Looking at lipid rafts?

Transcription

1 One of the most intriguing issues in membrane biology is whether non-caveolar glycosphingolipid (GSL)-enriched domains 1 actually exist in the plane of the plasma membrane. One reason for the high interest in the domains is that, if they exist, they would underscore the importance of lateral organization in the plasma membrane for complex activities such as signal transduction 2. It was postulated initially that lipid domains in Golgi membranes act as an apical sorting device in epithelial cells 3 based on the resistance of specialized lipid fractions and apically directed glycosylphosphatidylinositol (GPI)-anchored proteins to extraction with cold non-ionic detergent 4. Whether such domains actually exist in vivo has been debated vigorously since the inception of the idea. A number of different acronyms have been attached to the operationally defined membrane fractions, including DRMs (detergent-resistant membranes) and DIGs (detergentinsoluble glycolipid-enriched membrane domains), but perhaps the most descriptive term for such domains in the plane of various membranes is simply lipid rafts 3. The detergent-resistant fractions include caveolae, a morphologically well-established tubovesicular invagination of the plasma membrane, thought to be involved in various signal-transduction processes and several different forms of transport 5. However, such fractions can also be isolated from cells not exhibiting overt caveolae or expressing caveolin 6. This, together with the fact that the size of detergent-resistant membranes as established by electron microscopy 4,7 is often much larger than individual caveolae suggests that caveolae are a subclass within the operationally defined detergent-resistant fractions 3,8,9. It is certainly plausible that non-caveolar, GSLenriched plasma membrane domains with a distinct lipid composition could form. Because of their special acyl chain composition (long chains, very little unsaturation), GSLs, sphingomyelins and saturated phospholipids would preferentially associate laterally with themselves and with cholesterol 9,10. Indeed, in model membranes enriched in sphingomyelin and cholesterol, these components form a detergent-resistant liquid-ordered phase in which the acyl chains are ordered but the components retain substantial lateral and rotational mobility 10,11. Adjacent GSLs could hydrogen bond through their glycosyl moieties, providing an additional cohesive energy to further enhance lateral stability of putative domains. It is interesting that the most stable membranes are those of fibre cells of the mammalian lens. These membranes, which can last a lifetime, are rich in sphingomyelin and its derivatives, as well as cholesterol 12,13. An opposing view, however, is that, in the context of the large number of plasma membrane lipid species, the favourable interactions required to form lateral domains would not be strong enough to create domains that are large enough or long-lived enough to be biologically significant 14. This view is tempered by the knowledge that, in model systems, the energies required to demix components to Looking at lipid rafts? Ken Jacobson and Christian Dietrich create domains are small 15. Moreover, the high concentrations of the interacting lipid species could drive even weak associations by the law of mass action. The protein components coisolating in the detergent-resistant fractions include various GPIanchored proteins, some transmembrane proteins, non-receptor tyrosine kinases, certain G-protein subunits and several transporters 3,5,8. Many of these components are dually lipidated with saturated acyl chains, which would give them a preference to reside in more-ordered lipid environments 8. This variety of components suggests that, if these fractions represent actual domains, they could serve as hot spots for signal transduction. For example, in lymphoid and endothelial cells, the transmembrane hyaluronan receptor, CD44, selectively associates with the Src-family kinases Lck and Fyn in detergent-resistant domains where CD44-mediated signalling might occur 16. Moreover, the ability of various components to translocate in and out of such fractions (domains) presents unexpected opportunities for regulation. For example, the IgE receptor (Fc R1) in rat basophilic leukaemia cells, when binding IgE and antigen, moves into detergent-resistant fractions, mediating increased signalling via Srcfamily kinases By contrast, in T cells, hyperphosphorylated p56 lck appears to be sequestered from its activating phosphatase, CD45, when in the detergent-resistant fraction 20. FORUM comment The notion that microdomains enriched in certain specialized lipids exist in membranes has been both attractive and controversial since it was first proposed that such domains, termed rafts, might act as apical sorting devices in epithelial cells. The observation that certain lipids are not extractable in cold nonionic detergent supports the raft concept, but the nature of the in vivo correlate of such detergent-resistant membranes remains enigmatic. In principle, microscopy should be able to determine whether the postulated rafts exist. This article focuses on recent microscopy experiments addressing this question. Several, but not all, results support the raft concept, but further definition of the structure, dynamics and function of lipid domains in various biological contexts is urgently required. The authors are in the Dept of Cell Biology and Anatomy (K. J., C. D.) and Lineberger Comprehensive Cancer Research Center (K. J.), University of North Carolina, Chapel Hill, NC , USA. frap@med. unc.edu trends in CELL BIOLOGY (Vol. 9) March /99/$ see front matter 1999 Elsevier Science. All rights reserved. 87 PII: S (98)

2 (a) Concentration depolarization FRET (b) Conventional FRET Excitation Emission Excitation Emission (c) SPT Polarized emission in dilute Depolarization by homoenergy transfer in concentrated Normal donor emission in dilute Red-shifted acceptor emission via donor acceptor energy transfer in concentrated Confinement zone ~300 nm FIGURE 1 (a) Top: normal fluorescence emission, symbolized by the green wave, is excited by the blue wave. The emission of the fluorophore is significantly polarized to a degree that is dependent on its orientation, rotational mobility and spectral properties. Bottom: concentration depolarization by fluorescence resonance energy transfer (FRET) between like fluorophores (homoenergy transfer) occurs in a concentrated two- or three-dimensional (when the fluorophores are within 10 nm of each other) but not in dilute. Energy transfer indicated by curved arrow. (b) Top: normal fluorescence emission, here depicted as a green arrow, which is excited by blue radiation (blue arrow). Bottom: conventional FRET results from donor excitation being transferred to a proximate acceptor with subsequent emission at longer wavelength (red arrow) by the acceptor. FRET (which can be measured as enhanced acceptor emission or donor quenching) occurs only when donor and acceptor are within 10 nm of each other as might occur between GPI-anchored proteins concentrated in the same raft. Energy transfer indicated by curved arrow. (c) Single-particle tracking (SPT) reveals existence of confinement zones of ~300 nm diameter (red circle) in which diffusion is slowed significantly. In principle, modern light-microscopic methods should provide key information on whether there is an in vivo correlate of the non-caveolar detergentresistant membrane fraction. At this juncture, however, the answer is not completely clear because several recent papers come to opposite conclusions regarding the existence of rafts. In this article, we review some recent microscopy studies and their impact on the raft hypothesis. Recent microscopic evidence for rafts A number of recent light-microscopy papers that employ sophisticated strategies provide evidence supporting the raft concept. In an elegant study, Stauffer and Meyer 19 produced green fluorescent protein (GFP)-tagged Src-homology 2 (SH2) domain constructs that transiently translocated from the cytoplasm to GM1-rich regions in the plasma membrane upon activation of rat basophilic leukaemia cells with antigen. Concomitant and transient redistribution of the IgE receptor to the GM1-enriched domains was also observed. This presumably reflects the recruitment of appropriate signalling kinases to such domains and would therefore provide an in vivo correlate of biochemical studies showing translocation of antigen-activated IgE receptor to detergent-resistant membranes enriched for the kinase Lyn 17. In addition, Harder et al. 21 have performed studies on BHK cells expressing, by transfection, a variety of GPI-anchored and transmembrane proteins. They showed that GPI-anchored proteins, such as placental alkaline phosphatase and Thy-1, colocalize with GM1-enriched domains. Transmembrane proteins, such as viral haemagglutinin, which are isolated in the detergent-resistant membranes, also colocalize in a patched pattern with GPI-anchored proteins, but others, for example the transferrin receptor, which are detergent-extractable, do not copatch. Copatching of components sequestered in rafts is thought to result from the affinity of rafts for one another. Patched raft components, such as Thy-1 22, can cap in fibroblasts, suggesting a linkage of these raft components, once patched, to the retrograde flow of the cortical cytoskeleton. One recent paper utilizing energy transfer in a very novel way strongly supports the existence of rafts. When expressed in transfected CHO cells, the folate receptor (FR), a GPI-anchored protein, shows spectroscopic evidence of being clustered 23. This conclusion is based on the principle that the emission from concentrated s of fluorophores is depolarized by homoenergy transfer 24 (Fig. 1a); this effect can only be relieved by dilution, which separates the fluorophores. Varma and Mayor 23 observed that the emission anisotropy (a measure of fluorescence depolarization) of a fluorescein-labelled folic acid analogue bound to the FR is the same for pixels with fluorescence intensities that vary by a factor of 200, and that the anisotropy could only be increased by either photobleaching the analogue or dilution with folic acid. These results can be used to argue that the FR is clustered into domains that are 70 nm in dimension or smaller (Fig. 2b). Moreover, coexpression of a second GPI-anchored protein, decay accelerating factor (DAF), leads to an increased anisotropy of the labelled FR, suggesting that FR and DAF can occupy the same raft (S. Mayor, pers. commun.). Consistent with the raft model, removal of cholesterol causes clusters to disperse; furthermore, a chimeric transmembrane form of the FR, lacking the GPI anchor, fails to exhibit clustering as judged by this assay. 88 trends in CELL BIOLOGY (Vol. 9) March 1999

3 Clustering of the FR is also supported by chemical crosslinking studies on similar FR-expressing cells 25. These studies also concentrated on a chimeric GPIanchored protein growth hormone fused to the GPI anchor from decay accelerating factor expressed in MDCK cells. Crosslinking indicated that clusters were composed of at least 15 molecules and that clustering was inhibited by cholesterol removal. By contrast, chemically crosslinking growth hormone fused to transmembrane anchors did not result in clustering in this assay. Further support for the raft concept comes from a new fluorescence spectral technique exploiting the fact that GFP exhibits an excitation spectrum that is dependent on contact between adjacent GFPs. In HeLa cells, GPI-anchored GFP shows an excitation spectrum consistent with a loose clustering of these proteins 26. The degree of clustering is significant compared with either soluble GFP or palmitoylated GFP but considerably less than that mediated by crosslinking with antibodies against GFP. Single-particle tracking (SPT) studies, in which the lateral mobility of single or small groups of membrane molecules is followed by attaching a small gold or fluorescent particle to the component of interest, are also consistent with the raft concept 27. These studies detect what are termed transient confinement zones (Fig. 1c) in which roughly one-third of both Thy-1, a GPI-anchored protein, and GM1, a GSL, are confined to zones nm in diameter. The number and size of these zones diminish when GSL synthesis is metabolically inhibited, but the number of zones is unaffected by extraction of the cells with cold Triton X-100 (although their size is reduced somewhat), suggesting that the zones are detergent resistant. Furthermore, fluorescein phosphatidylethanolamine (Fl-PE), a fully saturated phospholipid analogue that should be accommodated to some extent by the ordered raft lipids, partitions into transient confinement zones but to a lesser degree than Thy-1 or GM1. The number, but not the size, of these zones markedly increases as temperature is lowered (K. Jacobson and C. Dietrich, unpublished), a condition that favours FIGURE 2 Cholesterol Sphingomyelin Phospholipid (a) (b) (c) GSL Antibody GPI-anchored protein ~100 nm ~>500 nm ~>500 nm Cytoskeleton Proteins involved in signal transduction Possible lipid microdomains (rafts) in the plane of the plasma membrane. In these diagrams, lipids in the liquid-ordered phase have black, bold acyl tails; the normal bilayer is in liquiddisordered phase, where lipids are depicted as having grey acyl chains. Cholesterol distribution reflects only a relative enrichment in the liquid-ordered phase; molecules are not drawn to scale. (a) Molecules are randomly distributed or clustered in such small microdomains, comprising just a few molecules (dashed boxes), that no appreciable fluorescence resonance energy transfer (FRET) occurs between labelled raft components. Such domains would be too small to mediate a biological function. (b) Upper panel: rafts of intermediate size (~100 nm) in which the lipids might be in a liquid-ordered phase that is enriched in cholesterol. One such domain could accommodate up to a maximum of ~600 proteins, assuming a molecular mass of 50 kda. Lower panel: raft size could be modulated by antibody-induced crosslinking [here, for simplicity, depicted as mediated by a single layer of bivalent antibodies against the generic glycosylphosphatidylinositol (GPI)-anchored protein]. (c) Upper panel: macrorafts of the order of >500 nm in dimension in which the lipids might be in a liquid-ordered phase that is enriched in cholesterol. The extent of raft development might be regulated by the membrane skeleton fence 29. Lower panel: model scenario in which the membrane skeleton fence confines laterally mobile intermediate-sized rafts to small regions of the plasma membrane defined by the effective mesh size of the fence (of the order of 500 nm). trends in CELL BIOLOGY (Vol. 9) March

4 the formation of detergent-resistant membranes. This observation and the fact that both GPIanchored and transmembrane isoforms of NCAM show similar-sized confinement zones 28 suggest that the membrane-associated cytoskeleton 29, in some way, regulates raft size, possibly by confining diffusing small rafts (see Fig. 2c, lower panel). Microscopic evidence that does not support the raft concept Conventional fluorescence microscopy employing direct immunofluorescence or fluorescent ligand analogues to visualize GPI-anchored proteins gives fairly uniform staining of live cells, with little evidence of punctate patterns indicative of clusters packed with these proteins. If rafts do exist, there are several possible explanations for this negative finding. The rafts could be below the re limit of the light microscope (~250 nm) and not be sufficiently concentrated in the GPI-anchored protein to appear as fluorescent dots at the re limit. It could also be that rafts are large enough to be resolvable but are simply not enriched enough in the fluorescent component to appear as distinct entities. In addition, it is possible that the fluorescent antibody or ligand employed disrupts the putative raft localization 5. Studies on fixed MDCK cells using fluorescence resonance energy transfer (FRET), which detects molecular proximity by donor-to-acceptor energy transfer when molecules are within 10 nm or less (Fig. 1b), showed little evidence of clustering of a transfected GPI-anchored protein 5 nucleotidase expressed in a steady-state distribution at the apical surface 30. In this study, s containing antibodies conjugated to either Cy3 (donor) or Cy5 (acceptor) in varying ratios were used to label the cells, providing variable acceptor-to-donor ratios. FRET was measured as a function of surface density of the enzyme and at increasing ratios of acceptor to donor. Predictions of the results were constructed from previously published theoretical models of energy transfer in a two-dimensional membrane for clustered, mixed clustered and random and totally random distributions of 5 nucleotidase. The results were consistent with a random distribution of this GPI-anchored protein and suggest that rafts are either very small (containing only one or two enzymes; Fig. 2a) or that the raft comprises the entire apical membrane. This is an especially striking result because it is based on the same principle, namely energy transfer between proximate fluorophores, but leads to a conclusion diametrically opposed to that concerning the clustering in rafts of other GPIanchored proteins 23. Another fluorescence microscopy study also appears to be inconsistent with the existence of rafts in the plasma membrane. Pagano and coworkers 31 used changes in the emission spectrum of a bodipy sphingomyelin analogue (C5 DMB-SM) to study where this lipid probe concentrates in membranes. As the probe becomes more concentrated, its emission shifts from green to red. Thus, the existence of rafts would have been expected to lead to small regions of more red fluorescence superimposed on the green emission. At low concentrations of C5 DMB-SM, only green emission is seen in the plasma membrane, indicating the absence of domains containing high concentrations of the probe. Under the same conditions, however, a relative concentration of the probe was indicated by red fluorescence from very early endosomes near the plasma membrane. It is quite possible, however, that the presence of the bodipy moiety at the end of a shortened acyl chain of this sphingomyelin analogue creates a more bulky probe that does not pack well into the ordered environment postulated for rafts. Conclusions and outlook Although each of these recent light-microscopy studies appears to be technically sound, specific caveats, arising from the particular methodologies, can be raised against each. Many are performed by transfecting the GPI-anchored protein into the cells thus expression levels will not be the same as in the wild-type cells. The FRET study finding no evidence for rafts used the photobleaching energy transfer method, which is more conveniently done with fixed cells; fixation could disrupt fragile lipid rafts. For both the fluorescence microscopy and SPT studies, labelled antibodies or other ligands are generally required, which have the potential to perturb the system. In particular, the SPT studies require a gold particle coated with antibody (or cholera toxin for GM1) that is 30 or 40 nm in diameter, giving several membrane-binding sites per bead, which could cause some aggregation of very small rafts. Keeping these limitations in mind, the various experimental results support a number of models for raft structure 9,11. Fig. 2 gives a range of possible raft models. Vanishingly small rafts are shown in Fig. 2a and are so small that they would have no significant biological function. Intermediate, ~100 nm scale, rafts are depicted in Fig. 2b (upper panel) that might, under crosslinking conditions, cluster together (Fig. 2b, lower panel) because of inter-raft affinity 21. Indeed, it is possible that the membrane constitutively contains many small domains and is poised to facilitate different functions by oligomerization-induced clustering 11. Finally, larger-scale macrodomains (of the order of 500 nm or larger) are depicted schematically in Fig. 2c (upper panel). The occurrence of larger domains could arise from large areas of the membrane being composed of such components or from cold detergent extraction causing coalescence of numerous smaller rafts 25,32. The possible involvement of the cytoskeleton underlying the membrane in the regulation of intermediatesized raft mobility is depicted in Fig. 2c (lower panel). We feel that the accumulating evidence favours the existence of some kinds of raft, but the important details regarding size and composition in different biological contexts remain to be elucidated. Indeed, the negative results and the arguments about the size of these domains indicate that the formation, stability and partitioning of molecules into rafts depends on subtle details. There is reason to 90 trends in CELL BIOLOGY (Vol. 9) March 1999

5 hope that the discordant microscopic results will be reconciled and a much clearer picture of lipid rafts in the plane of the membrane will emerge soon. Furthermore, the application of additional tools, including laser tweezers 29 and image-correlation spectroscopy 33, should enhance the definition of these potentially most interesting plasma membrane structures. Among the most pressing issues are: the dynamics and size of rafts and the molecular determinants of these characteristics; further definition of the affinities of various membrane components for rafts, how such affinities might be functionally regulated 11,21, and the related issue of the duration that particular components reside in a given raft; the relationship of putative rafts to the actinbased cytoskeleton; how the outer monolayer, containing GPIanchored proteins, GSLs, sphingomyelin and cholesterol, is coupled to the inner monolayer into which acylated signal-transduction components insert 18,34. References 1 Thompson, T. E. and Tillack, T. W. (1985) Annu. Rev. Biophys. Biophys. Chem. 14, Cinek, T. and Horejsi, V. (1992) J. Immunol. 149, Simons, K. and Ikonen, E. (1997) Nature 387, Brown, D. A. and Rose, J. K. (1992) Cell 68, Anderson, R. G. W. (1998) Annu. Rev. Biochem. 67, Parton, R. and Simons, K. (1995) Science 269, Schnitzer, J. E. et al. (1995) Science 269, Brown, D. A. and London, E. (1997) Biochem. Biophys. Res. Commun. 240, Harder, T. and Simons, K. (1997) Curr. Opin. Cell Biol. 9, Brown, R. E. (1998) J. Cell Sci. 111, Brown, D. A. and London, E. (1998) J. Membr. Biol. 164, Zelenka, P. S. (1984) Curr. Eye Res. 3, Borchman, D., Byrdwell, W. C. and Yappert, M. C. (1996) Ophthalmic Res. 28, Edidin, M. (1997) Curr. Opin. Struct. Cell Biol. 7, Parsegian, V. A. (1995) Mol. Membr. Biol. 12, Ilangumaran, S., Briol, A. and Hoessli, D. C. (1998) Blood 91, Field, K. A., Holowka, D. and Baird, B. (1997) J. Biol. Chem. 272, Sheets, E. D., Holowka, D. and Baird, B. Curr. Opin. Chem. Biol. (in press) 19 Stauffer, T. P. and Meyer, T. (1997) J. Cell Biol. 139, Rodgers, W. and Rose, J. K. (1996) J. Cell Biol. 135, Harder, T. et al. (1998) J. Cell Biol. 141, Holifield, B. F., Ishihara, A. and Jacobson, K. (1990) J. Cell Biol. 111, Varma, R. and Mayor, S. (1998) Nature 394, Weber, G. (1969) in Fluorescence and Phosphorescence Analysis (Hercules, D. M.), pp , John Wiley 25 Friedrichson, T. and Kurzchalia, T. V. (1998) Nature 394, De Angelis, D. A. et al. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, Sheets, E. et al. (1996) Biochemistry 36, Simson, R., Sheets, E. and Jacobson, K. (1997) Biophys. J. 69, Kusumi, A. and Sako, Y. (1996) Curr. Opin. Cell Biol. 8, Kenworthy, A. K. and Edidin, M. (1998) J. Cell Biol. 142, Chen, C-S., Martin, O. C. and Pagano, R. E. (1997) Biophys. J. 72, Mayor, S. and Maxfield, F. R. (1995) Mol. Biol. Cell 6, Srivastava, M. and Petersen, N. O. (1996) Meth. Cell Sci. 18, Brown, D. A. and London, E. (1998) Annu. Rev. Cell Dev. Biol. 14, Acknowledgements We thank R. Anderson, J. Costello, M. Edidin, A. Kenworthy and S. Mayor for their helpful comments. This work was supported by NIH GM and a DFG fellowship (Di 691/1-1). GFP in Motion CD-ROM With last month s special issue, we gave all subscribers a free copy of our GFP in Motion CD-ROM. This is a collection of over 150 movies made using green fluorescent protein. The CD was compiled by Beat Ludin and Andrew Matus, and was funded through generous sponsorship from Applied Imaging, Chroma Technology, Clontech, Leica Microsystems, Life Imaging Services, Life Science Resources and Universal Imaging. We are distributing copies of the CD to nonsubscribers, subject to availability. If you are not a TCB subscriber but would like a copy of the CD, please send your name and address to tcb@elsevier.co.uk. trends in CELL BIOLOGY (Vol. 9) March