I. Introduction

Top Previous Next References

Well before the era of molecular biology, electron microscopists of the 1950s were describing the ultrastructural components of the cell. Among these were 50- to 100-nm invaginations of the plasma membrane referred to as either caveolae intracellulare, due to their cave-like appearance, or plasmalemmal vesicles (Palade, 1953; Yamada, 1955). It would be approximately forty years before the molecular nature of caveolae could be explored following the identification of its signature protein, caveolin. Since that time, the field of caveolin/caveolae research has blossomed, with caveolae being implicated and demonstrated to be important in a variety of cellular functions including endocytic processes, cholesterol and lipid homeostasis, signal transduction, and tumor suppression. Although caveolae and the caveolins are continuously implicated in an increasing array of cellular processes, it is clear that their physiological roles are vastly different depending on the cell type and organ system examined. For example, their endocytic and vasoregulatory functions likely predominate in the vasculature, whereas they play an important role in the structural integrity of the musculature. In this regard, insights into caveolar function will not only be interesting from the standpoint of cell biology but will be rewarding in understanding mammalian physiology with applications to human disease.

In accordance with increasing knowledge and understanding, the subject of caveolae and the caveolins has been the focus of numerous review articles, with most confined to certain aspects of their function (Parton, 1996; Anderson, 1998; Okamoto et al., 1998; Kurzchalia and Parton, 1999; Smart et al., 1999; Razani et al., 2000b; Schnitzer 2001). Recently, the field has been invigorated by the characterization of caveolin/caveolae-deficient mouse models, thus for the first time enabling investigators to analyze the cellular functions of caveolae with respect to mammalian physiology. As a consequence, it is the purpose of this review to provide a broad overview of the field with detailed discussions on its salient features and to provide a link between the understanding of caveolae at the cellular level and their emerging roles at the organismal level.

	II. Caveolae

Top Previous Next References

Originally caveolae were given the exclusive electron microscopic description of membrane invaginated "smooth" vesicles of 50 to 100 nm in size (as opposed to the more electron-dense and larger "coated" vesiclesi.e., clathrin-coated pits). However, with further investigation the definition of caveolae has expanded to also include vesicles detached from the plasma membrane proper, associated in groups as grape-like clusters or rosettes, and even in fused form as elongated tubules or trans-cellular channels (Fig. 1A). Intriguingly, these "nontraditional" forms of caveolae are more often found in certain tissues. For example, grape-like clusters of caveolae are highly abundant in developing skeletal muscle cells, with rosettes in adipocytes, and detached vesicles/tubules in endothelial cells (Simionescu et al., 1975; Scherer et al., 1994; Parton et al., 1997). Figure 1B shows an electron micrograph of an adipocyte with the several rosette structures and an endothelial cell with traditional caveolae and detached plasmalemmal vesicles. Therefore, as more is learned about these structures, caveolae can no longer be considered static in-pocketings of the plasma membrane, but can take on disparate shapes and forms by conglomerating and/or fusing with one another. The processes involved in the formation of these various caveolar morphologies are largely unknown.

View larger version (39K): [in this window] [in a new window]   Fig. 1.   Caveolae: morphological definitions and plasticity. A, caveolae can exist in many alternate forms than the traditional membrane-invaginated variety. They can be found in vesicular form or in aggregates such as grape-like clusters, rosettes, and even elongated tubules. In fact, in endothelial cells, the elongated tubes can sometimes adjoin with the abluminal side, thus creating a trans-cellular channel. B, two electron micrographs, an adipocyte (shown on the left) and an endothelial cell (shown on the right) were taken at a magnification of 16,000×. Caveolar rosettes are frequently observed in adipoctyes, and endothelial cells are often found with caveolae in all states of membrane associationfrom membrane-bound to fully invaginated.

At the ultrastructural level, morphologically identifiable caveolae can be found at the plasma membrane of numerous tissues and cell types, albeit in vastly differing abundance. A compilation of morphological data from the past several decades reveals that although most cell types contain some caveolae, certain cells have an extraordinary abundance of caveolae, namely adipocytes, endothelial cells, type I pneumocytes, fibroblasts, smooth muscle cells, and striated muscle cells (Palade, 1953; Napolitano, 1963; Mobley and Eisenberg, 1975; Gabella, 1976; Gil, 1983). It has been reported that in tissues such as the lung up to 70% of the alveolar plasma membrane area (composed of type I pneumocytes and endothelial cells) could be occupied by caveolae (Gil, 1983). However, it should be noted that certain cells are curiously devoid of these invaginations (e.g., central nervous system neurons and lymphocytes) (Fra et al., 1994; Cameron et al., 1997). The reasons for such wide-ranging tissue- and cell-type-specific caveolae expression are not yet known. However, as will be discussed below, the answer might lie in the fact that some of the proposed functions for caveolae would suit certain cell types better than others.

Traditionally, the lipid bilayer of the plasma membrane has been viewed as a two-dimensional "fluid mosaic" (Singer and Nicolson, 1972). In this so-called "liquid-disordered" or "liquid-crystalline" state, the plasma membrane is mostly composed of loosely packed phospholipids capable of rapid lateral diffusion. It has been known for some time, however, that membranes also exist in "liquid-ordered" states, where bilayer assembly is more rigid with confined movement of lipids. Based on a variety of observations, it is now becoming clear that liquid-ordered and -disordered states can coexist on the same plasma membrane and that liquid-ordered domains can remain insular and maintain their relative rigidity among the neighboring phospholipid bilayer. Such domains form via a coalescence of cholesterol and sphingolipids (glycosphingolipids and sphingomyelin) and thus have been termed lipid rafts (Brown and London, 1998; Simons and Toomre, 2000) (Fig. 2A).

View larger version (127K): [in this window] [in a new window]   Fig. 2.   Detailed organization of lipid rafts and caveolae membranes. A, lipid rafts: the liquid-ordered phase is dramatically enriched in cholesterol (shown in yellow) and exoplasmic oriented sphingolipids (sphingomyelin and glycosphingolipids) (shown in orange). In contrast, the liquid-disordered phase is composed essentially of phospholipids, such as phosphatidylcholine, phosphatidylethanolamine, and phosphatidylserine (shown in green). B, caveolae: the liquid-ordered and liquid-disordered phases are illustrated as in panel A. Upon integration of the caveolin-1 protein, liquid-ordered domains form small flask-shaped invaginations called caveolae. Caveolin-1 monomers assemble into discrete homo-oligomers (shown as dimers for simplicity) containing ~14 to 16 individual caveolin molecules. Adjacent homo-oligomers are thought to pack side-by-side within caveolae membranes thereby providing the structural meshwork for caveolae invagination. Caveolin-1 oligomers are red and the caveolin-1 oligomerization domain is shown in blue.

Given their shared biochemical properties, caveolae have traditionally been considered a specialized form of lipid raft (i.e., an invaginated/vesicular form) (Fig. 2B) (Brown and London, 1998; Simons and Toomre, 2000). This generalization is probably not entirely accurate as it is now known that certain proteins preferentially partition into lipid rafts or caveolae but not both (Liu et al., 1997); the reader is referred to a more detailed review of lipid rafts (Brown and London, 1998; Simons and Toomre, 2000). The unusual lipid composition of lipid rafts/caveolae imparts particular properties to these microdomains, namely a highly reduced density compared with their phospholipid counterparts and resistance to solubilization by mild nonionic detergents such as Triton X-100 at 4°C. These properties form the basis for the biochemical identification, purification, and characterization of lipid rafts/caveolae (Brown and Rose, 1992; Lisanti et al., 1994b). For example, one of the simplest and most commonly used purification techniques (sucrose gradient ultracentrifugation) utilizes the detergent resistance and buoyancy of these microdomains to separate them from all other cellular constituents (Lisanti et al., 1994b).

Although caveolae and lipid rafts share certain biochemical properties, the localization of the caveolin proteins to caveolae distinguishes these membrane domains. The caveolins serve as selective markers for caveolae (Fig. 2B) and thus allow for the specific analysis of caveolae function.

	III. The Caveolin Gene Family

Top Previous Next References

The discovery of caveolin, the original member of the three-protein caveolin family, and its relationship to caveolae was converged upon by investigators from different fields with disparate research interests. In an antibody screen for tyrosine-phosphorylated substrates in Rous sarcoma virus-transformed fibroblasts, Glenney and Zokas (1989) isolated four predominant proteins. Tyrosine phosphorylation of one of these proteins, a 22-kDa molecule, was strongly correlated with the transformation potential of Rous sarcoma virus-transformed cells suggesting a potential role in cellular oncogenesis (Glenney, 1989). Interestingly, antibodies raised against this protein showed punctate staining on the plasma membrane, with a tendency for staining at cellular margins or linear membrane arrays (Glenney and Zokas, 1989; Rothberg et al., 1992). This nonrandom plasma membrane distribution was curiously similar to that observed for the flask-shaped caveolae (i.e., concentrations along the leading edge and the linear stress fibers of cells) and led Rothberg and colleagues to a series of landmark experiments linking this 22-kDa protein and caveolae (Rothberg et al., 1992). The answer was arrived at ultrastructurally when immunogold electron microscopy indicated nearly complete association between the 22-kDa protein and caveolae. More importantly, this protein also seemed to be a major structural component of caveolae. Using rapid-freeze deep-etch techniques, caveolae were shown to be composed of a series of concentric striated rings that stained with antibodies directed against the 22-kDa protein. Moreover, treatment of cells with cholesterol binding agents, such as nystatin and filipin, had profound effects on caveolar morphology, leading to flattening of the vesicles and dissociation of the 22-kDa protein-rich striations. Because this protein was so intimately associated with the structural components of caveolae, it was named caveolin, the first bona fide marker of caveolae microdomains (Rothberg et al., 1992).

The subsequent cloning of the caveolin gene led to yet another surprise about the possible functions of the protein (Glenney, 1992; Glenney and Soppet, 1992). In an attempt to identify the cellular machinery involved in the differential sorting of vesicles to the apical or basolateral surface of polarized epithelial cells, Simons and colleagues had cloned VIP-21¹ (vesicular integral protein of 21 kDa), an integral membrane protein component of trans-Golgi-derived transport vesicles (Kurzchalia et al., 1992). As it turned out, the caveolin sequence was identical to that of VIP-21, thereby showing that the same protein could possibly serve as a structural component of plasma membrane caveolae, as well as have roles in oncogenesis and vesicular traffickingall at the same time (Glenney 1992). This series of discoveries marked the beginning of the past decade's investigations into the intimate role between caveolae and caveolin and their intricate functions, topics that we will in turn discuss below.

Since the initial characterization of caveolin, two additional caveolin gene family members have been discovered (caveolin-2 and -3); thus, the original caveolin protein is now known as caveolin-1 (Cav-1) (Way and Parton, 1995; Scherer et al., 1996; Tang et al., 1996). Caveolin-2 (Cav-2) was cloned in an effort to identify other novel resident proteins of adipocyte caveolae (where there is an abundance of caveolae and caveolin-1); caveolin-3 (Cav-3) was found by classic cDNA library screening for Cav-1 homologous genes. In addition to the full-length proteins (-isoform), both caveolin-1 and -2 have other smaller sized isoforms. The -isoform of Cav-1 is derived from an alternative translational start site that creates a protein truncated by 32 amino acids (Scherer et al., 1995). The - and -isoforms of Cav-2 have not been analyzed in detail.

An alignment of the human caveolin sequences and an overview of the three caveolin proteins is provided in Fig. 3 and Table 1. There is a relatively high degree of identity between Cav-1 and -3, whereas Cav-2 is by far the most divergent of the three proteins. The highest degree of identity among all three caveolins in the majority of species examined is a stretch of amino acids "FEDVIAEP" and is now referred to as the "caveolin signature motif". The structural/functional significance of this motif remains unknown. Other important domains in the caveolin sequences are the membrane-spanning, oligomerization, and scaffolding domains, all of which will be discussed below (Fig. 3).

View larger version (71K): [in this window] [in a new window]   Fig. 3.   Sequence alignment of the human caveolin gene family. An alignment of the protein sequences of human caveolin-1, -2, and -3 is shown. Identical residues are boxed and highlighted in red. Note that caveolin-1 and -3 are most closely related, while caveolin-2 is divergent. Translation initiation sites are circled. In addition, the positions of the membrane-spanning segment (green), the oligomerization domain (blue), and the scaffolding domain (a subregion of the oligomerization domainhashed blue) are indicated.

All three proteins are present at their highest levels in terminally differentiated cells. As all caveolins are classically known as markers of caveolae, it is not surprising that tissues expressing caveolin-1, -2, and/or -3 are also abundant in caveolae. Interestingly, the expression pattern of caveolin-1 and -2 are largely distinct from that of caveolin-3. Adipocytes, endothelial cells, pneumocytes, and fibroblasts have the highest levels of caveolin-1 and -2, whereas caveolin-3 expression is limited to muscle cell types (i.e., cardiac, skeletal, and smooth muscle cells) (Table 1) (Scherer et al., 1994, 1996; Tang et al., 1996). Certain cell types (namely smooth muscle) have expression of all three proteins possibly due to the hybrid fibroblastic/muscle-like nature of this cell. The expression profiles of these proteins in mammalian tissues is intriguing in light of the following observations: 1) despite the divergence of the Cav-2 sequence, it is selectively detected in Cav-1 expressing tissues; and 2) the highly non-overlapping expression patterns of Cav-1 and -3 indicates that the two proteins might have disparate functions in vivo in spite of their high degree of identity.

1. Caveolin Membrane Topology. Early work on caveolin-1 showed that it is resistant to extraction with sodium carbonate, a property indicative of its integral association with the plasma membrane (Kurzchalia et al., 1992; Sargiacomo et al., 1993). Based on various topological analyses, it is now well accepted that the N and C termini of Cav-1 remain cytoplasmic. For example, i) antibodies recognizing the N and C termini of Cav-1 can bind the protein only when cells are permeabilized with detergents (Dupree et al., 1993); ii) membrane-associated Cav-1 remains sensitive to proteolysis, a condition indicative of cytoplasmically directed N and C termini (Monier et al., 1995); and iii) there are known phosphorylation and palmitoylation sites in the N-and C-terminal domains of Cav-1, respectively (Dietzen et al., 1995; Li et al., 1996b). By extension, it is reasonable to assume that Cav-1 is a double-pass membrane-spanning protein. However, cell-surface biotinylation of cells shows no labeling of caveolin-1, indicating the inaccessibility of its domains to the extracellular milieu (Sargiacomo et al., 1995). Analysis of the Cav-1 protein sequence indicates that it contains a single hydrophobic region (residues 102-134). As a 32-amino acid stretch is not long enough to enable a complete hairpin loop through the lipid bilayer, a topological view of Cav-1 that took into account its tight association with the plasma membrane was a conundrum.

View larger version (35K): [in this window] [in a new window]   Fig. 4.   The current view of caveolin-1 membrane topology. Caveolin-1 exists as either a homo-oligomer of ~14 to 16 monomers (shown as a dimer for simplicity) or as a hetero-oligomer with caveolin-2 (not shown). Via its hydrophobic trans-membrane (TM) domain (red), Cav-1 is believed to penetrate the membrane. The protein is also bound to membranes through tight association between the N-MAD and C-MAD (shown in lavender and green, respectively). Homo-oligomerization is mediated by a 40-amino acid stretch (residues 61-101; which incidentally encompasses N-MAD) known as the oligomerization domain (OD) (hashed brown). Adjacent oligomers interact via the terminal domain (TD) (purple). Sites of palmitoylation (Cys133, -143, and -156) are shown in blue.

2. Caveolin Oligomerization. Using velocity gradient centrifugation, it was first discovered that caveolin-1 migrates as a high molecular weight complex of approximately 350 to 400 kDa in vivo (Monier et al., 1995; Sargiacomo et al., 1995). This complex was exclusively composed of the Cav-1 protein and was dissociated only under harsh detergent treatment at elevated temperatures, indicating the presence of a highly stable caveolin-1 homo-oligomer of approximately 14 to 16 monomers (Monier et al., 1995; Sargiacomo et al., 1995; Li et al., 1996c). Furthermore, pulse-chase analysis showed that the complex was formed relatively rapidly after synthesis of Cav-1 in the ER and prior to completion of Golgi transit (Monier et al., 1995).

3. Structural Relationships between the Caveolins. Although it is well established that Cav-1 can form large homo-oligomeric complexes in vivo, Cav-1 has recently been shown to undergo higher ordered interactions with caveolin-2 as well. Cav-1 and -2 can interact to form very stable high molecular mass hetero-oligomers, akin to the large homo-oligomeric complexes observed for Cav-1 (Scherer et al., 1997). Interestingly, in the absence of Cav-1, Cav-2 is not capable of forming large homo-oligomers and rather exists in a monomeric/dimeric form and is retained at the level of the Golgi compartment (Scherer et al., 1996; Mora et al., 1999; Parolini et al., 1999). Upon Cav-1 expression, the Cav-1 and Cav-2 proteins form a large complex and are found predominantly localized to plasma membrane caveolae (Scherer et al., 1996; Mora et al., 1999; Parolini et al., 1999). Therefore, Cav-2 is dependent on caveolin-1 both structurally and for appropriate subcellular trafficking (Mora et al., 1999; Parolini et al., 1999). Based on the primary sequence, the oligomerization domain of Cav-2 is ~50% identical/75% similar to that of Cav-1, and it was believed that oligomer formation between the two proteins is mediated by their respective oligomerization domains. However, recent deletion mapping studies surprisingly indicate that the interaction is mediated by the membrane-spanning domains of Cav-1 and -2 (Das et al., 1999). The tissue-specific expression of Cav-1 and Cav-2 is highly coregulated (i.e., all tissues with Cav-1 expression are also the primary sites of Cav-2 expression) (Scherer et al., 1996). Therefore, despite the well established fact that Cav-1 can form large homo-oligomeric complexes in vivo, it seems more likely that the Cav-1 protein is constitutively associated with Cav-2 under physiological conditions.

	IV. Caveolar Biogenesis

Top Previous Next References

The crucial role of cholesterol in caveolar biogenesis has been evident for some time. Treatment of cells with cholesterol binding agents such as nystatin, filipin, or cyclodextrin leads to a complete ablation of morphologically identifiable caveolae (Rothberg et al., 1992; Schnitzer et al., 1994; Hailstones et al., 1998). Furthermore, absolute cellular levels of cholesterol need to rise above a certain threshold level before caveolae formation can occur (Hailstones et al., 1998). Although caveolae microdomains are highly enriched in cholesterol, the selective requirement for cholesterol in caveolar biogenesis is not straightforward. Recent reports have clearly shown that manipulation of cellular cholesterol levels also has effects on the biogenesis of clathrin-coated endocytic vesicles and synaptic vesicles, membrane entities that differ in many respects from lipid rafts and caveolae (Rodal et al., 1999; Thiele et al., 2000). If the formation of many types of post-Golgi-derived vesicles depends on critical levels of cholesterol, the formation of caveolae must involve more than just the role played by cholesterol.

Overexpression of caveolin-1 in cells lacking endogenous caveolin/caveolae (lymphocytes and transformed fibroblasts) results in the de novo production of ~50- to 100-nm membrane invaginations and vesicles (Fra et al., 1995b; Engelman et al., 1997; Li et al., 1998). Conversely, the targeted down-regulated caveolin-1 in cells with endogenous caveolae results in the loss of caveolae (Galbiati et al., 1998; Liu et al., 2001). These experiments clearly link caveolin expression with caveolae formation.

Several characteristics of caveolin-1 suggest the way in which it transforms the morphology of the plasma membrane into flask-shaped caveolae. Cav-1 binds the two primary components of lipid rafts, namely cholesterol and sphingolipids, both in vitro and in vivo (Fra et al., 1995a; Murata et al., 1995; Thiele et al., 2000). In fact, cholesterol has an unusually high affinity toward Cav-1, remaining associated with caveolin oligomers even in the presence of harsh detergents such as SDS (Murata et al., 1995). A high local concentration of cholesterol and caveolin maintained at a critical level could provide the appropriate lipid-protein microenvironment for membrane invagination (Fig. 5).

View larger version (32K): [in this window] [in a new window]   Fig. 5.   Caveolin-1 oligomerization and caveolae biogenesis. At the level of the ER, caveolin-1 self-associates to form high molecular mass homo-oligomers that contain ~14 to 16 individual caveolin-1 molecules. These caveolin-1 oligomers represent the functional assembly units of caveolae. Then, these caveolin-1 homo-oligomers undergo a second stage of oligomerization during transport to the plasma membrane, most likely at the level of the trans-Golgi. In this second stage of oligomerization, individual caveolin-1 oligomers interact with each other via their C-terminal domains, forming an extensive network or meshwork on the underside of the plasma membrane. This large meshwork of oligomers may act synergistically with cholesterol (yellow) to distort the membrane and to drive the invagination of caveolae.

The tendency of caveolins to form high molecular weight oligomeric complexes likely contributes to caveolae formation. As mentioned before, caveolin-1 is able to form extremely strong detergent-resistant homo-oligomeric complexes composed of ~14 to 16 monomers (Monier et al., 1995; Sargiacomo et al., 1995). If these oligomers are purified from plasma membrane lysates and detergents are subsequently removed by dialysis, the oligomers self-associate into very large/sedimentable complexes of ~20- to 40-nm in diameter (Sargiacomo et al., 1995). Furthermore, deletion analysis has revealed that the C-terminal domain of Cav-1 (namely residues 135-178), is able to interact with the full-length Cav-1 molecule (Song et al., 1997b; Schlegel and Lisanti, 2000). Taken together, this evidence suggests that caveolin oligomers can form higher order oligomer-oligomer complexes by interactions of the respective C-terminal domains. This meshwork of caveolin protein complexes most likely elicits the bending of the plasma membrane, a process similar to that of the clathrin triskelion in coated-pit formation (Fig. 5).

	V. Functional Significance of Caveolae/Caveolins

Top Previous Next References

Since their discovery in the 1950s, the function of caveolae has been a matter of speculation. Consequently, numerous roles have been attributed to these microdomains and their caveolar marker protein, the caveolins. Here, we will discuss the most thoroughly studied roles of caveolae and their potential relevance to physiology.

A. Vesicular Transport

1. Transcytosis. Based purely on the vesicular morphology and abundance of caveolae in endothelial cells (Palade 1953), caveolae were hypothesized to function as conduits for the general transport of proteins through capillary cells (i.e., transcytosis) (Fig. 6) (Simionescu et al., 1975). The primary approach in these initial reports and numerous follow-up studies involves the use of labeled "tracer" serum proteins (e.g., gold-conjugated albumin) and following their dynamic interaction with the capillary endothelium via transmission electron microscopy (Simionescu et al., 1975; Ghitescu and Bendayan, 1992; Schnitzer et al., 1994; Predescu et al., 1997, 1998). Despite very convincing time-lapse imaging of caveolae-mediated tracer transport to the abluminal side of endothelial cells, the functionality of caveolae as transcytotic vesicles has been contentious. It is not clear whether this observed mode of transport is specific to caveolae or is simply a nonspecific component of bulk fluid flow, which would include paracellular transport. Furthermore, it has been argued that the observed tracer labeling of invaginated caveolae (i.e., cytoplasmic caveolae in transit to the abluminal side) could actually be labeling of a continuation of the plasma membrane only appearing to be intracellular due to microscopic sectioning artifacts.

View larger version (34K): [in this window] [in a new window]   Fig. 6.   Caveoale vesicular trafficking: transcytosis, endocytosis, and potocytosis. Caveolae appear to mediate the selective uptake and transport of several molecules via different processes (transcytosis, endocytosis, and potocytosis). In transcytosis, caveolae transport proteins from the luminal side of the endothelial cell to the interstitial compartment for subsequent uptake by underlying tissues. In caveolae-mediated endocytosis (distinct from that of clathrin-coated pits), caveolae bud off from the plasma membrane and fuse with various intracellular compartments. Possible transport routes include the recently characterized caveolae-caveosome-ER pathway. In potocytosis, caveolae mediate the uptake of small solutes (<1 kDa) by pinching off but remaining associated with the plasma membrane. The molecular machinery involved in this caveolar fission/fusion is the same as that used for numerous other vesicular transport processes with requirements for dynamin, VAMP, SNAP-25, the SNARE complex, and GTP hydrolysis.

2. Endocytosis. Caveolae have also been implicated in endocytic processes. Traditionally, endocytosis has been almost entirely associated with clathrin-coated vesicle transport and the molecular mechanisms underlying this process have been studied in detail (for review, see Takei and Haucke, 2001). Numerous receptors and their cognate ligands are taken up via this pathway leading to termination or desensitization of signaling cascades, among other things. However, it has become clear that certain receptors and extracellular macromolecules are exclusively transported via caveolae rather than clathrin-coated pits (Fig. 6). The first demonstration of this selectivity came when investigators found that caveolae can bind and internalize cholera and tetanus toxins (Montesano et al., 1982). Biochemical evidence bolstered this argument when it was shown that treatment of cells with cholesterol binding agents (nystatin or filipin) abrogated the endocytosis of certain macromolecules (e.g., albumin) without affecting the uptake of clathrin-dependent ones (Schnitzer et al., 1994). Similar results have been seen with several membrane-bound proteins including alkaline phosphatase (Montesano et al., 1982; Anderson et al., 1992; Parton et al., 1994). This selectivity is in large part due to the caveolar localization of the cognate receptor for these molecules [e.g., the glycosphingolipid GM1 in the case of cholera toxin or the glycosylphosphatidylinositol (GPI) modification of alkaline phosphatase] (Fig. 6) (see also Table 2 for a more complete list of caveolae-localized proteins).

3. Mechanisms of Endocytosis/Transcytosis. A variety of proteins have been identified that comprise the vesicular fission/fusion machinery required for endocytosis and transcytosis. In a series of studies primarily by Schnitzer and colleagues, it was shown that caveolae harbor the same components used for the budding and docking of other vesicles (Schnitzer et al., 1995b). Early indications for the presence of such classical transport machinery was based on the observation that N-ethylmaleimide (NEM) can inhibit the transcytosis/endocytosis of endothelial cell caveolae (Predescu et al., 1994; Schnitzer et al., 1995a). Purification of these caveolar microdomains showed that several components used by cells in general vesicle formation, docking, and fusion (e.g., NSF, SNAP, VAMP, and GTPases) are concentrated in caveolae and possibly associated with caveolin-1 (Schnitzer et al., 1995b; Predescu et al., 2001) (see also Table 2). Furthermore, cleavage of VAMP by specific neurotoxins abrogates the endocytosis and fusion of caveolae with intracellular compartments (McIntosh and Schnitzer, 1999). Therefore, it has become clear that caveolae can co-opt the same mechanisms used in the trafficking of other vesicles to transport cargo throughout the cell, e.g., from the plasma membrane to internal compartments (Fig. 6).

4. Potocytosis. Caveolae have also been implicated in a unique form of solute uptake, termed potocytosis (or "cellular drinking"). First championed by Anderson and colleagues (1992), potocytosis is the process by which cells can transport small molecules (<1 kDa) without having to endocytose vesicles to internal endosomal/lysosomal compartments. Using microscopy, they demonstrated that the folate receptor, a GPI-linked protein, primarily localizes to caveolae (Rothberg et al., 1990). Remarkably, caveolae then seem to facilitate the uptake of folate by closing their necks, albeit remaining associated with the plasma membrane (Anderson et al., 1992). In this way, a highly concentrated pool of folate is created, which is subsequently fluxed to the cytoplasm by a three-step process: 1) lowering of caveolar pH by a caveolae-localized proton pump, 2) low pH dissociation of folate from its receptor, and 3) flux into the cytoplasm by a caveolae-localized anion carrier (Kamen et al., 1991) (Fig. 6). At present, the mechanism of the dynamic closing and opening of the caveolar "mouth" is not understood. Furthermore, there is controversy as to whether cells even use potocytosis in the uptake of small molecules or simply co-opt classical receptor-mediated endocytic processes (Mayor et al., 1998). More detailed analysis of the general applicability of caveolae to the uptake of other small molecules needs to be conducted before potocytosis can be considered a process distinct from caveolar endocytosis.

As briefly described above, a unique relationship exists between caveolins/caveolae and cholesterol. Ever since the discovery of caveolin-1 as a marker of caveolae, it has been known that caveolae are highly sensitive to cholesterol depletion; treatment of cells with cholesterol binding agents flattens these structures and disaggregates their caveolin-rich "striated coats" (Rothberg et al., 1992). The first indication of a direct link between the caveolins and cholesterol was demonstrated in vitro, where purified caveolin-1 was shown to reconstitute only into lipid vesicles containing cholesterol (Murata et al., 1995; Li et al., 1996c). Further analysis found that Cav-1 can actually form a complex with cholesterol in these lipid mixtures, on the order of ~1 to 2 cholesterol molecules per caveolin molecule (Murata et al., 1995). Recently, a photo-activatable form of cholesterol was shown to cross-link a ~21- to 24-kDa band in cells (identified as Cav-1), thereby making caveolin-1 one of only a few proteins with a demonstrated ability to bind cholesterol in vivo (Thiele et al., 2000). Not surprisingly, the intricate relationship with cholesterol renders Cav-1 highly sensitive to cholesterol perturbations, both in terms of its subcellular localization and transcriptional regulation.

1. The Effect of Cholesterol on Caveolin-1. Treatment of cells with cholesterol oxidase, which selectively converts caveolar cholesterol to cholestenone, causes the movement of caveolin-1 to the ER and subsequently to the Golgi compartment; removal of cholesterol oxidase allows the return of Cav-1 and free cholesterol to caveolae (Smart et al., 1994; Conrad et al., 1996). Although it is not known how Cav-1 can retreat intracellularly, it is clear that its subcellular transport is intricately linked to cholesterol transport.

2. Intracellular Transport of de Novo Synthesized cholesterol. De novo synthesis of cholesterol occurs in the ER, and upon membrane partitioning, it is rapidly transported to various membrane compartments, of which the plasma membrane is a major site (containing up to 90% of cellular cholesterol at steady state) (Simons and Ikonen, 2000). However, much less is known about the molecular machinery and trafficking pathways mediating this cholesterol transport. Based on a variety of observations, it is believed that cholesterol is predominantly transported to the cell surface by a Golgi-independent route (DeGrella and Simoni, 1982; Kaplan and Simoni, 1985; Urbani and Simoni, 1990; Heino et al., 2000). Considering its high binding capacity for cholesterol, caveolin-1 is a potential mediator of this function.

View larger version (28K): [in this window] [in a new window]   Fig. 7.   The role of caveolin-1 in the efflux/influx of cholesterol and intracellular cholesterol trafficking. Intracellular free cholesterol is derived from two main sources (externally from LDL by either receptor-mediated endocytosis or by membrane loading) or from de novo synthesis. This cholesterol is then shuttled throughout the cell to all membrane compartments. Caveolin-1 can then bind cholesterol in the ER and transport it to caveolae at the plasma membrane. At this juncture, the cholesterol can either be effluxed to extracellular lipoproteins (namely HDL) or be siphoned into the bulk plasma membrane. Caveolin-1 can then retreat to the ER/Golgi compartments and repeat this transport cycle.

3. Cholesterol Efflux from Cells. Cellular cholesterol is typically derived from two major sources: de novo synthesis (as described above) or exogenous uptake primarily from serum low-density lipoproteins (LDLs) (Fielding and Fielding, 1997). To maintain steady levels intracellularly, peripheral cells can also efflux excess free cholesterol from the exoplasmic leaflet of the plasma membrane to high-density lipoproteins (HDLs) for delivery to the liver, a process known as reverse cholesterol transport (Fielding and Fielding, 1997). In a series of studies by Fielding and Fielding, it appears that caveolae play an important role in this process as well. If the intracellular cholesterol pool is increased above baseline by incubating cells with cholesterol-donating LDL, excess cholesterol is eventually observed in caveolae; caveolae then act as portals for the efflux of this cholesterol upon incubation of cells with HDL (Fig. 7) (Fielding and Fielding, 1995). Caveolin-1 regulates this process, because antisense-mediated down-regulation of caveolin-1 decreases cellular cholesterol efflux (Fielding et al., 1999; Arakawa et al., 2000). In support of these data, in vivo expression of caveolin-1 in the mouse liver via adenoviral-based strategies causes an increase in plasma HDL cholesterol levels (Frank et al., 2001). Therefore, it appears that Cav-1 enhances the availability and delivery of cholesterol to plasma membrane caveolae for subsequent efflux. At present however, the molecular mechanisms of this efflux process from caveolae to HDL are not known, but most likely involve an association between the HDL particle or its major apolipoprotein, ApoAI, and caveolae.

C. Signal Transduction Mechanisms

1. Caveolae As Signalosomes: Compartmentalized Signaling. Caveolin-1 was not only the first protein to be localized to caveolae but due to its apparent involvement in the structural integrity of caveolae was also the first caveolar "marker protein" (Rothberg et al., 1992). The issue that required clarification, however, was whether caveolae could also serve as platforms for the aggregation and/or concentration of other proteins. Clearly, evidence for the presence of other caveolar resident proteins would be important in the understanding of caveolar function. In this regard, Lisanti and coworkers were the first investigators to broadly address this issue (Sargiacomo et al., 1993; Lisanti et al., 1994b). Using the insolubility of caveolae in mild detergents and their buoyancy in sucrose gradients, they were able to biochemically separate caveolae membranes and, in turn, determine the identity of cosegregated proteins. Of the numerous proteins identified in this manner, it was surprising to find that a large majority were signal transduction molecules, some at concentrations manyfold higher than the bulk plasma membrane (Sargiacomo et al., 1993; Lisanti et al., 1994b). This observation led Lisanti and colleagues to put forth the "caveolae/raft signaling hypothesis": the compartmentalization of such molecules has distinct advantages as it provides a mechanism for the regulation of subsequent signaling events and explains cross-talk between different signaling pathways (Lisanti et al., 1994a).

2. The Caveolins As Modulators of Signaling. A topic neglected by this signaling concept was the functional significance of the caveolin proteins (were they simply structural bystanders or could they actively contribute in the retention and modulation of signal transduction proteins). After all, the concept of "scaffolding proteins" is not a new one, and the literature is replete with examples of molecules that act to restrict, organize, and/or regulate the subcellular distribution of other signaling molecules (Pawson and Scott, 1997). For example, a paradigm class of proteins with such functions are the AKAPs (A-kinase anchor proteins), known to contain the localization and activation of protein kinase A to different subcellular compartments, thereby limiting the spurious activation of this potent kinase (Colledge and Scott, 1999).

View larger version (39K): [in this window] [in a new window]   Fig. 8.   CSD and caveolin binding motif reciprocal interactions. The sequence of the caveolin-scaffolding domain and the caveolin binding sequence motifs within several caveolae-localized signaling molecules are shown. These include G-protein  subunits (Gi2), eNOS, Src family tyrosine kinases, receptor tyrosine kinases (EGF-R), and PKC isoforms (PKC). In most cases, this caveolin interaction is inhibitory, leading to inactivation of the signaling molecules and modulation of downstream signal transduction.

3. Caveolin-2 and -3 As Signaling Modulators. Primarily due to the overwhelming focus on the archetypal caveolin-1, the capacity of the other members of the caveolin family to act as scaffolding proteins remains unknown. Most of what is known about these proteins is by analogy with Cav-1. A glance at the protein alignment in Fig. 3 illustrates that Cav-1 and Cav-3 have highly homologous CSDs (especially in the spacing of the aromatic residues alluded to above), whereas Cav-2 is largely divergent. Therefore, it can be surmised that in muscle cells, where Cav-3 is selectively expressed, it can act as an effective scaffolding protein in the absence of other caveolins. Indeed, the few experiments that have been conducted with Cav-3 and its CSD seem to corroborate this idea (Li et al., 1995; Yamamoto et al., 1998; Razani et al., 1999). At present however, the dearth of data on the functions of caveolin-2 preclude its implication in any known signaling pathway.

4. Signaling Spotlight: Modulation of Endothelial Nitric-Oxide Synthase Function. Probably the best studied of the signaling molecules that are modulated by caveolin is eNOS. Acylation of eNOS serves to target it to lipid-raft domains of the plasma membrane and the Golgi apparatus (Garcia-Cardena et al., 1996b; Shaul et al., 1996; Sowa et al., 2001; Fulton et al., 2002) (Fig. 9). eNOS localized to lipid rafts/caveolae, but not interacting with Cav-1, has optimal enzymatic function; however, interaction with Cav-1 inhibits eNOS function (Garcia-Cardena et al., 1996a; Liu et al., 1996a; Michel et al., 1997a; Sowa et al., 2001). The region of Cav-1 responsible for the inhibition of eNOS has been mapped to the CSD (Garcia-Cardena et al., 1997; Ju et al., 1997). Point mutations in the CBD of eNOS results in the loss of Cav-1 binding without the loss of eNOS enzymatic activity (Garcia-Cardena et al., 1997). Thus a model has been proposed in which eNOS is targeted to caveolae/lipid rafts where the CSD/CBD interaction serves to maintain eNOS inhibition until the CSD/CBD interaction is disrupted due to an increase in cytosolic calcium followed by calcium-calmodulin/eNOS interaction (Fig. 9) (Garcia-Cardena et al., 1997; Michel et al., 1997a, b).