Proc. Natl. Acad. Sci. USA Vol. 90, pp. 10909-10913, December 1993 Review Caveolae: Where incoming and outgoing messengers meet Richard G. W. Anderson Department of Cell Biology and Neuroscience, University of Texas Southwestem Medical Center, Dallas, TX 75235

ABSTIRACT Plasmalemmal caveolae ing. At the same time, this information is This portable, membrane-bound com- were flrst identified as an endocytic com- used to construct several models that partment has been found to contain a partment In endothelial cells, where they illustrate the different ways that caveolae number of molecules that are known to appear to move molecules across the cell might function in both intracellular and participate in cell signaling. There are by transcytosis. More recently, they have intercellular communication. three classes of molecules: enzymes that been found to be sites where small mole- generate messengers from substrates in cules are concentrated and internalized by Caveolae the environment, high-affinity binding a process called potocytosis. A growing sites that concentrate chemical signals, body of biochemical and morphological Each caveola is a dynamic piece ofmem- and substrates that are enzymatically evidence indicates that a variety of mole- brane that is either open for receiving and converted into messengers. cules known to function directly or indi- releasing material or closed for process- GPI. Insulin was the first hormone rectly in signal transduction are enriched ing, storage, and delivery to the cell (11). suspected of using inositol phosphogly- in caveolae. This raises the possibility that The exact nature of the closed compart- can (IPG) or a molecule derived from IPG a third function for caveolae is to process ment is still unclear. In endothelial cells, as a second messenger (19-21). Since hormonal and mechanical signals for the caveolae appear to pinch off and form then, hormones as diverse as interleukin cell. Insights gained from studying poto- vesicles that sequester low molecular 2 (22), nerve growth factor (23), TGF-l31 cytosis suggest several different ways that weight tracer molecules introduced into (24), and thyroid-stimulating hormone this membrane specialization might func- the blood vessel (12-14). Similar vesicles (25) have been found to stimulate the tion to integrate incoming and outgoing have not been found during receptor- release of IPG from cells. High concen- cellular messages. mediated potocytosis, even though a trations of synthetic IPG (26), IPG from closed caveolar compartment is easily hormone-stimulated cells (21), or IPG The surfaces of most cells are studded detected by biochemical and morpholog- isolated from trypanosomes (27) can with tiny, flask-shaped membrane invagi- ical methods (11, 15). Caveolae do not mimic some ofthe effects ofeach ofthese nations called caveolae (1-5). These in- merge with other endocytic pathways but hormones. Even a GPI toxin from ma- vaginations are the preferred location for appear to remain associated with the cell laria has been identified (28). Finally, an glycosylphosphatidylinositol (GPI)-an- surface. This is significant because in antibody directed against IPG inhibits chored membrane proteins (6-8). Some many instances caveolae are grouped to- second messenger-mediated stimulation of these GPI-anchored proteins function gether in specific regions of the cell by insulin (29). These data strongly im- to concentrate small molecules and ions where they appear to open and close but plicate IPG as a messenger molecule. in the sequestered space created when remain localized. The only known cellular sources of caveolae transiently seal off from the The invagination of caveolae may be these glycans are the externally oriented, extracellular environment. The concen- controlled by the distinctive coat mate- GPI-anchored proteins and lipids of sur- trated ligands then flow into the cell rial that decorates the inner membrane face membranes (30, 31). How could an through carriers or channels embedded in surface (Fig. 1). In stromal cells such as IPG derived from these molecules ever the membrane. This process is called fibroblasts, this coat has a striated ap- get into the cell unless they were first potocytosis (9, 10). pearance (7, 16, 17), but in epithelial cells captured by an endocytic pathway? This An unexpected complexity is begin- it looks more globular (18). The caveolar is where caveolae might have an impor- ning to emerge about caveolae. In addi- coat is quite unlike other coats found on tant function. Nearly every caveolae has tion to being a site of entry for small budding membranes because it is con- a cluster of GPI anchored proteins, and molecules and ions by potocytosis, evi- structed from membrane proteins that within a cluster the protein density can dence is accumulating that they also play cannot be removed by either high salt or as as molecules a major role in coordinating the interac- high pH (7). As a consequence, the coat reach high 30,000 per tion of the cell with its environment. The may be permanently attached to caveolae ,um2 (15). Therefore, the IPG is naturally molecular composition of caveolae indi- and remain there during all phases of the at a high concentration in caveolae. Ifthe cates that they have the capability to internalization cycle. hormone-stimulated release of glycans store and process messengers such as were coordinated with the closure of the cAMP, calcium, or adenosine and to use Storage and Processing of Messengers caveolae (Fig. 2A), then just a few IPG non-receptor tyrosine kinases to initiate in Caveolae molecules generated in this space would crucial phosphorylation cascades. They be sufficient to establish a concentration may actually be able to form chemical Caveolae can either be open in direct gradient capable of driving transport synapses between non-neuronal cells. communication with the extracellular through a membrane carrier into the cell. Remarkably, the formation of these sig- space or be closed to process trapped Another possibility is that a subset of nals may be dependent on membrane molecules or store them for later use. caveolae may be kept closed and used as cholesterol, since the structure of cave- Their structure and cellular distribution IPG storage sites. Hormones such as olae is disrupted when cells are depleted suggest that caveolae are a semiperma- insulin would then allow IPG entry into of this sterol. nent component of the cell surface that In this article, I outline the data that can be moved to different locations to Abbreviations: GPI, glycosylphosphatidyli- support a role for caveolae in cell signal- meet the changing demands of the cell. nositol; IPG, inositol phosphoglycan. 10909 Downloaded by guest on September 24, 2021 10910 Review: Anderson Proc. Natl. Acad. Sci. USA 90 (1993) A - Hormone -+

Caveola

0

FIG. 1. Rapid-freeze, deep-etch view of the striated coat that decorates the inside membrane surface of each human fibroblast caveola. (Bar = 50 nm.) C the cell by opening a carrier in the cave- olar membrane. cAMP-Binding Protein. Some GPI- anchored proteins might be binding sites for molecules or ions that function as messengers. Analysis ofthe folate recep- tor in MA104 monkey kidney epithelial cells suggests that GPI-anchored, high- affinity binding sites can very effectively concentrate molecules in caveolae. For example, this receptor can sequester as many as 700 molecules of folate in each FIG. 2. A model for intracellular signaling by caveolae. (A) Hormonal transduction. caveola (15). Recently a set of related Hormones (m) such as insulin bind to their receptor on the cell surface and elicit a signal. a Caveolae respond to this signal by closing off from the cell surface and opening a membrane cAMP-binding proteins that contain channel that allows molecules or ions (o), which are either sequestered or stored in the GPI anchor have been found in mamma- compartment, to enter the cell. (B) Receptor transduction. Some GPI-anchored proteins in lian (32) and yeast (33) cells. Therefore, caveolae are hormone receptors (GPI receptors). Caveolae have Src kinases (o) on their in certain cells caveolae may function as cytoplasmic surface. Ligand (m) binding causes the GPI receptor to associate with the kinase storage and release sites for cAMP (Fig. and stimulate a phosphorylation cascade. Alternatively, stimulation occurs indirectly through 2A). an intermediate generated in caveolae after ligand binding. (C) Mechanical transduction. Membrane Carriers and Pumps for Cal- Caveolae interact with elements of the cytoskeleton through actin filaments. Stretching or a second mes- compressing the cell changes the organization of the cytoskeleton, which is transmitted to the cium. Ca2+ is ubiquitous caveolae by a change in filament tension. The force applied by the actin filaments deforms the senger (34). For some time there has been caveolae, which opens membrane channels/carriers that allow messenger molecules or ions (o) speculation that caveolae might be a Ca2+ to flow into the cytoplasm. entry site, similar to the T-tubule of mus- cle cells (35). Histochemical methods time, only the amount ofCa2+ thatis stored phorylation cascade at the cell surface have even indicated that high concentra- in the vesicle would enter the cell, which (50). Several new observations suggest tions of Ca2+ are present at these sites, would help control Ca2+ toxicity and allow that this cascade might involve mem- which suggests a storage role (36, 37). for a more precise delivery of the signal brane-bound tyrosine kinases that are Two recent reports now put these obser- (41). The quantal release of Ca2+ from attached to the cytoplasmic surface of vations on a firmer footing: a form of the caveolae could also be used to generate each caveola (Fig. 2B). inositol 1,4,5-trisphosphate-regulated Ca2+ oscillations that code for vital intra- Three different laboratories have re- Ca2+ channel (38) and a Ca2+-pump of a ATPase have been localized to caveolae cellular messages (42, 43). A final idea is ported that the immunoprecipitation (39). This places two membrane mole- based on the finding that caveolae are often GPI-anchored membrane protein copre- cules known to be crucial for regulating asymmetrically distributed on the surface cipitates tyrosine kinase activity (44, 51, Ca2+ fluxes (40) in a location where they ofthe cell. Most ofthe caveolae in migrat- 52). In each case, the kinase has been can be used to store and release this ing fibroblasts, for example, are positioned found to be a member of the Src family. messenger. These findings also suggest at the cell margin and the leading edge (7). Clustered and unclustered (8, 15) GPI- that caveolae have the capability to func- In these locations caveolae could control anchored proteins are present on the cell tion as an intracellular Ca2+ store. the spatial distribution ofcytoplasmic Ca2+ surface, and either type might be the form Packaging high concentrations of Ca2+ by regulating the entry and retrieval ofthe that interacts with these kinases. How- in a vesicle-like compartment at the cell ion at these sites. ever, the GPI-anchored clusters have re- surface could offer several unique ways to Src Kinases and Caveolin. Antibodies cently been purified and found to contain use this ion as a signal (Fig. 2A). If Ca2+ directed against several different GPI- multiple GPI-anchored proteins as well were to enter the cell from a highly con- anchored membrane proteins have been as Src-related tyrosine kinase activity centrated source such as the caveolae, found to elicit a signal response in certain (53-55). These studies did not determine then steep gradients of Ca2+ would be cells (28, 44-49). Presumably the stimu- whether the tyrosine kinase-containing created across the membrane. At the same lation involves the activation of a phos- GPI clusters were derived from caveolae. Downloaded by guest on September 24, 2021 Review: Anderson Proc. Natl. Acad. Sci. USA 90 (1993) 10911 Sargiacomo et al. (56) now have evidence that they are associated. Using antibod- ies against caveolin, a membrane marker for caveolae (7), they have developed a purification procedure that yields caveo- lin-rich fractions of membrane from Unstimulated MDCK canine kidney epithelial cells. ucleus These fractions have a 50- to 400-fold Receptors enrichment in the non-receptor tyrosine kinase, c-Yes, along with a 150-fold in- crease in the concentration of GPI- i anchored proteins. Taken together, these Receptor -- ," studies suggest that both Src-related ty- - rosine kinases and GPI-anchored pro- _;i teins are resident proteins of caveolae. Cav0eol One of the target substrates for v-Src tyrosine kinase is caveolin (57, 58). The Caveolae phosphorylation of caveolin by pp6Osrc has been linked to the malignant trans- formation of chicken embryo fibroblasts (58). This suggests that in normal cells caveolae are important for cell behavior and that in transformed cells tyrosine phosphorylation of caveolin may inter- fere with an essential signaling activity. Caveolin may function to transmit signals directly from individual GPI-anchored proteins to a resident tyrosine kinase (Fig. 2B) (56). Another possibility is that caveolin directly or indirectly controls the opening and closing of caveolae and that phosphorylation of caveolin inhibits the normal internalization cycle. This would prevent the processing and/or FIG. 3. A model for intercellular signaling by caveolae. Caveolae are sometimes grouped storage of any essential messengers that together in patches that are located at sites of cell-cell contact. Closed caveolae could function might originate in this compartment. in this location as storage sites for messenger molecules such as adenosine (o, in caveolae of 5'-Nucleotidase. Thus far in the discus- unstimulated cells). Stimulation of the cell, for example by the binding of a hormone to its sion, caveolae have been portrayed as receptor (n, stimulated), causes the caveolae to open and release the messenger into the sites that ulti- intercellular space. The messenger then binds to a specific receptor on the adjacent cell, which storage and processing initiates a signal cascade in that cell. The messenger is quickly degraded and the cascade is mately deliver messages to the inside of turned off. Thus, caveolae may be able to form chemical synapses between neuronal and the cell. They also could be used to store non-neuronal cells. signaling molecules that are released on demand into the extracellular space. Ca- closed caveolae on one cell are adjacent to can potentially control the opening and veolae sometimes aggregate together in clusters of purinoceptors on the other closing cycle as well as their spatial dis- specific regions ofa cell and form patches (Fig. 3, Unstimulated). 5'-Nucleotidase tribution. In addition, there may also be containing 10-100 densely packed units converts AMP to adenosine in caveolae, multiple mechanisms for controlling the (Fig. 1). If a patch were abutting an where it is stored until an appropriate entry into the cytoplasm of material adjacent cell, then any substance re- stimulus causes caveolae to open and re- stored in caveolae. leased from those caveolae into the ex- lease the signal (Fig. 3, Stimulated). The Cells are constantly bombarded by me- tracellular space would have only a short released adenosine interacts with the re- chanical stimuli from cells distance to travel to interact with that cell ceptor and then is quickly degraded by surrounding adenosine deami- and the extracellular matrix. These im- (Fig. 3). Adenosine is a candidate mole- another ecto-enzyme, pulses must be correctly interpreted if a cule for intercellular signaling by caveo- nase (62, 63). The result is a chemical lar patches. "synapse" between two non-neuronal cell is to maintain a proper relationship this with its neighboring cells in the tissue. Adenosine monophosphate (AMP) has cells. While entirely speculative at of the cell is recently been found to be a paracrine point, such a synapse might be capable of The shape and malleability factor secreted by neutrophils that, upon propagating messages by stimulating controlled by a cytoskeleton that is partly conversion to adenosine, stimulates C1- membrane depolarization or inducing the composed of actin filaments. Recently, secretion from nearby intestinal epithelial release of a second chemical signal. An- an actin-binding protein has been found cells (59). GPI-anchored 5'-nucleotidase, other potential function for caveolar syn- associated with caveolae (64). In addi- the ecto-enzyme responsible for convert- apses is to separate multiple signaling tion, myosin subfragment 1 is able to bind ing AMP to adenosine, is clustered in pathways that are operating in the same to a molecule in the striated caveolar coat caveolae (60). Furthermore, many differ- cell. This would provide the cell with a in vitro (17, 18). The presence of these ent cells have P1 purinoceptors that bind way to control the spatial distribution of binding sites in caveolae suggests that adenosine with high affinity (61). These signal input and output. this portion of membrane might be linked receptors are usually linked to adenylate to the cytoskeleton. If so, then mechan- cyclase and can either stimulate (A1 re- Release of Messengers from Caveolae ical perturbations of the cell could di- ceptors) or inhibit (A2 receptors) the pro- rectly cause the release of a message that duction of cAMP (61). Imagine two Caveolae may be a versatile location for is stored in this compartment (Fig. 2C). closely apposed cells where patches of messenger processing because the cell Actin may also control caveolar patch Downloaded by guest on September 24, 2021 10912 Review: Anderson Proc. Natl. Acad. Sci. USA 90 (1993)

formation at regions of the cell that are close working relationship that I enjoy with 27. Misek, D. E. & Saltiel, A. R. (1992) J. actively making contact with the substra- my colleagues, Drs. Barton Kamen, Karen Biol. Chem. 267, 16266-16273. tum or with other cells (65). These Rothberg, Eric Smart, and Stephen Lacey. 28. Schofield, L. & Hackett, F. (1993) J. patches could be positioned to optimize Exp. Med. 177, 145-153. the conversion offocal, mechanical input 1. Bundgaard, M. (1991) J. Struct. Biol. 29. Romero, G., Gamez, G., Huang, L. C., 107, 76-85. Lilley, K. & Luttrell, L. (1990) Proc. into local changes in cell behavior. 2. Severs, N. J. (1988) J. Cell Sci. 90, 341- Natl. Acad. Sci. USA 87, 1476-1480. Hormones may also control the open- 348. 30. Low, M. G. (1989) Biochim. Biophys. ing of the caveolar channels that allow 3. Palade, G. E. (1953) J. Appl. Phys. 24, Acta 988, 427-454. potential signals to be released into the 1424 (abstr.). 31. Ferguson, M. A. & Williams, A. F. cell (Fig. 2A). Inositol trisphosphate re- 4. Palade, G. E. (1958) Anat. Rec. 130, 467 (1988)Annu. Rev. Biochem. 57, 285-320. ceptors, for example, are hormonally (abstr.). 32. Muller, G., Dearey, E. A. & Punter, J. regulated Ca2+ channels. Many different 5. Yamada, E. (1955) J. Biophys. Biochem. (1993) Biochem. J. 289, 509-521. hormones are known to induce inositol Cytol. 1, 445-458. 33. Muller, G., Schubert, K., Fiedler, F. & trisphosphate formation (43) and one of 6. Rothberg, K. G., Ying, Y.-S., Kamen, Bandlow, W. (1992) J. Biol. Chem. 267, B. A. & Anderson, R. G. W. (1990) J. 25337-25346. the targets for inositol trisphosphate is Cell Biol. 111, 2931-2938. 34. Rasmussen, H. & Barrett, P. Q. (1984) this group ofrelated channel proteins that 7. Rothberg, K. G., Heuser, J. E., Physiol. Rev. 64, 938-984. control Ca2+ entry into the cytoplasm. A Donzell, W. C., Ying, Y.-S., Glenney, 35. Popescu, L. M. (1974) Studia Biophys. similar signal cascade may control the J. R. & Anderson, R. G. W. (1992) Cell 44, 141-153. entry ofother messengers via channels or 68, 673-682. 36. Suzuki, S. & Sugi, H. (1989) Cell Tissue carriers that reside in caveolar mem- 8. Ying, Y.-S., Anderson, R. G. W. & Res. 257, 237-246. branes. Rothberg, K. G. (1992) Cold Spring Har- 37. Sugi, H., Suzuki, S. & Daimon, T. (1982) Caveolae cannot carry out any of the bor Symp. Quant. Biol. 57, 593-604. Can. J. Physiol. Pharmacol. 60, 576- proposed functions unless they work 9. Anderson, R. G. W., Kamen, B. A., 587. properly. Recent evidence indicates that Rothberg, K. G. & Lacey, S. W. (1992) 38. Fujimoto, T., Nakade, S., Miyawaki, A., Science 255, 410-411. Mikoshiba, K. & Ogawa, K. (1992) J. membrane cholesterol is essential for the 10. Anderson, R. G. W. (1993) Trends Cell Cell Biol. 119, 1507-1514. normal function of this organelle (66). Biol. 3, 69-71. 39. Fujimoto, T. (1993) J. Cell Biol. 120, Lowering the cholesterol content ofa cell 11. Kamen, B. A., Wang, M. T., Streck- 1147-1157. causes disruption ofthe caveolar coat (7), fuss, A. J., Peryea, X. & Anderson, 40. Berridge, M. J. (1993) Nature (London) unclustering of GPI-anchored proteins R. G. W. (1988) J. Biol. Chem. 263, 361, 315-325. (7), and an inhibition of potocytosis (66). 13602-13609. 41. Allbritton, N. L., Meyer, T. & Stryer, L. Most likely, cholesterol helps to organize 12. Simionescu, N., Simionescu, M. & (1992) Science 258, 1812-1815. caveolar components in late portions of Palade, G. E. (1972) J. Cell Biol. 53, 42. Fewtrell, C. (1993) Annu. Rev. Physiol. the This is where GPI 365-392. 55, 427-454. Golgi apparatus. 13. Simionescu, N., Simionescu, M. & 43. Berridge, M. J. (1992) Adv. Second Mes- anchors begin to cluster (67) and it is a Palade, G. E. (1975) J. Cell Biol. 64, sengerPhosphoprotein Res. 26, 211-223. region of the Golgi that contains caveolin 586-607. 44. Shenoy-Scaria, A. M., Kwong, J., Fu- (7, 68). This sterol may act like a lipid 14. Simionescu, M., Simionescu, N. & jita, T., Olszowy, M. W., Shaw, A. S. & "glue" that holds together the protein and Palade, G. E. (1982) J. Cell Biol. 94, Lublin, D. M. (1992) J. Immunol. 149, lipid components of the caveolar mem- 406-413. 3535-3541. brane. Thus, too much or too little cho- 15. Rothberg, K. G., Ying, Y.-S., Kolhouse, 45. Korty, P. E., Brando, C. & Shevach, lesterol may influence the transmission of J. F., Kamen, B. A. & Anderson, E. M. (1991) J. Immunol. 146, 4092- messages that originate in caveolae. R. G. W. (1990) J. Cell Biol. 110, 637- 4098. 649. 46. Hahn, A. B. & Soloski, M. J. (1989) J. 16. Peters, K.-R., Carley, W. W. & Palade, Immunol. 143, 407-413. Outlook G. E. (1985) J. CellBiol. 101, 2233-2238. 47. Davis, L. S., Patel, S. S., Atkinson, 17. Izumi, T., Shibata, Y. & Yamamoto, T. J. P. & Lipsky, P. E. (1988) J. Immunol. The next step is to learn how caveolae (1988) Anat. Rec. 220, 225-232. 141, 2246-2252. help the cell to receive and interpret 18. Izumi, T., Shibata, Y. & Yamamoto, T. 48. Malek, T. R., Ortega, G., Chan, C., environmental stimuli. The vesicular ar- (1989) J. Electron Microsc. 38, 47-53. Kroczek, R. A. & Shevach, E. M. (1986) chitecture of the caveolae suggests the 19. Larner, J., Huang, L. C., Schwartz, J. Exp. Med. 164, 709-722. possibility that they are sites where con- C. F., Oswald, A. S., Shen, T. Y., 49. Thompson, L. P., Leudi, J. M., Glass, tinuous signals received from a hormonal Kinter, M., Tang, G. Z. & Zeller, K. A., Low, M. G. & Lucas, A. H. (1989)J. stimulus are converted (1988) Biochem. Biophys. Res. Commun. Immunol. 143, 1815-1821. or mechanical 151, 1416-1426. 50. Carpenter, G. (1992) FASEB J. 6, 3283- into quantal signals that control cell be- 20. Romero, G., Luttrell, L., Rogol, A., 3289. havior. The frequency, the concentra- Zeller, K., Hewlett, E. & Lamer, J. 51. Stefanova, I., Horejsf, V., Ansotegui, tion, and the type of messenger released (1988) Science 240, 509-511. I. J., Knapp, W. & Stockinger, H. (1991) from each caveola may encode the actual 21. Saltiel, A. R. & Sorbara-Cazan, L. R. Science 254, 1016-1019. message that the cell receives. If caveo- (1987) Biochem. Biophys. Res. Commun. 52. Thomas, P. M. & Samelson, L. E. (1992) lae are arranged into patches at specific 149, 1084-1092. J. Biol. Chem. 267, 12317-12322. locations on the cell surface, then these 22. Eardley, D. D. & Koshland, M. E. 53. Draberova4, L. & Draber, P. (1993) Proc. messages provide critical spatial informa- (1991) Science 251, 78-81. Natl. Acad. Sci. USA 90, 3611-3615. And when these are 23. Ohmichi, M., Decker, S. J. & Saltiel, 54. Bohuslav, J., Cinek, T. & Horejsf, V. tion. messages A. R. (1992) J. Biol. Chem. 267, 21601- (1993) Eur. J. Immunol. 23, 825-831. passed between cells they become an 21606. 55. Cinek, T. & Horejsf, V. (1992) J. Immu- important form of intercellular commu- 24. Vivien, D., Petitfrere, E., Martiny, L., nol. 149, 2262-2270. nication. Caveolae may be "message Sartelet, H., Galera, P., Haye, B. & 56. Sargiacomo, M., Sudol, M., Tang, Z. & centers" for the cell. Pujol, J. P. (1993) J. Cell Physiol. 155, Lisanti, M. P. (1993) J. Cell Biol. 122, 437-444. 789-808. _I thank Drs. Michael S. Brown, Joseph L. 25. Jacquemin, C. (1991) Biochimie 73, 37- 57. Glenney, J. R. & Zokas, L. (1989) J. Cell Goldstein, Fredrick Grinnell, George Bloom, 40. Biol. 108, 2401-2408. and William Snell for all of their helpful com- 26. Plourde, R., d'Alarcao, M. & Saltiel, 58. Glenney, J. R. (1989) J. Biol. Chem. 264, ments during the preparation of this manu- A. R. (1992) J. Org. Chem. 57, 2606- 20163-20166. script. These ideas have evolved from the 2616. 59. Madara, J. L., Patapoff, T. W., Gillece- Downloaded by guest on September 24, 2021 Review: Anderson Proc. Natl. Acad. Sci. USA 90 (1993) 10913

Castro, B., Colgan, S. P., Parkos, C. A., 63. Casad6, V., Lluis, C., Canela, E., 66. Chang, W.-J., Rothberg, K. G., Kamen, Delp, C. & Mrsny, R. J. (1993) J. Clin. Franco, R. & Mallol, J. (1992) Neuro- B. A. & Anderson, R. G. W. (1992) J. Invest. 91, 2320-2325. chem. Res. 17, 129-139. Cell Biol. 118, 63-69. 60. Forsman, C. A. & Gustafsson, L. E. 64. Rogalski-Wilk, A. A., Skepner, A., 67. Brown, D. A. & Rose, J. K. (1992) Cell (1985) J. Neurocytol. 14, 551-562. Venugopal, G. & Donati, R. (1992) Mol. 68, 533-544. 61. Olsson, R. A. & Pearson, J. D. (1990) Biol. Cell 3, 268a (abstr.). 68. Kurzchalia, T. V., Dupree, P., Parton, Physiol. Rev. 70, 761-845. 65. Izumi, T., Shibata, Y. & Yamamoto, T. R. G., Kellner, R., Virta, H., Lehnert, 62. Andy, R. J. & Kornfeld, R. (1982) J. (1991) J. Electron Microsc. Tech. 19, M. & Simons, K. (1992) J. Cel Biol. 118, Biol. Chem. 257, 7922-7925. 316-326. 1003-1014. Downloaded by guest on September 24, 2021