Model membrane platforms to study protein-membrane interactions

Abstract

Constituting functional interactions between proteins and lipid membranes is one of the essential features of cellular membranes. The major challenge of quantitatively studying these interactions in living cells is the multitude of involved components that are difficult, if not impossible, to simultaneously control. Therefore, there is great need for simplified but still sufficiently detailed model systems to investigate the key constituents of biological processes. To specifically focus on interactions between membrane proteins and lipids, several membrane models have been introduced which recapitulate to varying degrees the complexity and physicochemical nature of biological membranes. Here, we summarize the presently most widely used minimal model membrane systems, namely Supported Lipid Bilayers (SLBs), Giant Unilamellar Vesicles (GUVs) and Giant Plasma Membrane Vesicles (GPMVs) and their applications for protein-membrane interactions.

Keywords::

• Membrane model
• membrane lipid
• membrane protein

Introduction

Biophysical studies generally aim at a quantitative determination of key parameters to characterize a functional biological system. A characteristic of many, if not all, biological systems is their enormous complexity. Therefore, the reconstitution of cellular or molecular subsystems of reduced complexity in cell-free, well-controlled environments has been, and remains, an important task in biological science. This approach thus refers to the modern field of Synthetic Biology, as it implies that a specific biological phenomenon may be approached ‘bottom-up', from a minimal assembly of its functional parts (Schwille and Diez 2009, Arumugam et al. 2011, Schwille 2011). In the present paper, we give a short overview over approaches to understand membrane related biological phenomena by employing suitable model systems simple enough to allow for quantitative analysis, but on the other side complex enough to retain the fundamental character of the phenomena to be observed.

The structural and functional characterization of lipid-protein interactions in biological membranes has always been an enormous challenge to cell biology. Although many biochemical techniques exist for measuring protein-protein interactions, there has been a lack of tools for protein-lipid interactions, since there are no specific targeting agents for lipids, as there are in the form of antibodies for proteins. Taking this into account, the possibility of addressing the specific lipid environment of a protein, or reversely, its segregation into membrane domains of a certain physical nature became an attractive approach. In addition to recent advances in microscopy and spectroscopy that have allowed investigation of protein-lipid or rather, protein-lipid membrane domain interactions in live cells (Sezgin and Schwille 2011), several advanced minimal model membrane systems have helped to address the issue from a ‘bottom-up' paradigm. Here, we summarize the presently most widely used model membrane systems, along with the key membrane properties relevant for protein-membrane interactions.

Historical background

The first experimental investigation of the remarkable physicochemical features of lipidic systems was documented by Benjamin Franklin, in order to understand the calming effect of ‘pouring oil on troubled water'. He remembers the first sparks of his curiosity, when he made the observation that emptying the ship kitchen's greasy cleaning water significantly smoothened their wake (Tanford 2004). In the description of the now-famous Clapham pond experiment, he noted how only a teaspoon of oil covered a significant region of the pond, making it look smooth as glass and explained the physical mechanisms of thin film formation on air-water interface (Tanford 2004). Thus, the first research on lipidic surfaces started with half a membrane – a monolayer (i.e., a monomolecular film formed by lipids on an air-water interface) – well before the concept of a cell membrane was proposed in 1899 by Overton (Overton 1899, Park 1968). Parallel with the discovery of the ultrathin structure of the plasma membrane in 1925 (Gorter and Grendel 1925), Langmuir undertook the first quantitative characterization of monolayers (Langmuir 1917). Decades later, liposomes, the first bilayer system, were discovered (Bangham and Horne 1964). These discoveries led to the ‘Fluid Mosaic' model of the cellular plasma membrane (Singer and Nicolson 1972), describing it as a homogenous lipid environment with embedded proteins, which appeared to be remarkably accurate.

The first major correction to the concept of a homogeneous fluid mosaic followed upon the discovery of detergent resistant membranes (DRMs), which required a model explaining a heterogeneous lipid structure of the plasma membrane (Yu et al. 1973). The 1980s were the golden age of research on pure lipid model membranes, when the phase separation phenomenon was established for monolayers, and supported lipid bilayers were first prepared (McConnell et al. 1984, Tamm and McConnell 1985). A few years later, Giant Unilamellar Vesicles were introduced as a membrane model system that allowed microscopic investigation of cell-sized membrane vesicles (Angelova and Dimitrov 1986). Meanwhile, the heterogeneous distribution of lipids in the apical, compared to the basolateral, plasma membrane in polarized epithelia was observed, strengthening the concept of heterogeneous membrane structure (Vanmeer et al. 1987). Although detergent resistant proteins were also observed in early 1990s (Brown and Rose 1992), it was 1997 when the raft concept was introduced, proposing functional lipid-induced domains (Simons and Ikonen 1997). To confirm and pinpoint the molecular details behind this phenomenon, phase separation was extensively investigated in synthetic membrane systems as models for lipid raft-mediated membrane heterogeneity (Korlach et al. 1999, Bagatolli and Gratton 2000). An important step to a more biological relevance of these phase-separating membrane models was the achievement of functional protein reconstitution into them (Bacia et al. 2004, Girard et al. 2004, Kahya et al. 2005). The most recent manifestation of this quest is the observation of liquid-liquid phase separation in isolated plasma membranes (GPMVs) (Baumgart et al. 2007), 30 years after their discovery (Scott 1976) (Figure 1).

Figure 1. Development of model membrane systems. This Figure is reproduced in colour in the online version of Molecular Membrane Biology.

Membrane features involved in protein-lipid interactions

Membrane composition

As membranes are composed of a collective of lipids and proteins, the main determinant of membrane organization is their constituent molecular composition. However, thorough characterization of the molecular composition of cellular membranes has remained challenging, due to the lack of appropriate techniques. Recent advances in analytical technologies have addressed this issue. For example, mass spectrometric proteomics and lipidomics (Figure 2A) have begun to define the compositions of individual cell compartments (Aebersold and Mann 2003, Marxer et al. 2005, Sampaio et al. 2011). Similarly, NMR and Raman spectroscopy have been used to determine the lipid/protein compositions of cell membranes (Wright et al. 1997, Aussenac et al. 2003, Boehme et al. 2009, Jakop et al. 2009, Estrada et al. 2010). The difficulties and ambiguities of determining lipid compositions of native cellular membranes have actually spurred the development of Giant Unilamellar Vesicles (GUVs), to be discussed in detail below, as a very controlled assembly of binary or ternary lipid mixtures into valuable model systems. It is foreseeable that in the same way as the improvement of analytical methods such as mass spectrometry will in the future allow for a better and more detailed lipidomics, the relevance of well-defined GUVs in biology will diminish.

Figure 2. Membrane features involved in protein-lipid interactions. (A) Lipid composition as determined by Mass Spectrometry. Lipid composition of MDCK cells varies during epithelial polarization. Individual lipid species can be quantified and have unique polarization-dependent behaviors (modified from Sampaio et al. 2011). (B) C-Laurdan spectra report cell-dependent membrane packing of plasma membranes (see Giant Plasma Membrane Vesicles) derived from different cell types. GP calculated as in Equation (1). (C) Membrane lateral pressure. Positive pressure at the head groups (1) and acyl chains (3) is balanced by lateral tension at the interface (2). (D) Hydrophobic match. Lipids adapt to the TMD length of the protein by interfacing long TMDs with long acyl chain lipids. (E) Phase separation. Confocal image of a Giant Unilamellar Vesicle (GUV) composed of DOPC/BSM/Chol shows Lo and Ld phases. Green is cholesterol (labeled with TopFluor) which prefers the Lo phase, while red is the Ld phase marker DiI. (F) Protein induced membrane curvature. This Figure is reproduced in colour in the online version of Molecular Membrane Biology.

Clearly, the lipid composition of a membrane is the key determinant in all protein-membrane interactions, with many proteins using specific lipids, e.g., gangliosides (Coskun et al. 2011), as major targets for their membrane attachment. One lipid of particular importance for many cellular processes, involving cell polarization, but also as the key interface between the membrane and the cytoskeleton is the phosphatidylinositol, PIP₂ (Ho et al. 2012), but other lipids may in the future be identified to provide similar hub functions for processes in and on membranes.

Lipid packing

In addition to the compositional complexity required for the many distinct functions of biological membranes, there exists a remarkable variety of possible lipid structures and consequently, local lipid packing. Lipids are generally grouped into classes by the structure of their hydrophilic head group (e.g., phosphatidylcholine), but the hydrophobic acyl chains also vary greatly in length and degree of saturation. These properties determine the relative packing of assembled membranes, with longer, more saturated lipids forming tightly packed, highly ordered membrane environments, whereas short and unsaturated lipids create looser membranes with more acyl chain flexibility. This packing can be studied by Electron Paramagnetic Resonance (EPR), Nuclear Magnetic Resonance (NMR) or fluorescence anisotropy (Baatz et al. 1990, Pincelli et al. 2000, Han et al. 2001, Vogel et al. 2005), although these methods typically report wholesale membrane properties, thus being of limited use to reveal the role of lateral heterogeneity. Recently, several fluorescent membrane probes have been developed and characterized (i.e., Laurdan, Prodan, C-Laurdan and Di-4-ANEPPDHQ) which are sensitive to water content, from which the relative packing of the membrane can be extrapolated at microscopic resolution (Parasassi et al. 1998, Jin et al. 2005, Kim et al. 2007). When these dyes are excited, their transition dipole moment increases, which results in a reorientation of the ambient water molecules, therefore their emission spectra depend on the aqueous content of the environment (Dietrich et al. 2001), with red-shifted emission indicative of more water and therefore less ordered membranes. A normalized, relative polarity index called generalized polarization (GP) characterizes the extent of emission shift between ordered (tightly packed, less aqueous) and disordered (loosely packed, more aqueous) environment. For example, the emission wavelength for Laurdan is 440 nm in highly ordered, and 490 nm in highly disordered environments (Figure 2B). The GP value is then calculated as:

with a theoretical range of −1 to +1, reflective of minimally and maximally ordered membranes, respectively.

Lipid packing, for instance, plays a role in apolipoprotein A-I (apoA-I) and high density lipoprotein (HDL) binding to membrane. By using Laurdan to probe the lipid packing, it was shown that apoA-I and HDL binding to the membrane prefer certain lipid packing (Sanchez et al. 2010). Dystrophin (a cytoplasmic protein which connects the cytoskeleton to the surrounding extracellular matrix) binding to di-oleyl phospholipids is also influenced by the lipid packing. Changing the lipid packing by introducing lipids such as phosphatidylethanolamine (PE) to the membrane leads to an approximately 7-fold increase in membrane binding affinity of Dystrophin, although Dystrophin does not bind directly to any PE lipid (Le Rumeur et al. 2007). Recently, it has been shown that cell membrane has the ability to form differential lipid packing which provides an important structural insight on the plasma membrane lateral organization (Levental et al. 2011, Sanchez et al. 2012).

Lateral pressure

Due to the amphiphilic nature of lipids, the pressure profile in the direction normal to the plane of the membrane is not constant, as shown in Figure 2C. A negative pressure (i.e., tension) is created at the interface between the water-rich polar headgroup region and the hydrophobic acyl chain region, in order to minimize the disfavored interactions between water molecules and the hydrophobic core. Meanwhile, a positive counteracting pressure is created by the conformational restriction of the hydrophobic acyl chains. These pressures and pressure gradients play important roles in the conformational states of transmembrane proteins (TMP), which may modulate their activity or interactions (Cantor 1997a, 1997b). As an example, dystrophin organization in the membrane was shown to be influenced by the lateral pressure profile (Sarkis et al. 2011). Although the lateral pressure profile, and therefore its effect on membrane-protein interactions, is difficult to measure experimentally, many simulation studies have investigated this phenomenon in detail (Marsh 2007, Ollila et al. 2011).

Hydrophobic (mis)matching

Transmembrane domains (TMDs) of TMPs are typically alpha helices with hydrophobic amino-acid side chains interfacing with the hydrophobic aliphatic tails of membrane lipids. To minimize the energetically unfavorable exposure of these hydrophobic residues to aqueous environments, lipids physically adapt to match TMDs of various lengths, a phenomenon described as hydrophobic matching. Hydrophobic matching brings together long TMD-containing proteins with long acyl chain lipids, resulting in local variations in membrane thickness (Jensen and Mouritsen 2004, Kaiser et al. 2011, Parton et al. 2011) (Figure 2D). This heterogeneity in bilayer thickness can serve to modulate membrane protein function (Andersen and Koeppe 2007). In addition, the thickening of the membrane due to enrichment of saturated and longer acyl chain lipids is supposed to be a characteristic feature of raft entities in cell membranes (Simons and Gerl 2010, Simons and Sampaio 2011), in the same way as the raft-like liquid-ordered phase in phase-separated model membranes (Heberle and Feigenson 2011, Mouritsen 2011). On the other hand, hydrophobic mismatching, i.e., a lack of match between TMD and membrane lipids, can also contribute to the organization of the membrane (Fattal and Benshaul 1993). In case of hydrophobic mismatch, proteins must either aggregate or change their conformation by reducing their effective hydrophobic length (Mouritsen and Bloom 1984, Schaefer et al. 2011), thereby affecting their functional state. The mismatch between POPC and the hydrophobic regions of BtuB (a bacterial outer-membrane beta barrel transport protein), around most of the BtuB circumference, is an example of hydrophobic mismatch which is thought to create the heterogeneity in bacterial outer membrane (Ellena et al. 2011).

Phase separation

Because of its unique interest to biophysics, but since the advent of the ‘lipid raft hypothesis' (Simons and Ikonen 1997) also to cell biology, one of the most widely investigated physicochemical phenomena of biological membranes is liquid-liquid phase separation resulting in lateral domain formation. Preferential interactions between saturated lipids and sterols can lead to the formation of a relatively tightly packed, ordered membrane environment, which readily separates from a relatively disordered phase enriched in unsaturated lipids (Figure 2E). The biological manifestation of this phenomenon, the ‘lipid raft hypothesis', proposes that segregated nanodomains in the cell membrane dynamically concentrate particular proteins/lipids in active platforms to take part in cellular processes (Lingwood and Simons 2010).

Phase separation can be readily visualized by fluorescent probes having a non-uniform partitioning between the coexisting phases. The phase preference of many such dyes, fluorescent lipid analogs, and fluorescently tagged proteins has been studied and utilized in the determination of membrane heterogeneity (Baumgart et al. 2007, Johnson et al. 2010, Juhasz et al. 2010, Sezgin et al. 2012b). Moreover, the membrane polarity-sensitive probes mentioned above are also extensively used for characterization of lipid domain properties (Bagatolli et al. 2003). Using these markers as references for the particular domain, phase preference of many proteins was tested in co-localization studies. The partitioning of SNAREs (Bacia et al. 2004), The β-secretase (BACE) (Kalvodova et al. 2005), the transmembrane linker for activation of T cells (LAT) (Levental et al. 2010b), Hemagglutinin of Influenza Virus (HA) (Nikolaus et al. 2010), GPI anchored proteins (Baumgart et al. 2007), inner leaflet-associated proteins (Sengupta et al. 2008) are some examples of protein partitioning in phase separated systems.

Membrane curvature

As a consequence of the chemical structure of the lipids, the activity of membrane proteins, or the presence of scaffolding proteins or the cytoskeleton, membranes can be curved at length scales that can be sensed by proteins and lipids (McMahon and Gallop 2005), (Figure 2F). Lipid-protein dynamics in a variety of membrane-related events (e.g., endocytosis and exocytosis) are directly driven by this membrane curvature (Kim et al. 1998, Biscari et al. 2002, Graham and Kozlov 2010, Baumgart et al. 2011). Pore formation by proapoptotic proteins (Basanez et al. 2001), interaction of the sterol carrier protein-2 with membranes (Huang et al. 1999), and binding of acyl-CoA binding protein (ACBP) to the membrane (Chao et al. 2002) are examples of processes in which curvature plays an important role. Moreover, membrane curvature itself may be regulated by particular protein-membrane interactions. BAR domain superfamily proteins, for example, are viewed as important regulators of membrane curvature (McMahon and Gallop 2005, Rao and Haucke 2011). BAR domain interaction with the membrane is accomplished by the amphipathic helices of the BAR domains (Bhatia et al. 2009) which proposes a general mechanism for curvature sensing and curvature-driven protein-membrane interactions (Hatzakis et al. 2009, Madsen et al. 2010).

Model systems for studying protein-lipid interactions

In the following, we summarize the three most frequently used biophysical assays to quantitatively address membrane-related phenomena. The first two (supported membranes, GUVs) were initially lipid-only systems, used to determine the physical chemistry of lipids. To specifically investigate the interactions of proteins with these membranes, several protocols have been developed (Kahya et al. 2001, Bacia et al. 2004, Richmond et al. 2011) in order to fully reconstitute the proteins into the membranes. GPMVs, on the other hand, represent a more top-down approach towards the study of lipid-protein systems, as they are derived from living cells by disrupting the membrane, and retain a large lipid complexity, along with an unknown number of integral and peripheral membrane proteins. Spanning from a single lipid bilayer (SLBs) to a complex cell membrane derived vesicle (GPMV), model membranes enabled membrane biologists to investigate the nature of protein-membrane interactions in a system with desired complexity, composition and structure.

Supported lipid bilayers (SLBs)

Supported lipid bilayers (SLBs) are flat membranes formed on a hydrophilic support (such as mica, glass or silica), containing a very thin (∼1 nm) layer of liquid between the support and the membrane. Thanks to the confinement on a solid support, changing physical conditions (e.g., buffer) without disrupting the membrane is much easier in SLBs than in free-standing membranes. Additionally, supported membranes are not only well accessible to all kinds of optical microscopy, but also to surface probe techniques such as atomic force microscopy (AFM), in contrast to free-standing bilayers (Figure 3). Thus, in contrast to free-standing membranes, SLBs provide access to parameters of membrane thickness and mechanical stability (Chiantia et al. 2007) by AFM, and allow to resolve membrane structures far below the optical resolution limit (Chiantia et al. 2008).

Figure 3. Phase separation in Supported Lipid Bilayers (SLBs). (A) Confocal image of a SLB prepared from DOPC/BSM/C18Cer/Chol mixture and stained with DiD (Ld marker). (B) Topological image of the same membrane obtained with AFM. (C) Intensity and height profile of the membrane above. The DiD containing Ld patches are thinner than the patches Lo regions (Image courtesy of Grzegorz Chwastek). This Figure is reproduced in colour in the online version of Molecular Membrane Biology.

The first SLBs were achieved by the sequential transfer of monolayers from an air-water interface to a solid substrate (Tamm and McConnell 1985). Later attempts used the bursting of small vesicles (Small Unilamellar Vesicles [SUVs] or Large Unilamellar Vesicles [LUVs]) on a substrate. The vesicles are first adsorbed on the surface, then ruptured and spread into planar membranes (Richter et al. 2006). The most recent technological development has been SLBs deposited by spin coating membranes directly onto supports (Simonsen and Bagatolli 2004).

Although a solid-supported membrane is quite artificial compared to a free-standing one, the mobility of individual lipids can be preserved (Loose and Schwille 2009), making it a well-controlled system for studying lipid/protein dynamics. Some technical improvements have been introduced to arrive at a relatively unbiased mobility even of integral proteins reconstituted into SLBs. For example, polymer cushions have been placed on the support to increase the hydration layer, to avoid non-specific trapping of proteins with large extramembrane domains at the support (Wagner and Tamm 2000). Recently, to eliminate the effect of the surface/lipid interaction, a lipid bilayer was suspended between narrowly separated solid supports or at a small orifice, allowing the formation of a free-standing planar bilayer. This assay has been employed for simultaneously measuring optical and electrical properties of the membrane (Samsonov et al. 2001, Honigmann et al. 2010). Similarly, bursting vesicles on porous substrates (like alumina or Si₃N₄) is an alternative way to eliminate the effect of support but preserving the experimental advantages of flat SLBs (Hennesthal et al. 2002, Simon et al. 2007, Heinemann and Schwille 2011).

Interaction of Cytochrome C and alkaline phosphatase with the membrane was investigated using SLBs (Milhiet et al. 2002, Choi and Dimitriadis 2004). They were also used to figure out the effect of glycine to alanine substitution on interactions of antimicrobial peptide latarcin 2a with the membrane lipid bilayer (Idiong et al. 2011). Selective binding of HPA3, an antimicrobial and cell-penetrating peptide, to DMPC, DMPG, POPC and POPG is another example where SLBs were used as a model membrane system (Hirst et al. 2011). SLBs are also extensively used to mimic the bacterial membranes which provided valuable information about the prokaryotic cellular structures (Loose et al. 2008).

Giant Unilamellar Vesicles (GUVs)

In aqueous environments, lipid assemblies are easily transformed morphologically by changes of the physical parameters in the solution. Mild fluid convection, but also slowly alternating electric fields can lead to the vesiculation of a flat lipid multilayer, allowing the formation of unilamellar vesicles much larger than those prepared by sonication or injection methods, termed Giant Unilamellar Vesicles (GUVs >1000 nm). Originally, GUVs were formed using DC electric fields (Angelova and Dimitrov 1986), though AC electric fields were later found to increase formation efficiency (Angelova and Dimitrov 1988).

The basic protocol for GUV preparation involves the spreading of a lipid mixture (in organic solvent) on a platinum electrode to form the lipid multilayer. Electrodes are then dipped into aqueous buffer and the electric field (usually between 2 and 3 V) is applied. GUVs can also be prepared by gentle hydration, where charged lipids are used to create electrostatic repulsion between the bilayers, which eventually facilitates unilamellar vesicle formation. For details on GUV preparation refer to ref (Garcia-Saez et al. 2010, Morales-Penningston et al. 2010).

Besides having a great potential to selectively study single lipid species in a model membrane, it is also possible to form GUVs with complex lipid mixtures. This capability led to reconstitution of phenomena structurally and functionally similar to cellular lipid rafts in GUVs. One early assumption was that lipid rafts are related to more ordered, or more tightly packed, liquid membrane phases from the bulk membrane. The first attempts to observe phase separation in GUVs were made with binary mixtures of saturated phospholipids, namely DPPC/DLPC and DPPC/DPPE (Korlach et al. 1999, Bagatolli and Gratton 2000). These mixtures indeed phase separate under near-physiological conditions; however, while one of the phases retains the liquid-crystalline properties of biological membranes, the other is a solid crystal (in membranes, termed the ‘gel' phase) with little translational or rotational diffusion. Due to the limited biological relevance of the gel phase, another component was added to recapitulate liquid-liquid phase coexistence. The key ingredient turned out to be cholesterol-ternary mixtures of a saturated lipid, an unsaturated lipid and cholesterol readily resulted in coexisting liquid ordered (Lo) and liquid disordered (Ld) domains (Dietrich et al. 2001) (Figure 4).

Figure 4. Phase Properties in GUVs. (A) 3D confocal image of phase separation in GUVs composed of DOPC/BSM/Chol (2:2:1; dyes are same as Figure 2A). (B) Laurdan GP image of a phase separated GUV (same composition as [A]). Ordered phases have higher GP values. (C) Diffusion of a fluorescent molecule in Lo and Ld phases of Giant Unilamellar Vesicles (GUVs) quantified by fluorescence correlation spectroscopy (same composition as [A]). Diffusion is approximately an order of magnitude slower in the Lo phase. (D) Height profile of a burst GUV (same composition as [A]) obtained by AFM. The Lo phase is thicker than Ld phase by about 1 nm. This Figure is reproduced in colour in the online version of Molecular Membrane Biology.

In addition to their crucial role in defining the physical chemistry of raft-mimetic Lo phases, GUVs have been widely used as platforms to investigate the role of individual lipids in protein activity or interactions. Profilin interaction with phosphatidylinositol (4,5)-bisphosphate (Krishnan et al. 2009), caspase-8/Bid-FL complex binding to cardiolipin (Jalmar et al. 2010), tBID interaction with BCL(XL) (Garcia-Saez et al. 2009), and the activity of pore forming peptides (Fuertes et al. 2010) are some of the biological processes where GUVs have been used as model platforms. The distribution and the mobility of the lipid-anchored oligopeptide-binding protein OppA, the translocator complex OppBCDF of the oligopeptide ABC transporter, the mechanosensitive channel of MscL and the secondary lactose transport protein LacS were studied using GUVs (Doeven et al. 2005). SNAREs (Bacia et al. 2004) and beta secretase (BACE) proteins (Kalvodova et al. 2005) were also reconstituted into GUVs to elucidate their Lo/Ld phase preference and their lipid-modulated bioactivity.

Giant plasma membrane vesicles

Although non-lipid cellular components (such as functional membrane proteins) can be reconstituted into GUVs, these experiments are rather work-intensive and still only capable of reproducing a limited complexity, compared to the physiological situation. This is particularly true for the extensive variety of lipids in a native cellular membrane. An optimal complement to simple GUV models is a free-standing membrane which retains at least the compositional lipid complexity of a live cell membrane. Such a system is provided by so-called membrane blebs, which can be grown and harvested from native cellular membranes (Sezgin et al. 2012a).

Blebs are pure cell membrane regions, detached from the cell cortex, which can naturally form during cell division, apoptosis or cell migration, or are induced by some chemical or physical destruction of the cytoskeleton. In dividing cells, blebs provide an attractive system for studying the cell mechanics and biophysics of membrane-cytoskeleton interactions in live cells (Charras et al. 2008).

Chemically-induced cell membrane vesiculation (so-called plasma membrane vesicles [PMVs]) was first observed in 1976, and used as a method to isolate plasma membrane (Scott 1976). Recently, phase separation was observed in Giant Plasma Membrane Vesicles (GPMVs) (Baumgart et al. 2007) (Figure 5), mirroring the behavior of the three component GUVs described above in a system retaining the compositional complexity of the native plasma membrane. Measurements of temperature (Johnson et al. 2010) and composition dependence of phase separation in GPMVs (Keller et al. 2009) have further confirmed these similarities. A similar method was developed to separate intact Plasma Membrane Spheres (PMS) from the cell interior (Lingwood et al. 2008), and phase separation was shown by confocal imaging and fluorescence correlation spectroscopy.

Figure 5. Phase Properties in Giant Plasma Membrane Vesicles (GPMVs). (A) Phase separation of GPMVs at 10°C stained with Topfluor Cholesterol (ordered phase), DiI (disordered phase) or Alexa647-Choleratoxin (ordered phase). (B) Laurdan GP image of a phase separated GPMV at 10°C. The order difference between coexisting phases is less pronounced than in GUVs shown in Figure 4B. (C) The diffusion of a fluorescent molecule in two phases of GPMVs measured by FCS. Diffusion in the ordered phase is slower, but much closer to the disordered phase than in GUV shown in Figure 4C. (D) AFM image of a burst GMPV. The surface is covered by proteins. The inset shows the corresponding confocal image of the membrane stained with DiI (disordered phase). This Figure is reproduced in colour in the online version of Molecular Membrane Biology.

These systems, in particular the GPMVs, have been used to confirm and extend existing hypotheses about the roles of lipid modifications on membrane proteins, for instance (Levental et al. 2010a, 2010b, Thaa et al. 2011). Additionally, interactions between interleukin receptors and dimer formation of interleukin-4 receptor (Worch et al. 2010) were addressed using this system. Moreover, they have been beneficial to illuminate the physical packing and molecular composition of the cell membrane (Kaiser et al. 2009, Levental et al. 2011).

Outlook – the future of model membranes

As summarized in Table I, there exist advantages and disadvantages of each membrane model system discussed above. Each system has its own difficulties and limitations and unfortunately, also produces its own artifacts. Moreover, all the model membranes mentioned above are at thermodynamic equilibrium, which does clearly not reflect the state of a living cell membrane. Similarly, membrane asymmetry, i.e., different compositions of the two opposing leaflets of a bilayer, is a central feature of biological membranes (Chiantia et al. 2011), while only the first steps have been taken to establish and investigate asymmetric model membranes (Collins and Keller 2008, Chiantia et al. 2011, Richmond et al. 2011) since the introduction of Montal-Muller membranes (Montal and Mueller 1972). Thus, the development and study of asymmetric membranes away from equilibrium presents the next challenge in the processive development toward truly biomimetic model membranes.

Table I. Pros and cons of model membranes.

	Advantages	Disadvantages
SLBs	Compositional control Simple buffer exchange High stability Eligible to surface techniques	Support effects Limited complexity Difficult to incorporate proteins
Suspended SLBs	Compositional control Simple buffer exchange Free-standing membranes Eligible to surface techniques	Limited complexity Difficult to incorporate proteins
GUVs	Compositional control Free-standing membranes	Limited complexity Difficult to incorporate proteins Difficult buffer exchange Low stability
GPMVs	Near-physiological complexity Proteins included Free-standing membranes	No compositional control Difficult buffer exchange Low stability

SLBs, Supported Lipid Bilayers; GUVs, Giant Unilamellar Vesicles; GPMVs, Giant Plasma Membrane Vesicles.

Acknowledgements

We thank Grzegorz Chwastek and Fabian Heinemann for Atomic Force Microscopy images and Ilya Levental for his careful reading of and valuable comments on the manuscript. Financial support by the German Research Foundation (DFG Project Number SO-818/1-1 and SFB TRR 83/1-2010) is gratefully acknowledged.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.