of clathrin lattices, Cell 65(1991) 621–632.
[6] Simons K., Ikonen E., Functional rafts in cell membranes, Nature387 (1997) 569–572.
[7] Zha X., Pierini L.M., Leopold P.H., Skiba P.J., Tabas I., MaxfieldF.R., Sphingomyelinase treatment induces ATP-independent en-docytosis, J. Cell Biol. 140 (1998) 39–47.
[8] Op Den Kamp J.A.F., Lipid asymmetry in membranes, Annu. Rev.Biochem. 48 (1979) 47–71.
[9] Devaux P.F., Zachowski A., Favre E., Fellmann P, Cribier S.,Geldwerth D., Hervé P., Seigneuret M., Translocation énergiedépendante des amino-phospholipides dans la membrane desglobules rouges, Biochimie 68 (1986) 383–393.
[10] Devaux P.F., Static and dynamic lipid asymmetry in cell mem-branes, Biochemistry 30 (1991) 1163–1173.
508 Devaux
[11] Cribier S., Sainte-Marie J., Devaux P.F., Quantitative comparisonbetween aminophospholipid translocase activity in human ery-throcytes and in K562 cells, Biochim. Biophys. Acta 1148 (1993)85–90.
[12] Müller P., Pomorski T., Herrmann A., Incorporation of phospho-lipid analogues into the plasma membrane affects ATP-inducedvesiculation of human erythrocyte ghosts, Biochem. Biophys. Res.Commun. 199 (1994) 881–887.
[13] Farge E., Increased vesicle endocytosis due to an increase in theplasma membrane phosphatidylserine concentration, Biophys. J.69 (1995) 2501–2506.
[14] Farge E., Ojcius D.M., Subtil A., Dautry-Varsat A., Enhancementof endocytosis due to aminophospholipid transport across theplasma membrane of living cells, Am. J. Physiol. 276 (1999)C725–C733.
[15] Kornberg R.D., Mc Connell H.M., Inside-outside translocation ofphospholipids in vesicle membranes, Biochemistry 10 (1971)1111–1120.
[16] Sheetz M., Singer J.S., Biological membranes as bilayer couples.A molecular mechanism of drug erythrocyte membrane, Proc.Natl. Acad. Sci. USA 71 (1974) 4457–4461.
[17] Helfrich W., Elastic properties of lipid bilayers: theory andpossible experiments, Z. Naturforsch. 28c (1973) 693–703.
[18] Miao L., Seifert U., Wortis M., Döbereiner H.G., Budding transi-tions of fluid bilayer vesicles: the effect of area differenceelasticity, Phys. Rev. E. 49 (1994) 5389–5407.
[19] Svetina S., Zeks B., Membrane bending energy and shape deter-mination of phospholipid vesicles and red blood cells, Eur.Biophys. J. 17 (1989) 101–111.
[20] Berndl J., Käs J., Lipowsky R., Sackmann E., Seiffert U., Shapetransformation of giant vesicles: extreme sensitivity to bilayerasymmetry, Eur. Phys. Lett. 13 (1990) 659–664.
[21] Lipowsky R., The conformation of membranes, Nature 349 (1991)475–481.
[22] Allan D., Walkins C.M., Endovesiculation of human erythrocytesexposed to sphingomyelinase C: a possible explanation for theenzyme resistant pool of sphingomyelin, Biochem. Biophys. Acta938 (1988) 403–410.
[23] Callisen T.H., Talmon Y., Direct imaging by cryo-TEM showsmembrane break-up by phospholipase A2 enzymatic activity,Biochemistry 37 (1998) 10987–10993.
[24] Schmidt A., Wolde M., Thiele C., Fest W., Kratzin H., Podtelejni-kov A.V., Wilker W., Huttner W.B., Söling H.D., Endophilin Imediates synaptic vesicle formation by transfer of arachidonate tolysophosphatidic acid, Nature 401 (1999) 133–141.
[25] Ferrell J.E., Huestis W.H., Phosphoinositide metabolism and themorphology of human erythrocytes, J. Cell Biol. 98 (1984)1992–1998.
[26] Traïkia M., Thèse de doctorat en Biophysique Moléculaire, Uni-versité Denis Diderot, Paris, France, 1999.
[27] Hope M., Redelmeier T.E., Wong K.F., Rodrigueza W., Cullis P.R.,Phospholipid asymmetry in large unilamellar vesicles induced bytransmembrane pH gradients, Biochemistry 28 (1989) 4181–4187.
[28] Mathivet L., Cribier S., Devaux P.F., Shape change and physicalproperties of giant phospholipid vesicles prepared in the presenceof an electric field, Biophys. J. 70 (1996) 1112–1121.
[29] Farge E., Devaux P.F., Shape change of giant liposomes inducedby an asymmetric transmembrane distribution of phospholipids,Biophys. J. 61 (1992) 347–357.
[30] Kas J., Sackman E., Shape transformation and shape stability ofgiant phospholipid vesicles in pure water by area-to-volumechanges, Biophys. J. 60 (1991) 825–844.
[31] Döbereiner H.G., Käs J., Noppi D., Sprenger I., Sackmann E.,Budding and fission of vesicles, Biophys. J. 65 (1993) 1396–1403.
[32] Mui B.L.S., Döbereiner H.G., Madden T.D., Cullis P.R., Influenceof transbilayer area asymmetry on the morphology of largeunilamellar vesicles, Biophys. J. 6 (1995) 930–941.
[33] Farge E., Devaux P.F., Size-dependent response of liposomes tophospholipid transmembrane redistribution: from shape change toinduced tension, J. Phys. Chem. 97 (1993) 2958–2961.
[34] Farge E., Scale-dependent elastic response of closed phospholipidbilayers to transmembrane molecular pumping activity: a key forexo-endocytosis physiological process, Il Nuovo Cimento 16D(1994) 1457–1470.
[35] Devaux P.F., Lipid transmembrane asymmetry and flip-flop inbiological membranes and in lipid bilayers, Curr. Opin Struct.Biol. 3 (1993) 489–494.
[36] Zachowski A., Phospholipids in animal eukaryotic membranes:transverse asymmetry and movement, Biochem. J. 294 (1993)1–14.
[37] Devaux P.F., Zachowski A., Maintenance and consequences ofmembrane phospholipid asymmetry, Chem. Phys. Lipids 73(1994) 107–120.
[39] Birchmeier W., Lanz J.H., Winterhalter K.H., Conrad M.J., ATP-induced endocytosis in human erythrocyte ghosts, J. Biol. Chem.254 (1979) 9298–9304.
[40] Seigneuret M., Devaux P.F., ATP-dependent asymmetric distribu-tion of spin-labeled phospholipids in the erythrocyte membrane;relation to shape change, Proc. Natl. Acad. Sci. USA 81 (1984)3751–3755.
[41] Daleke D.L., Huestis W.H., Incorporation and translocation ofaminophospholipids in human erythrocytes, Biochemistry 24(1985) 5406–5416.
[42] Pomorski T., Müller P., Zimmermann B., Burger K., Devaux P.F.,Herrmann A., Transbilayer movement of fluorescent and spin-labeled phospholipids in the plasma membrane of human fibro-blasts: a quantitative approach, J. Cell Sci. 109 (1996) 687–698.
[43] Marx U., Polawkowski T., Pomorski T., Lang C., Nelson H.,Nelson N., Herrmann A., Rapid transbilayer movement of fluo-rescent phospholipid analogues in the plasma membrane ofendocytosis-deficient cells does not require the Drs2 protein, Eur.J. Biochem. 263 (1999) 254–263.
[44] Tang X., Halleck M.S., Schlegel R.A., Williamson P., A subfamilyof P-type ATPase with aminophospholipid transporting activity,Science 272 (1996) 1495–1497.
[45] Dolis D., Moreau C., Zachowski A., Devaux P.F., Aminophospho-lipid translocase and proteins involved in transmembrane phos-pholipid trafic, Biophys. Chem. 68 (1997) 221–231.
[47] Zhou Q., Zha J., Stout J.G., Luhm R.A., Wieder T., Sims P.J.,Molecular cloning of human plasma membrane phospholipidscramblase, J. Biol. Chem. 272 (1997) 18240–18244.
[48] Bassé F., Stout J.G., Sims P.J., Wiedmer T., Isolation of anerythrocyte membrane protein that mediates Ca2+ dependenttransbilayer movement of phospholipid, J. Biol. Chem. 271 (1996)17205–17210.
[49] Stout J.G., Bassé F., Luhm R.A., Weiss H.J., Wiedmer T., SimsJ.P., Scott syndrome erythrocytes contain a membrane proteincapable of mediating Ca2+-dependent transbilayer migration ofmembrane phospholipids, J. Clin. Invest. 99 (1997) 2232–2238.
[50] Lipowsky R., Domain-induced budding of fluid membranes,Biophys. J. 64 (1993) 1133–1138.
[51] Dekkers D.W.C., Comfurius N., Schroit A.J., Bevers E.M., ZwalF.A., Transbilayer movement of NBD-labeled phospholipids in redblood cell membranes: outward-directed transport by the multi-drug resistance protein (MRP1), Biochemistry 37 (1998)1433–1437.
[52] Gaffet P., Bettache N., Bienvenüe A., Phosphatidylserine exposureon the platelet plasma membrane during A23187 induced activa-tion is independent of cytoskeleton reoganisation, Eur. J. Cell Biol.67 (1995) 336–345.
[53] Gallet F.P., Zachowski A., Julien R., Fellmann P., Devaux P.F.,Maftah A., Transbilayer movement and distribution of spin-labeledphospholipids in the inner mitochondrial membrane, Biochim.Biophys. Acta 1418 (1999) 61–70.
Is lipid translocation involved during endo- and exocytosis? 509
