Physiological Reviews

Role of Caveolae and Caveolins in Health and Disease

Alex W. Cohen, Robert Hnasko, William Schubert, Michael P. Lisanti


Although they were discovered more than 50 years ago, caveolae have remained enigmatic plasmalemmal organelles. With their characteristic “flasklike” shape and virtually ubiquitous tissue distribution, these interesting structures have been implicated in a wide range of cellular functions. Similar to clathrin-coated pits, caveolae function as macromolecular vesicular transporters, while their unique lipid composition classifies them as plasma membrane lipid rafts, structures enriched in a variety of signaling molecules. The caveolin proteins (caveolin-1, -2, and -3) serve as the structural components of caveolae, while also functioning as scaffolding proteins, capable of recruiting numerous signaling molecules to caveolae, as well as regulating their activity. That so many signaling molecules and signaling cascades are regulated by an interaction with the caveolins provides a paradigm by which numerous disease processes may be affected by ablation or mutation of these proteins. Indeed, studies in caveolin-deficient mice have implicated these structures in a host of human diseases, including diabetes, cancer, cardiovascular disease, atherosclerosis, pulmonary fibrosis, and a variety of degenerative muscular dystrophies. In this review, we provide an in depth summary regarding the mechanisms by which caveolae and caveolins participate in human disease processes.


Since their initial discovery in the early 1950s, caveolae, with their unique flask-shaped morphology, have provoked a multitude of conjecture as to their functional significance. Although detailed morphological examinations have provided some insight into their function, it was not until the discovery of the caveolin coat proteins (caveolin-1, -2, -3) that the true nature and importance of these organelles was realized. Since that time, the exponential growth of the caveolae field has provided numerous clues as to the physiological functions of both the caveolae organelle and the caveolin coat proteins. The recent generation of mice deficient in the various caveolin genes has provided new in vivo animal models with which to elucidate the exact physiological significance of a given caveolin gene product.

Indeed, evidence is accumulating that implicates the caveolin gene family in the pathogenesis of numerous human diseases, including cancer, muscular dystrophy, and type II diabetes. This review describes the 1) biochemical properties of the caveolin protein products, 2) the physiological consequences of caveolin gene ablation in mice, and 3) the relationship between caveolin expression and the development/progression of certain human diseases.


A. Discovery and Morphology

Caveolae are morphologically identifiable plasma membrane invaginations that are distinct from the larger electron-dense clathrin-coated pits. Originally identified in the 1950s by electron microscopists investigating cellular ultrastructure, caveolae appear as “smooth” uncoated pits or vesicles at the plasma membrane, typically observed using conventional resin-embedded techniques. In general, caveolae are 50- to 100-nm flask-shaped invaginations of the plasma membrane that can be singular or found in detached grapelike clusters (rosette formation) and long tubular structures thought to evolve from the fusion of individual caveolae (Fig. 1). Isolated caveolae appear as vesicular or curved U-shaped structures possessing a distinctive granular appearance using techniques such as transmission electron microscopy and low-angle platinum shadowing (Fig. 2). While the overall function of the prototypical caveolae organelle has been an area of intense exploration, little attention has focused on the specialized function, if any, of the various morphological subsets of caveolae. Consequently, the functional significance of these caveolae-related organelles remains unknown.

FIG. 1.

Stylized view of the cell depicting morphological variants of caveolae and select subcellular compartments. These include 1) fenstra, 2) a transcellular channel, 3) traditional caveolae, 4) plasmalemmal vesicles (fully invaginated, static caveolae), 5) a vesiculo-vacuolar organelle (a grapelike cluster of interconnected caveolae and vacuoles), 6) cavicles (mobile, internalized caveolae not associated with the plasma membrane), and 7) a caveosome (a slow moving, irregularly shaped, cytoplasmic organelle). Golgi, dark blue; endoplasmic reticulum, yellow.

FIG. 2.

Ultrastructural analysis of purified caveolae. A, top: caveolae purified from mouse lung appear as ∼100-nm vesicles (arrowheads) and as curved or U-shaped structures (arrows) by transmission electron microscopy. Bar, 0.1 μm. Bottom: these domains were fixed with paraformaldehyde and immunolabeled with anti-caveolin-1 IgG and 10-nm gold-conjugated secondary antibodies. Note that these domains contain an abundance of caveolin-1. To preserve immunoreactivity, it was necessary to exclude fixation with glutaraldehyde and OsO4. [Modified from Lisanti et al. (131).] B, top: isolated mouse lung caveolae as seen by low-angle platinum shadowing. These domains are ∼50–100 nm in diameter and possess a distinctive granular appearance. Bar, 0.1 μm. Bottom: staining with anti-caveolin-1 IgG and 10-nm gold-conjugated secondary antibodies reveals caveolin-1 in these domains (magnification is twice that of the top panel). [Modified from Lisanti et al. (130).]

B. Tissue Distribution

Caveolae were simultaneously identified in capillary endothelial cells and epithelial cells from the mouse gall bladder (181, 285). Since then, caveolae have been identified in a wide variety of tissues and cell types. While no all-encompassing ultrastructural study has been undertaken, a review of published literature reveals that caveolae are present to some degree in most differentiated cell types. Particularly, caveolae have been well-described in adipocytes, where they are extremely abundant, endothelial cells, type I pneumocytes of the lung, and striated and smooth muscle cells. Because of their relative abundance in endothelial cells and type I pneumocytes, the two major constituents of lung alveoli, the lung stands out as one of the most abundant sources of identifiable caveolae, second only to adipocytes. Ultrastructural analysis of adipocytes has shown that as much as 20% of the total plasma membrane is occupied by caveolae (53). Thus caveolae can greatly increase the surface area of numerous cell types, an observation that lends credence to the original speculation that caveolae are involved in macromolecular transport and mechanotransduction events.

C. Biochemical Properties

Unlike earlier views of the plasma membrane as a “fluid mosaic” (235), where integral membrane proteins were thought to float and diffuse freely through a sea of homogeneous lipids, a more contemporary view of the plasma membrane is that proteins are much more heterogeneously distributed and can be found clustered within specialized microdomains, termed lipid rafts. These lipid rafts are thought to form via the aggregation of glycosphingolipids and sphingomyelin in the Golgi apparatus (held together by transient and weak molecular interactions) and are then delivered to the plasma membrane as concentrated units (263, 264). These lipid rafts are also enriched in cholesterol and several resident proteins, including glycophosphatidylinositol (GPI)-linked proteins. Relative to the plasma membrane proper, which contains an abundance of cis-unsaturated phospholipids, the sphingolipids in lipid rafts contain primarily saturated fatty acyl chains, allowing tighter molecular packing that results in a higher melting temperature (Tm ∼41°C vs. Tm <0°C for phospholipids) (223). The high cholesterol and sphingolipid content of lipid rafts imparts a resistance to extraction in nonionic detergents such as Triton X-100 at 4°C and a light buoyant-density in sucrose gradients, properties instrumental for their purification and biochemical characterization. It should be mentioned that the exact nature and defining characteristics of lipid rafts, as well as the techniques involved in isolating them, are now quite well-developed (for recent reviews of this subject, see Refs. 90, 165, 187).

Caveolae represent a morphologically identifiable subset of lipid rafts. They contain the coat protein caveolin, which is essential for the invagination of the plasma membrane through a largely unknown process, giving them their characteristic flasklike appearance. While the overall biochemical composition of lipid rafts and caveolae is thought to overlap, these microdomains are not completely equivalent. In addition to the caveolins, several proteins have been shown to preferentially localize to either caveolae or lipid rafts, respectively (134).


A. Initial Discovery

In the late 1980s, investigators attempting to identify possible targets for phosphorylation-mediated cellular transformation used a then recently produced antiphosphotyrosine antibody (PY-20) to affinity purify several phosphotyrosine-containing proteins from Rous sarcoma-transformed chicken embryonic fibroblast (77). One of the four major proteins identified was found to be resistant to extraction with nonionic detergents and demonstrated a staining pattern that was simultaneously punctate, concentrated at cell margins, and focused in parallel arrays along actin stress fibers. The localization of this 22-kDa protein was also shown to change after cellular transformation; indeed, the tyrosine phosphorylation of this protein was dependent on transformation with v-Src, suggesting a role for this protein in oncogenesis (76). This 22-kDa protein was localized to caveolae by immunoelectron microscopy, and ultrastructural analysis revealed that numerous striated filaments coat the cytoplasmic surface of these organelles. In searching for the protein components of these striations, it was found that antibodies generated against the 22-kDa protein stained the striated filaments, and thus the 22-kDa protein was termed caveolin (now called caveolin-1). Furthermore, the caveolin filaments were found to be resistant to extraction with high salt, and treatment of cells with cholesterol-chelating agents resulted in the flattening of membrane caveolae, with disassociation of the striated coat (203). Later, using oligonucleotides predicted from its peptide sequence, the caveolin gene was cloned. Overexpression of the caveolin-1 cDNA in fibroblasts resulted in caveolin-1 localization to caveolae, confirming that it is the major resident protein of this organelle (75).

Working on a completely different aspect of cell biology, Kurzchalia et al. (115) identified a set of CHAPS-insoluble proteins, one of which they termed VIP21 (vesicular integral-membrane protein of 21 kDa). The cDNA for this protein was cloned and transiently expressed in mammalian cells. By immunofluorescence microscopy, VIP21 was found to localize to the Golgi apparatus, the plasma membrane, and membrane-bound vesicles (115). In later work, Glenney (75) showed that the sequences for VIP21 and caveolin were identical and thus these two research groups had independently identified the same protein by distinct biochemical methods.

B. The Caveolin Gene Family

To date, three members of the caveolin (CAV) gene family have been identified (Fig. 3). Caveolin (later termed caveolin-1; Ref. 216) was the first gene discovered and is composed of three exons that are highly conserved in sequence and structure across species. Caveolin-2 was discovered when micro-sequencing of purified adipocyte caveolae membrane domains revealed a strikingly similar peptide sequence to that of caveolin-1, differing in several key conserved caveolin-1 residues. Cross-referencing this new peptide sequence with known expressed sequence tags (EST) revealed an EST encoding a caveolin-1-like protein of 162 amino acids, subsequently termed caveolin-2 (216). Caveolin-3 was cloned from a cDNA library using a caveolin-1-related sequence found immediately downstream of the rat oxytocin receptor (256).

FIG. 3.

Schematic depiction of the caveolin gene family. Color-coded boxes indicate the exon arrangement of each caveolin family member. The numbers within each box refer to the number of nucleotides in each exon.

In addition to these two new members of the caveolin gene family, it was also found that both caveolin-1 and -2 have multiple isoforms. Caveolin-1 has two isoforms, termed α and β, with the α-isoform consisting of residues 1–178 and the β-isoform containing residues 32–178, resulting in a protein ∼3 kDa smaller in size (128, 217). The β-isoform is thought to derive from an alternate translation initiation site occurring at a methionine in position 32. When expressed in Sf21 insect cells, which lack endogenous caveolae, both isoforms of caveolin-1 (Cav-1α and Cav-1β) are capable of driving caveolae formation (128). In addition, analysis of the size distribution of caveolae in this setting revealed a mean of 80 ± 14.8 nm with 95% of caveolae being between 50–105 nm (Fig. 4). Furthermore, whole-mount electron microscopic evaluation of isolated recombinant caveolae revealed that both the α- and β-isoforms of caveolin-1 generate morphologically similar cup-shaped caveolae (Fig. 4). While the exact functional significance of these distinct isoforms remains unclear, studies have suggested that the caveolin-1α isoform is localized predominantly to deeply invaginated caveolae and can more efficiently drive the formation of caveolae than the β-isoform (65, 217). Caveolin-2 has three identified isoforms, the full-length caveolin-2α, and two truncated variants, termed caveolin-2β and -2γ. The β-isoform is thought to be an alternate splice variant, with a distinct subcellular distribution from full-length caveolin-2α (113). Little is known about the functional significance, if any, of the Cav-2β and -2γ isoforms.

FIG. 4.

Caveolin-1 expression generates caveolae organelles. Expression of mammalian caveolin-1 isoforms (α and β) in Sf21 insect cells. A: Western blot analysis of baculovirus-based recombinant expression of caveolin-1α and caveolin-1β in insect cells reveals that the α-isoform is ∼3 kDa larger that the β-isoform. B: transmission electron microscopy of Sf21 insect cells expressing caveolin-1α. Note the accumulation of caveolae-sized vesicles, similar in size and shape to those seen in mammalian cells (∼50–100 nm in diameter). Comparable results were obtained with caveolin-1β expressing cells. Bar, 100 nm. C: quantitation of the size distribution of reconstituted caveolae in Sf21 insect cells. Measurements of over 100 vesicular profiles revealed a mean of 80.3 ± 14.8 (SD) nm in diameter. Ninety-five percent of the caveolae-sized vesicles were 50–105 nm in diameter (2 SD). Identical results were obtained with the expression of caveolin-1α and -1β. D: purified recombinant caveolae isolated from Sf21 insect cells have a similar morphological appearance to those isolated from mammalian cells. Note that these caveolae appear as cup-shaped vesicular (∼50–100 nm) structures (arrows). Identical results were obtained with caveolin-1α and -1β. Bar, 100 nm. [Modified from Li et al. (128).]

The various caveolin genes and proteins share significant homology. Human caveolin-2 is ∼38% identical and ∼58% similar to human caveolin-1, while caveolin-3 is ∼65% identical and ∼85% similar to caveolin-1. In addition, the caveolin proteins are highly conserved throughout evolution (Fig. 5). Moreover, a short stretch of eight amino acids has been identified (FEDVIAEP) that constitutes the “caveolin signature sequence,” a motif that is identical between all three caveolin proteins.

FIG. 5.

Protein sequence alignment of caveolin-1 across several species. Sequences are presented in decreasing order of similarity to human caveolin-1. Identical residues are not shown in the bovine, mouse, chick, and fugu sequences, while changed amino acids are indicated. A − indicates a space inserted to maximize homology among species. Colors indicate proline (purple, P); aspartic or glutamic acid (red, D and E); arginine or lysine (blue, R and K); phenylalanine, isoleucine, methionine, valine, tryptophan, or tyrosine (green, F, I, L, M, V, W, and Y); and all other amino acids alanine, cysteine, glycine, histidine, asparagine, serine, and threonine (black, A, C, G, H, N, Q, S, and T). The single red box outlines the oligomerization domain, the double red box outlines the scaffolding domain, and the blue box outlines the membrane spanning domain. The Swiss-Prot accession numbers are as follows: human (P49817), bovine (P79132), mouse (P49817), chick (P35431), and fugu (Q9YGM8).

The Cav-1 and Cav-2 genes have a relatively ubiquitous distribution pattern, being coexpressed in most differentiated cells types, with the notable exception of skeletal muscle fibers and cardiac myocytes (215). Initial examination of murine tissue by Northern blot showed that Cav-3 expression is limited to skeletal muscle, the diaphragm, and the heart (256, 272). Additionally, the expression of Cav-3 has been shown to be dependent on the degree of muscle cell differentiation, as Cav-3 mRNA is only detectable in differentiated, nonproliferating C2C12 myoblasts versus the proliferating precursor myoblast cell type in which both Cav-1 and Cav-2 are highly expressed (256, 272).

C. Caveolin Protein Characterization

As the prototypical caveolin family member, caveolin-1 has been the primary focus of many biochemical studies and provides the basis of our knowledge regarding the interaction of these proteins with their subcellular environment. Therefore, the attributes of caveolin-1 are considered here; however, it can be inferred that the majority of these characteristic are valid for the other caveolin family members. Caveolin-1 is a 22-kDa protein of 178 amino acids with adipocytes, endothelial cells, fibroblasts, and type I pneumocytes having the highest levels of expression. It is now known that caveolin-1 expression is sufficient and necessary to drive the formation of morphologically identifiable caveolae (128, 197). In addition, caveolin-1 expression is necessary for the stable expression and membrane localization of caveolin-2. Indeed, caveolin-2 alone is insufficient to induce caveolae biogenesis (199). The Cav-3 protein is muscle specific and, like Cav-1, is sufficient to drive the formation of caveolae (67).

1. Membrane topology and attachment

Initial studies into the subcellular localization of caveolin-1 revealed that it is an integral membrane protein, as it was found to be resistant to extraction with sodium carbonate and high salt concentrations (115, 203, 212). In addition, several lines of evidence have contributed to the generally accepted view that both the NH2 and COOH termini of caveolin-1 face the cytoplasm, with an intervening hydrophobic domain inserted into the membrane (Fig. 6). Preliminary findings indicative of this unusual topology were suggested by Dupree et al. (42) after antibodies directed against either the NH2- or COOH-terminal domains of caveolin-1 were unable to label the protein in the absence of cellular permeablization. Additional studies demonstrated that the COOH terminus of caveolin-1 is palmitoylated and the NH2 terminus is tyrosine phosphorylated, two posttranslational modifications that require both ends of the protein to remain cytoplasmic (39, 127). Furthermore, surface biotinylation studies failed to label caveolin-1, further supporting the idea that no portion of the caveolin-1 protein is extracellular (211). The membrane insertion of caveolin-1 occurs via the classical endoplasmic reticulum (ER) machinery, resulting in an unusually short transmembrane domain of 32 hydrophobic amino acids (residues 102–134) thought to form a unique hair-pin loop configuration that prevents caveolin-1 from completely spanning the plasma membrane in a traditional double-pass fashion (Fig. 6) (159).

FIG. 6.

Caveolin-1 membrane topology and protein domains. In this view, caveolin-1 is depicted as a homodimer for simplicity. Mutational analysis has shown that the C-MAD (blue; residues 135–150) and N-MAD (yellow; residues 82–101) are important for membrane attachment, while the transmembrane domain (red; residues 102–134) is thought to insert into the membrane. Oligomerization is mediated by residues 61–101 (hashed pink). The scaffolding domain (yellow; residues 82–101) recognizes a hydrophobic caveolin-binding motif present within many signaling molecules. Caveolin-1 is also palmitoylated on three conserved cysteine residues (green; 133, 143, and 156). Note that caveolin-1 is not a conventional transmembrane protein. It is thought to have a unique hairpin topology, with no exposure to the extracellular environment.

Mutational analysis and domain mapping experiments have demonstrated the importance of two other regions of the caveolin-1 protein for membrane attachment. Interestingly, the postulated membrane-spanning region (residues 102–134) was not found to be essential for this function. Rather, two adjacent regions that flank this central hydrophobic domain, residues 82–101 and residues 135–150, were found to bind to membranes with high affinity (2, 145, 220, 221). These regions are now referred to as the NH2-terminal membrane attachment domain (N-MAD) and COOH-terminal membrane attachment domain (C-MAD), respectively. The C-MAD contains a Golgi-targeting sequence dissimilar from other known Golgi-associated sequences and was found to localize an otherwise cytoplasmic green fluorescent protein (GFP) to the cis-Golgi when expressed as a fusion protein (145, 220). When GFP was fused to the C-MAD of caveolin-1, the resultant protein was resistance to extraction in both Triton X-100 at 4°C and alkaline carbonate buffer, two properties consistent with an integral membrane protein (220). Several critical experiments reveal the importance of caveolin-1 residues 82–101 (N-MAD) in membrane attachment (221). Deletion mutagenesis showed that a caveolin-1 mutant containing only residues 1–101 remained associated with membranes, while another mutant composed of residues 1–81 localized to the cytoplasmic compartment (221). Like the C-MAD, the N-MAD-GFP fusion protein was also targeted to membranes. However, while the C-MAD-GFP fusion protein showed a predominant Golgi-like distribution, the N-MAD-GFP fusion protein localized to plasma membrane caveolae (145, 220). Further mutational analysis of the 20-amino acid N-MAD domain identified a short membrane attachment sequence (KYWFYR) that was sufficient to confer membrane localization to a GFP fusion protein (280). Although the KYWFYR motif targets GFP to the membrane, the entire 20-amino acid N-MAD is necessary for caveolar targeting.

2. Oligomerization domain

Caveolin-1 isoforms form high-molecular-mass oligomers of ∼400 kDa, as demonstrated using velocity gradient centrifugation (159, 211). Interestingly, deletion mutagenesis studies showed that caveolin-1 contains an oligomerization domain mapping to residues 61–101, which mediates the homo-oligomerization of 14–16 individual caveolin-1 molecules (Fig. 6) (211). These caveolin-1 oligomers are thought to undergo a second stage of oligomerization during transport from the trans-Golgi to plasma membrane caveolae, whereby several oligomers self-associate via COOH-terminal interactions, forming a large network of caveolin (243). Sargiacomo et al. (211) examined the ultrastructure of purified caveolin-1 homo-oligomers derived from rat lung tissue and found that these individual homo-oligomers appear as globular particles of 4–6 nm in the presence of the nonionic detergent octyl glucoside (Fig. 7). After removal of this detergent, caveolin-1 homo-oligomers associated into larger, nonlinear polymers of ∼25 nm in an apparent side-by-side fashion, further supporting the notion that this may be the mechanism by which individual homo-oligomers associate into larger multimeric caveolar complexes in vivo.

FIG. 7.

Low-angle platinum shadowing of purified caveolin-1 homo-oligomers. Caveolin-1 was purified by velocity gradient centrifugation from isolated murine lung caveolae. A: caveolin-1 homo-oligomers appear as 4- to 6-nm individual globular particles in the presence of the nonionic detergent octyl glucoside. B: after removal of octyl glucoside, individual caveolin homo-oligomers self-associate into ∼25-nm nonlinear structures, reminiscent of intact caveolae. The boxed area is shown at higher magnification to illustrate the side-by-side packing of caveolin oligomers. Bar, 0.05 μm. [Adapted from Sargiacomo et al. (211).]

In contrast, caveolin-2 is incapable of forming high-molecular-mass homo-oligomeric complexes (216), a property that may contribute to its inability to form caveolae by itself. However, caveolin-2 does form hetero-oligomeric complexes with caveolin-1 in the ER, an association that serves to both stabilize caveolin-2 against proteasomal degradation and allow its transport from the Golgi to the plasma membrane (160, 185, 198, 214).

Caveolin-3 is known to form high-molecular-mass homo-oligomeric complexes in situ, in a fashion similar to that of caveolin-1 (239, 256). In addition, while it has long been thought that caveolin-3 does not bind to caveolin-2, recent findings in isolated neonatal cardiac myocytes and smooth muscle cells suggest otherwise (206, 278). In these systems, it was found that caveolin-2 coimmunoprecipitates with caveolin-3 and cofractionates with caveolin-3 in sucrose density gradient centrifugation (206, 278).

3. The caveolin scaffolding domain

The caveolin-1 scaffolding domain (CSD) was originally defined by a series of experiments, including deletion mutagenensis of Cav-1-GST fusion proteins, which identified a region within caveolin-1 (residues 82–101) capable of mediating protein-protein interactions (3335, 104, 124, 243). Using a GST-fusion protein containing the caveolin-1 scaffolding domain as a receptor to select peptide ligands from a bacteriophage display library, two related but distinct caveolin binding motifs (CBM) were identified in most proteins shown to interact with caveolin-1 (CBM; ΦXXXXΦXXΦ and ΦXΦXXXXΦ, where Φ represents an aromatic amino acid) (33). The scaffolding domain has since been shown to serve a dual role, acting both as an anchor holding various proteins within caveolae as well as a regulatory element capable of either inhibiting or enhancing a given protein's signaling activity.


A. Vesicular Transport

Based purely on their ultrastructural appearance as plasma membrane invaginations, caveolae were originally thought to function as macromolecular transport vesicles. Initially, their proposed function was limited to the process of pinocytosis or “cellular drinking.” However, with the advent of new tools to investigate their function, their role as vesicular transporters has expanded to include transcytosis and endocytosis.

1. Transcytosis

The term transcytosis was first used by Simionescu et al. (233) to describe the movement of macromolecules from the luminal side of capillary endothelial cells to the interstitial space via membrane-bound vesicles. This early research involved ultrastructural quantification of probes, such as gold-labeled albumin and peroxidase-labeled small heme derivatives, in what appeared to be transendothelial channels formed by the coalescence of caveolae (74, 233). While these reports demonstrate an accumulation of labeled probes in both abluminal caveolae and the interstitial space, indicative of transcytosis, the possibility remained that these tracers were subject to paracellular transport (between cells). Furthermore, their presence in abluminal caveolae could also have been attributed to retrograde filling, i.e., paracellular transport followed by entry into caveolae.

It was not until two major advancements in the field, 1) the introduction of novel caveolae-specific tracer molecules and 2) the development of caveolin knockout mice, that the role of caveolae in transcytosis could be resolved. Recently, Schnitzer and co-workers (151) used a novel approach that employed antibodies generated against antigens present only on the luminal surface of caveolae. Using these antibodies, this group demonstrated that caveolae are capable of macromolecular transcytosis, as they found an ∼50-fold increase in the interstitial accumulation of caveolae-specific antibodies, compared with control antibodies (151). Because their specific probe accumulates in tissue, while a control probe of similar structure does not, these authors conclude that nonspecific pathways, such as paracellular transport, cannot account for the interstitial accumulation.

Measurements of transcytosis in caveolin-1 knockout mice, which lack detectable caveolae in cell types such as endothelial and epithelial cells, show no interstitial accumulation of gold-labeled albumin in knockout, but significant accumulation in wild-type mice (224, 225). Similarly, when incubated with radio-iodinated albumin, isolated aortic rings from caveolin knockout mice showed no uptake of this tracer, while aortic rings from wild-type animals showed temperature- and time-dependent albumin uptake (225). In contrast to these findings however, Drab et al. (40) found no change in the apparent transcytosis of albumin into the cerebrospinal fluid (CSF) of their independently generated Cav-1 null mice. Here, these authors assessed albumin transcytosis across endothelial cells by measuring the concentration of albumin in the CSF of wild-type and Cav-1 null mice using SDS-PAGE. These authors conclude that either 1) caveolae are not involved in endothelial cell transcytosis or 2) an unidentified compensatory mechanism exists. To further clarify this issue, Schubert et al. (225) examined the microvascular permeability of wild-type and Cav-1 null mice, finding that indeed Cav-1 null mice demonstrate significant increases in the extravascular deposition of radioiodinated albumin. Additional experiments led us to conclude that defective transcytosis of albumin is compensated for by increased paracellular transport between endothelial cells, a process mediated by increases in plasma nitric oxide levels. Thus with these studies it now seems clear that caveolae do indeed mediate transcytosis (i.e., transcellular transport) of specific macromolecules in endothelial cells.

2. Endocytosis

While clathrin-mediated endocytosis stands as the prototypical pathway for the internalization of many extracellular substances, alternative mechanisms also exist, including those mediated by caveolae. Caveolar transport may overlap with those events mediated by coated pits, but caveolae may serve selective transport functions as well.

This alternative caveolar pathway was first recognized by the observation that cholera toxin was preferentially bound and internalized via caveolae (177). Additional observations suggesting a role for caveolae in the process of endocytosis include 1) the identification of several molecules involved in vesicle docking and fusion (i.e., SNARE proteins) shared by clathrin-coated pits (CCPs) and 2) the detection of the small GTPase dynamin, which is known to be involved in the internalization of CCPs, in the necks of caveolae following specific stimuli (91, 173, 222). Taken together, these data have helped solidify the idea that caveolae are indeed functional endocytic vesicles. Dynamin seems to play a homologous role in the internalization of both CCPs and caveolae, as overexpression of a dominant negative dynamin mutant inhibits endocytosis by both pathways (173). Although the exact process mediating caveolar endocytosis remains unknown, several recent findings suggest that the activation of tyrosine kinase-dependent signaling is an important step (190). Indeed, treating cells with phosphatase inhibitors increases endocytosis via caveolae. Similarly, ligands that are known to be internalized via caveolae, including several pathogenic agents, activate tyrosine kinase signaling pathways upon binding to their cellular receptors (21, 170, 171, 188, 192).

Perhaps the most well-documented infectious agent that selectively uses caveolae to enter cells is simian virus 40 (SV40). It has been shown that SV40 is internalized specifically by caveolae by several means, including that the overexpression of a dominant negative epidermal growth factor (EGF) receptor pathway mutant (Eps15) will selectively inhibit clathrin-mediated endocytosis without effects on SV40 internalization (192). Additionally, the overexpression of a dominant negative caveolin mutant, which inhibits caveolar-mediated endocytosis, resulted in little or no SV40 infection (191, 204). Following its initial binding, SV40 accumulates in what has now been termed a “caveosome,” an endocytic vesicle containing caveolin-1 and devoid of markers of clathrin-mediated endocytosis (191). After accumulation, SV40 is delivered to the ER, completing a mechanism that bypasses the lysosomal compartment thereby preventing SV40 inactivation (171, 191). Although SV40 is the only virus shown unequivocally to enter cells via caveolae en route to the ER, this may prove prototypical for many viruses (170). It is unclear if SV40 is an infectious agent to humans or a mediator of disease, but its exploitation of caveolae to gain entry into cells and avoid degradation may prove a robust strategy co-opted by many pathogens. It should also be noted that viral endocytosis via caveolae may be a relatively slow process and that caveolae, on the whole, are thought to be relatively stable structures (190, 260); however, this remains a hotly debated topic (164).

A growing list of pathogens, including viruses, bacteria and their associated toxins, fungi, and even prions, can interact with caveolae membrane domains (Table 1). The intracellular trafficking of these agents via caveolae differs dramatically from the usual route of ligands internalized by clathrin-medited endocytosis. The use of caveolae for cellular entry allows the pathogen to avoid classical endosome-lysosome trafficking and, consequently, avoid such degradative compartments in the cell. The mechanism of pathogen binding usually involves the utilization of one or more of the glycolipid or GPI-anchored moieties that cluster within the caveolae of host cells. It appears that disparate pathogens can utilize common binding sites localized within caveolae (170, 231, 232). The numbers of currently known pathogens that can use caveolae for cellular invasion probably represents a small fraction that will ultimately be shown to use caveolae-mediated cell entry (Table 1). Interestingly, the Marburg and Ebola viruses, which result in lethal hemorrhagic fever, have been shown to utilize the folate receptor α as a cofactor to gain entry into cells (20). Consequently, these pathogens may exploit caveolar potocytosis as a mechanism for cellular invasion. In the same auspice, strategies aimed to exploit the folate receptor for therapeutic gene transfer may prove useful for the treatment of human disease (37, 236). Importantly, as pathogens exploit caveolae as a common mechanism to gain entry into cells and avoid degradative membrane compartments, this may benefit pharmaceutical strategies aimed at reengineering common pathogen moieties to facilitate entry, translocation, and delivery of therapeutic agents.

View this table:

Pathogens linked to caveolar internalization and/or signaling

B. Cholesterol Homeostasis

The interrelationship between cholesterol and caveolin has a wide and varied history. Rothberg et al. (203) first recognized the importance of cellular cholesterol balance on caveolar structure when they found that treating cells with cholesterol binding agents, such as filipin and nystatin, resulted in the flattening of morphologically identifiable caveolae and the disruption of the striated caveolin protein coat. Additionally, cholesterol regulates caveolin-1 expression at the level of transcription through two steroid regulatory binding elements in the caveolin-1 promoter. This results in decreased caveolin-1 mRNA levels during periods of cholesterol depletion, while cholesterol loading has the opposite effect (8, 58). Additionally, the oxidation of cholesterol into cholesterone by cholesterol oxidase functionally disrupts cellular cholesterol balance and results in the internalization of caveolin-1 from the plasma membrane with its translocation to the ER and the Golgi apparatus (237). Thus caveolin-1 expression and its intracellular distribution are clearly dependent upon cellular cholesterol levels.

In an attempt to determine the interrelationship between caveolin-1 and the cellular cholesterol transport machinery, Frank et al. (63) identified a direct role for cholesterol in the stabilization of caveolin-1. These authors found that the expression of endogenous caveolin-1 increased ∼18-fold when human embryonic kidney (HEK) 293 cells were transfected with the scavenger receptor B1 (SR-B1) cDNA, but not with a cDNA encoding CD36 (63). Because SR-B1, but not CD36, results in the accumulation of intracellular cholesterol, these authors postulated that cholesterol, not SR-B1 itself, was responsible for the increase in caveolin-1 expression. Indeed, when HEK 293 cells were loaded with cholesterol, caveolin-1 protein expression increased dramatically by Western blot analysis. In addition, the distribution of caveolin-1 expression was also affected, with the intracellular clustering of caveolin-1 in cholesterol-loaded COS-7 cells. Thus, in addition to positive transcriptional regulation, cholesterol also directly modulates caveolin-1 expression via stabilization of the protein itself.

This interrelationship between caveolin-1 and cholesterol is further complicated by recent evidence suggesting that caveolin-1 can regulate cholesterol levels by modulating its cellular influx and efflux. Caveolin-1 has been shown to transport newly synthesized cholesterol from the ER to membrane caveolae, where it is delivered to plasma high-density lipoproteins (HDL). Furthermore, while extracellular cholesterol primarily enters cells via clathrin-mediated endocytosis of low-density lipoproteins (LDL), a secondary pathway involving caveolae may mediate the influx of cholesterol from HDL particles by way of the scavenger receptor B1 (SR-B1), which has recently been shown to colocalize with caveolin-1 in caveolae (81). Therefore, caveolae may be the principal location whereby plasma membrane cholesterol is exchanged between HDL and the cell membrane.

C. Signal Transduction

Perhaps one of the most important realizations concerning caveolae and caveolins was that these elements play an important functional role in the modulation of cell signal transduction. For a more in-depth review of caveolae and caveolins in signal transduction, the reader is referred to References 117, 200, 219, 234. Lisanti and colleagues (131, 212) were the first to recognize this property when they made the surprising discovery that biochemically purified caveolae microdomains contained an abundance of signaling molecules, such as Src-like tyrosine kinases and heterotrimeric G proteins (131, 212). Such initial discoveries led these authors to propose that caveolin-1 may serve to compartmentalize certain signaling molecules within caveolae, thereby concentrating and localizing these elements within the cell with the prospect of rapidly and selectively modulating cell signaling events (130). Thus caveolae may serve as docking points for numerous cell surface receptors, which when activated by ligand binding are recruited to caveolae (Fig. 8). Here, activated receptors interact with their respective caveolae-associated signaling components, leading to their rapid and selective activation in the confines of a specific subcellular environment. This hypothesis is strengthened by studies showing that the GTPase activity of G protein α-subunits could be suppressed by a peptide derived from the NH2 terminus of caveolin-1, the caveolin scaffolding domain, demonstrating the interdependence of these proteins for functional activity (126).

FIG. 8.

Signaling through caveolae. In this view, two separate receptors are shown docking with a cholesterol and caveolin-enriched caveola organelle, following ligand-mediated stimulation. The β-adrenergic receptor (β-AR; blue) is a conventional G protein-coupled receptor with seven membrane-spanning domains. When stimulated, this receptor initiates a signaling cascade conveyed through several caveolae-localized proteins, beginning with the activation of Gs subunits. This, in turn, leads to the activation of adenylyl cyclase, which increases intracellular cAMP concentrations, resulting in the activation of protein kinase A (PKA). On the right, an activated epidermal growth factor receptor (EGF-R) is also shown docking with the caveola, leading to the activation of a proliferative pathway involving several caveolae-associated proteins of the p42/44 mitogen-activated protein kinase cascade (Ras/Raf/MEK/ERK).

Caveolins are now thought to act as scaffolding proteins, concentrating specific signaling molecules within caveolae via an interaction between the CSD (caveolin-1 residues 82–101) and an aromatic amino acid-based caveolin-binding domain (CBD; described above) usually found in the active catalytic domain of a given caveolae-associated protein. Following these early findings, a rapid succession of studies showed that an enormity of signaling molecules copurify with caveolae by detergent-resistant extraction and sucrose gradient centrifugation (reviewed in Ref. 200). Furthermore, many of these proteins have now been shown to be regulated by an interaction with caveolin-1, including H-Ras, Src-family kinases, and endothelial nitric oxide synthase (eNOS). Caveolin-1, for instance, binds to wild-type H-Ras with high affinity in vitro but does not bind to mutationally activated H-Ras (G12V) (241). Furthermore, a nonfarnesylated mutant of H-Ras (C186S), which normally remains cytoplasmic, is recruited to caveolae membranes by overexpression of the caveolin-1 cDNA (241). Thus this prototypical set of experimental data demonstrates the dual role of caveolin-1 as both a regulator of signal transduction as well as a scaffolding protein, sequestering agents within caveolae. Such experimental findings led to the proposal of the “caveolae signaling hypothesis,” which attributes the regulation of numerous signaling molecules to interactions with caveolae organelles and the caveolin proteins (130, 176). It is important to note that while caveolin seems to be a negative regulator of the vast majority of signaling proteins with which it interacts, at least one protein, the insulin receptor, is positively regulated by an interaction with caveolin-1 (27, 30, 172, 286).

D. Proteomics

Proteomic analysis of purified caveolae has identified a wide assortment of proteins that are localized to these structures. The first detailed proteomic analysis of caveolae was carried out in 1994. Based on their buoyancy and resistance to detergent solubilization, Lisanti et al. (131) used sucrose density ultracentrifugation to purify caveolae-rich membrane domains from murine lung tissue. This procedure allowed for the exclusion of ≥98% of an integral plasma membrane protein marker, while retaining ∼85% of total caveolin and ∼55% of GPI-linked marker proteins. When these purified caveolae were analyzed by protein microsequencing and Western blotting, these authors provided the first glimpse into resident caveolar protein components, identifying a preponderance of cytoplasmically oriented signaling molecules (Src-like kinases and heterotrimeric G protein subunits), including the small GTPases Rap 1, Rap 2, and TC 21, and cytoskeletal elements, such as monomeric G-actin and myosin regulatory light chain (Fig. 9A). In addition, several transmembrane proteins were identified, including CD36 and RAGE, as well as an abundance of GPI-linked proteins (131). This initial proteomic analysis of caveolae allowed for the first large-scale characterization of caveolae-enriched protein constituents and provided the basis for many follow-up investigations into the functional significance that caveolar localization may confer upon these proteins. Importantly, extremely similar results were obtained during the proteomic analysis of purified adipocyte caveolae (Table 2).

FIG. 9.

Proteomic analysis of purified caveolae and caveolin-associated proteins. A: caveolae proteomics. Proteins copurifying with caveolae organelles were further partitioned into hydrophobic and hydrophilic fractions with the detergent Triton X-114, resolved by SDS-PAGE, and stained with Coomassie blue. Note that few proteins segregate into the hydrophilic aqueous portion (Aq), while the vast majority of proteins partition into the hydrophobic detergent fraction (Det). Several prominent bands were subjected to microsequencing, and their assigned identities are indicated. [Modified from Lisanti et al. (131).] B: proteomics of caveolin-associated proteins. Proteins copurifying with caveolin-1, after a cycle of disassembly and reassembly, were resolved by SDS-PAGE and visualized by Ponceau S. These proteins were subjected to microsequencing analysis to determine their identity, as indicated. [Modified from Scherer et al. (217).]

View this table:

Internal microsequencing of purified murine adipocyte caveolae

To identify proteins that are tightly associated with caveolin-1, purified lung caveolae were first dissociated in octyl-glucoside and then allowed to reassociate after removal of the detergent, and subjected to a second round of purification (217). These reassembled caveolae were then analyzed by protein microsequencing (Fig. 9B). With the use of this caveolar assembly/disassembly approach, it was observed that caveolin-1 forms a tight complex with a number of proteins, including a transmembrane protein (CD36), four GPI-linked proteins (membrane dipeptidase, 5′-nucleotidase, ceruloplasmin, carbonic anhydrase IV), actin, a calcium-binding protein (calsequestrin), cytoplasmically oriented signaling molecules (Lyn tyrosine kinase; Gi-2α), and two novel 45-kDa proteins (now known as flotillin-1/2) (217).

More recently, Sprenger et al. (247) explored the proteomic composition of endothelial cell caveolae isolated by two distinct methodologies. In their first approach, these authors coated the outer plasma membrane of cultured human endothelial cells with positively charged silica, allowing subfractionation of luminal membrane components. Homogenization of the silica-bound membranes in cold Triton X-100 allows for the release of caveolar vesicles, which can be subsequently purified by sucrose density gradient centrifugation. However, when the resultant fractions were analyzed by two-dimensional gel electrophoresis and matrix-assisted laser desorption/ionization (MALDI), it was found that this methodology fails to purify caveolae, as almost no known caveolar resident proteins were identified. Furthermore, it seems that this methodology provides for an enrichment of ER proteins, as these were the main resident proteins identified (247).

When these authors next used a procedure to isolate caveolae based on their resistance to extraction in cold Triton X-100 and their propensity to float in a sucrose density gradient, they were able to confirm the validity of this methodology, showing an abundance of known caveolar resident proteins, such as caveolin-1 and flotillin-1. Subsequent two-dimensional gel and MALDI analysis of these isolated microdomains provided insight into their protein components, identifying a host of proteins similar to those described above. In addition, these authors recognized, for the first time, the existence and caveolar localization of several proteins whose identity had only been postulated based on analysis of the human genome (247). These proteins include a stomatin-like protein (SLP-2), a glucose-regulated-like protein (SGRP58), as well as the X-linked retinitis pigmentosa 2 protein (XRP-2). Because at least one of these genes has been identified as causing human disease (XRP-2), its localization to caveolae provides a new avenue by which its molecular function may be addressed. Taken together, proteomic studies such as those described above provide powerful tools by which new caveolar protein components can be identified and may present a basis for the further implication of these structures in human disease.


The recent generation of caveolin (−/−) null mice has made it possible to evaluate the significance of each caveolin protein in the context of whole animal physiology, while yielding valuable new models for the study of human disease. Perhaps one of the more surprising findings regarding these mice was that all of the caveolin-deficient mouse models generated (Cav-1 null, Cav-2 null, Cav-3 null, and Cav-1/3 double knockout mice) are viable and fertile. This is remarkable given the diverse attributes of the caveolin proteins and their widespread tissue distributions. Furthermore, ultrastructural analysis of Cav-1/3 double knockout mice revealed a complete loss of caveolae in both nonmuscle and striated muscle tissues, respectively, thus confirming the essential role of these proteins in caveolar biogenesis. However, these results contrast with those observed in Cav-2 null mice, which retain morphologically identifiable caveolae. While these mice have, thus far, provided unique animal models by which to study the physiological roles of the caveolin proteins, their potential as empirical models of human disease is only now being fully realized. In the sections below, we outline our current understanding regarding the phenotypic characteristic of each caveolin null mouse (Cav-1, Cav-2, Cav-3 single-knockouts, and the Cav-1/3 double-knockout) and how these changes relate to physiological processes and the pathogenesis of human disease.

A. Caveolin-1

Cav-1 null mice were generated and initially characterized by two independent groups simultaneously (40, 198). As reported by both groups, these mice have a complete absence of morphologically identifiable caveolae in all tissues and cell types that normally express caveolin-1 (i.e., endothelia and adipocytes), while retaining identifiable caveolae in tissues that normally express caveolin-3 (i.e., skeletal and cardiac myocytes). Indeed, the Cav-1 null mouse unequivocally demonstrates the essential role of caveolin-1 in caveolar biogenesis in nonmuscle cells. An interesting caveat of caveolin-1 ablation is a consequential loss of the caveolin-2 protein by ∼90%, in all caveolin-1-expressing tissues. This finding was not completely unexpected considering the known role of caveolin-1 in its hetero-oligomerization with caveolin-2, and the ability of caveolin-1 to recruit caveolin-2 from the Golgi to the plasma membrane (125, 160, 185). This reduction in caveolin-2 protein levels is the result of caveolin-2 destabilization and subsequent proteasomal degradation, rather than altered transcriptional regulation. Indeed, the treatment of mouse embryonic fibroblasts (MEFs) derived from Cav-1 null animals with the proteasomal inhibitor MG-132 is sufficient to rescue the caveolin-2 protein expression (197). These data were supported by the observation that, in untreated Cav-1 null MEFs, caveolin-2 remained trapped in the Golgi apparatus, apparently unable to transit to plasmalemmal caveolae in the absence of its molecular chaperone, caveolin-1. In MEFs treated with MG-132, caveolin-2 expression increased to levels near those in wild-type cells; however, the protein remained confined to the Golgi (197). Thus Cav-1 null mice are essentially caveolin-2 deficient as well, and this must be considered in the phenotypic characterization of Cav-1-deficient mice. In support of this notion, our group has shown that the restrictive lung phenotype initially described in Cav-1 null animals is actually due to a selective loss of caveolin-2, as determined through the generation and characterization of Cav-2 null mice (199).

The sections below detail the various phenotypes identified to date in Cav-1 null mice and their relation to human disease.

1. Cellular transformation and tumorigenesis

Caveolin-1 was initially identified as a major substrate for tyrosine phosphorylation in Rous sarcoma virus-transformed chicken embryonic fibroblasts, suggesting that it may be a target for inactivation during oncogenesis (76). Thus it has been proposed that caveolin-1 acts as a tumor suppressor protein, inhibiting the functional signaling activity of several protooncogenes and consequently disrupting the process of cellular transformation (4851, 66, 70, 114, 123, 209).

Numerous follow-up studies designed to test this hypothesis have contributed a myriad of evidence suggesting that caveolin-1 may indeed possess tumor suppressor capabilities. For instance, caveolin-1 mRNA and protein expression are downregulated in NIH-3T3 cells transformed with several activated oncogenes, such as v-Abl, Bcr-Abl, and H-Ras (G12V) (48, 114). The ability of these transformed cells to grow in soft agar, a hallmark of cellular transformation, was found to inversely correlate with caveolin-1 protein levels. The reintroduction of caveolin-1 under the control of an inducible promoter was sufficient to inhibit the anchorage-independent growth of these cells, thus reverting their transformed phenotype (48). In addition, some studies have identified similar reductions in caveolin-1 protein and mRNA levels in breast cancer cell lines (47, 51, 60, 94, 123, 292).

In further support of a role of caveolin-1 as a tumor suppressor, it has been shown that targeted downregulation of caveolin-1 using a vector-based anti-sense approach resulted in the transformation of NIH-3T3 cells, enhanced their anchorage independent growth, and hyperactivated the Ras-p42/44 mitogen-activated protein (MAP) kinase cascade (70). When injected into nude mice, these NIH-3T3 cells expressing the caveolin-1 anti-sense construct were capable of forming large tumors, compared with matched NIH-3T3 cells lacking the caveolin-1 anti-sense vector.

Importantly, these results were reversible, as loss of the caveolin-1 anti-sense vector, and thus reexpression of caveolin-1, reverted the transformed phenotype. Although it is generally accepted that caveolin-1 is indeed targeted for downregulation during oncogenesis, the mechanisms required for this reduction in mRNA levels have, for the most part, remained enigmatic (52). The identification of c-Myc-repressive and p53-responsive elements in the caveolin-1 promoter may provide common targets by which various oncogenes regulate caveolin-1 gene expression (183, 196).

Genetic evidence supporting the role of caveolin-1 as a tumor suppressor has emerged from gene mapping studies, which revealed that the human CAV-1 gene maps to the long arm of human chromosome 7 (7q31.1). This region, the D7S522 locus, encompasses a known fragile site (FRA7G) and is often associated with loss of heterozygosity (LOH) in various cancers, including breast, prostate, ovarian, and renal carcinomas (105, 110, 289291). The deletion of this region and its association with the pathogenesis of several different types of cancers lends credible support to the presence of a tumor suppressor gene within this genetic locus. While no genes have been directly mapped to the D7S522 locus, the closest genes to this region encode caveolin-2 and caveolin-1 (Fig. 10) (50, 51). The CAV-2 gene is found ∼67 kb downstream of the microsatellite repeat marker D7S522, while the CAV-1 gene is located ∼86 kb downstream. Furthermore, the CAV-1 promoter has been reported to be hypermethylated in several cancer cell lines, including those derived from breast and prostate, suggesting that transcriptional silencing may abrogate caveolin-1 expression in human cancers (36, 51). As others have not shown significant changes in the methylation status of the caveolin-1 gene in human tumor cell lines, a result which may reflect technical differences, the transcriptional downregulation of caveolin-1 by promoter silencing will require further clarification (99).

FIG. 10.

Genomic organization of human chromosome 7 (7q31.1) relative to the D7S522 locus, CAV-1, and CAV-2. Colored boxes represent exons and their corresponding number of nucleotides. Open boxes represent introns and their corresponding size in kilobases (kb).

If caveolin-1 truly possesses tumor suppressor characteristics, then by what means does it inhibit cell growth and transformation? Because caveolin-1 is a known scaffolding protein, interacting with and regulating a variety of signaling pathways, it is possible that this protein may similarly regulate protooncogenes. Indeed, overexpression of caveolin-1 in cultured cells is sufficient to inhibit signaling from several proliferative pathways. For instance, it has been shown that key components of the Ras-p42/44 MAP kinase cascade (MEK and ERK) reside within caveolae and that these and other members of this signaling cascade are negatively regulated by a direct interaction with caveolin-1 (46, 70, 137, 139, 238, 241). It has been shown that transient transfection of caveolin-1 dramatically inhibits signaling along the Raf-1/MEK/ERK pathway, and the kinase activity of MEK-1 and ERK-2 are inhibited by incubation with caveolin-1 scaffolding domain-based peptides in vitro (46). Furthermore, a reciprocal relationship exists between caveolin-1 expression and the activation of p42/44 MAP kinase, suggesting that caveolin-1 normally functions to inhibit signaling through this cascade (48, 52, 70, 114).

With so much data indicating that caveolin-1 possesses important tumor suppressor properties, it was somewhat surprising that Cav-1 null mice do not form spontaneous tumors. However, the characterization of the proliferative capacity of cells derived from these mice did reveal their hyperproliferation. For instance, cultured primary mouse embryonic fibroblasts (MEFs) derived from Cav-1 null mice show a marked increase in growth rate, compared with matched control wild-type MEFs (197). In addition, the lungs of Cav-1 null animals were found to be hypercellular, indicative of altered cell cycle regulation. Although Cav-1 null mice do not show any increased incidence of spontaneous mammary tumor formation, they develop significant mammary epithelial cell hyperplasia as early as 6 wk of age, in the nulliparous state. In 9-mo-old Cav-1 null mice, this mammary epithelial cell hyperplasia is also accompanied by an increase in lobular development, with acini formation and fibrosis (121).

Because these results indicate that loss of caveolin-1 may be a contributing factor in oncogenesis, Capozza et al. (18) subjected the skin of Cav-1 null mice to the carcinogen 7,12-dimethylbenzanthracene (DMBA). Here, it was shown that repeated applications of this compound resulted in a 10-fold increase in tumor incidence, a 15-fold increase in tumor multiplicity, and a 35-fold increase in tumor area per mouse, compared with wild-type mice treated identically (18). In addition, before the formation of overt skin tumors, Cav-1 null mice showed severe epidermal hyperplasia, accompanied by increases in cyclin D1 expression and the hyperactivation of the p42/44 MAP kinase cascade. These findings suggest that loss of caveolin-1 sensitizes mouse keratinocytes to oncogenic transformation and that caveolin-1 does indeed function as a tumor suppressor gene.

Following a similar rationale, Cav-1 null mice were interbred with a tumor-prone transgenic model of breast cancer (MMTV-PyMT). MMTV-PyMT mice develop spontaneous mammary tumors by ∼8 wk of age involving the entire mammary (146). When crossed with MMTV-PyMT mice, Cav-1 null mice develop multifocal dysplastic mammary lesions at a much earlier age (3 wk) than MMTV-PyMT mice alone (274). At this early age, there is an approximately twofold increase in the number of lesions per mouse, as well as an approximately sixfold increase in the area occupied by these lesions in PyMT/Cav-1 (−/−) mice compared with PyMT/Cav-1 (+/+) mice. Furthermore, PyMT/Cav-1 (−/−) lesions were of a higher histological grade and contained many more papillary projections than those of PyMT/Cav-1 (+/+) mice. Western blot analysis of these dysplatic lesions revealed a dramatic elevation in cyclin D1 protein expression, thus providing a potential mechanism by which loss of caveolin-1 accelerates the appearance of these lesions.

Thus the loss of caveolin-1 alone appears insufficient to induce cell transformation in vivo, but loss of caveolin-1 potentiates this process when combined with a transforming agent (a carcinogen or tumor-prone genetic background) (18, 274). This is not surprising, as the process of cell transformation and the development of cancer in vivo is a multistep process that involves the selective and progressive loss of several tumor suppressors (such as p53, INK4a, and Rb), as well as the mutational activation or upregulation of certain key protooncogenes [Ras(G12V), c-Myc, and c-Neu/ErbB2]. Indeed, the etiology of most cancers does not reflect alterations in a single gene, but rather the functional loss or induction of a series of key regulatory proteins that, in combination, disrupts the normal regulation of the cell cycle and subsequently leads to uncontrolled cell growth (140).


Fueled by evidence suggesting a role for caveolin-1 in tumorigenesis, many clinically oriented studies have been undertaken to examine the regulation of caveolin-1 in a variety of human cancers. Analysis of human breast cancer samples revealed that up to 16% of these cancers have a CAV-1 gene point mutation (P132L), with the majority of the mutations being found in invasive carcinomas (88). Following up on this finding, Lee et al. (121) examined the behavior of the Cav-1 (P132L) mutant in a mammary cell culture model and found that it is mislocalized, being retained in the Golgi apparatus. Furthermore, coexpression of the P132L mutant along with the wild-type caveolin-1 cDNA resulted in the retention of the wild-type caveolin-1 protein in the Golgi, indicating that the P132L mutant acts in a dominant negative manner, promoting the retention of wild-type caveolin-1. This suggests that a single allelic adulteration of the CAV-1 gene, the P132L mutation, is sufficient to render caveolin-1 inactive in cells.


In gastrointestinal cancers, the role of caveolin-1 seems less clear. For instance, the expression of both the caveolin-1 mRNA and protein in human colon carcinoma cell lines is reduced (4). In contrast to these findings, several studies have shown elevations in caveolin-1 expression in colonic tumor tissue. In comparing normal colonic mucosa to that of an adenocarcinoma, it was found that caveolin-1 levels were increased in ∼78% of adenocarcinomas, compared with 6% of normal epithelial tissue samples (59). In a similar study, it was found that up to 45% of patients with squamous cell carcinoma of the esophagus had elevated levels of caveolin-1 expression (109). Furthermore, caveolin-1 expression negatively correlated with patient long-term survival and prognosis.

Yet, in an examination of frank colonic tumors, it was shown that ∼67% of these colon cancers had reduced levels of caveolin-1 expression. When a cDNA encoding full-length caveolin-1 was transfected into several colon cancer cell lines, the ability of these cells to form tumors in nude mice was decreased drastically (4).

Thus it is difficult to draw any definitive conclusions about the contribution of caveolin-1 expression levels to the pathogenesis of gastrointestinal cancers in the context of this apparent dichotomy. Perhaps additional cell-type or tissue-specific alterations, in combination with the dysregulation of caveolin-1 expression, are necessary to promote the pathogenesis of these cancers (122).


Although no specific mutations have been identified in the caveolin-1 gene in either prostate cancers or prostate cancer-derived cell lines, caveolin-1 is upregulated in many primary prostate tumors. Additionally, caveolin-1 expression is even further upregulated in metastatic prostate cells (288). This observation is supported by the detection of a significant number of lymph node metastases in both human and mouse prostate cancers that expresses caveolin-1. Interestingly, normal prostate epithelial cells, which are the likely precursors of malignant cells, have undetectable levels of caveolin-1.

During the process of prostate cancer metastasis, focal areas in the primary tumor upregulate caveolin-1, a process that may be androgen dependent. Indeed, testosterone transcriptionally activates caveolin-1 expression via an androgen receptor-dependent mechanism (282). Furthermore, when stably transfected with caveolin-1 antisense cDNA, metastatic prostate cancer cell lines result in tumors that are 10% smaller in intact mice, but 40% smaller in castrated animals, compared with vector alone (259). Additionally, both intact and castrated mice showed a 17% decrease in incidence of metastasis and 52% reduction in tumor volume. These data suggest that caveolin-1 expression in androgen-responsive mouse prostate cancer cells suppresses apoptosis in vitro and in vivo.

Although the expression of caveolin-1 appears to be associated with the development of metastatic prostate cancer, the role caveolin-1 plays in malignant progression remains controversial. Indeed, caveolin-1 is a suspected tumor suppressor, so how can caveolin-1 contribute to the metastatic progression of prostate cancer via antiapoptotic actions and act as a tumor suppressor role in other cells? The capacity of this alternating function of growth suppression and growth promotion for caveolin-1 has not yet been resolved.

Caveolin-1 has been detected in the serum of patients with prostate cancer, an observation supported by detectable caveolin-1 in mouse and human prostate cancer cell conditioned media. Furthermore, secreted caveolin-1 is bioactive, capable of stimulating cell viability and clonal growth of a prostate cancer cell line, a process inhibited by neutralizing caveolin-1 antibodies (162). The phosphorylation of caveolin-1 on serine-80 within the ER converts caveolin-1 from an integral membrane protein to a luminal secretory protein (136, 218).

Thus conversion of caveolin-1 to a secreted protein provides yet another mechanism to subvert the tumor suppressor effects of caveolin-1 by altering its membrane topology. For example, phosphorylation at this site results in the regulated secretion of caveolin-1 from human prostate cancer cells that, in turn, can stimulate cancer cell growth in an autocrine/paracrine manner (255, 282).

Caveolin-1 expression is also associated with the metastasis of lung and pancreatic adenocarcinomas and negatively correlates with patient survival (95, 252).


The development of resistance to a variety of chemotherapeutic agents is often the primary cause for treatment failure in cancer patients. In vitro studies have shown that cells that develop resistance to one type of drug often have resistance to a multitude of other structurally and functionally distinct compounds, a phenomenon termed multidrug resistance (MDR). A variety of cellular changes have been found to accompany the development of the MDR phenotype, including the “classic” activation of P-glycoprotein, an ATP-dependent drug efflux pump (12). Additionally, increases in plasma membrane sphingolipid and cholesterol content, the components of lipid rafts and caveolae, often accompany the development of MDR. Such changes are thought to influence the action of P-glycoprotein, which has been shown to associate with lipid rafts and caveolae, by altering the availability of drugs to the cellular efflux machinery (120). In addition, caveolin-1 has been identified as an upregulated gene in a number of MDR human cancer cell lines derived from a variety of primary tumor sources, including breast, colon, ovary, and lung (119, 287). Many of these cell lines also show increased expression of caveolin-2, as well as an increase in the number of identifiable surface caveolae. In MDR, increases in caveolin-1 expression may serve to accelerate the efflux of drugs by facilitating the transport of these substances from intracellular compartments to plasma membrane efflux pumps, via routing through the intracellular cholesterol transport pathway (133).

Although increased expression of caveolin-1 may enhance the MDR phenotype, it may also serve to decrease the transformed phenotype of the same cells by promoting cell adhesion and senescence. As mentioned above, several proproliferative pathways are negatively regulated by caveolin-1, including cyclin D1 expression and Ras-p42/44 MAP kinase cascade activation, a process that would promote cell senescence. In addition, it has been reported that MDR cells undergo a “reverse transformation,” demonstrating a reduced ability to form secondary tumors in athymic mice, with a concomitant decrease in anchorage-independent growth (7). Thus the acquisition of MDR may allow cells to survive high concentrations of cytotoxic agents, but it conversely impedes their transformed phenotype.

This process of sequential loss/induction of regulatory genes is well documented during the process of cancer metastasis (56). The widespread association of caveolin-1 and caveolae with multiple signal transduction pathways suggests that the loss of caveolin-1 serves as a common point that functionally regulates many signaling pathways and, consequently, a plethora of downstream events. In this model, the combined loss of caveolin-1, with that of loss of a tumor suppressor or addition of an activated oncogene, may be sufficient to 1) induce cell transformation and/or 2) potentiate the transformed phenotype, by disrupting associated downstream signaling events involved in cell cycle regulation.

2. Glucose metabolism, insulin signaling, and diabetes

Type II diabetes is a genetically heterogeneous metabolic disease characterized by chronic hyperglycemia and abnormalities in lipid metabolism afflicting nearly 5% of the Western world. While a definitive understanding of the molecular mechanisms involved in the pathogenesis of this disease has remained elusive, numerous genetic candidates have been identified that, when dysfunctional, result in defective peripheral insulin action. Shortly after their initial discovery, caveolins were implicated in insulin signaling and have thus become an interesting candidate disease genes in type II diabetes.

Initial interest in caveolae as sites for insulin signaling came from ultrastructural studies of rat adipocytes, which showed that gold-labeled insulin ligand clustered within plasmalemmal caveolae (78, 79). Further studies showed that caveolae are highly enriched in adipocytes and that both caveolin-1 mRNA and protein levels increase over 25-fold during the differentiation of mouse 3T3-L1 fibroblasts into adipocytes (an insulin-dependent phenomenon), also suggestive of a role for these microdomains in insulin signaling (215). Additional experimentation revealed that insulin treatment of these cells results in the association of the glucose transporter GLUT4 with caveolin-enriched membrane microdomains, as well as the tyrosine phosphorylation of caveolin-1 on residue 14 (22, 31, 32, 112, 122, 215). Intriguingly, in adipocytes, caveolin-1 tyrosine phosphorylation is specific for insulin, as EGF, platelet derived growth factor (PDGF), tumor necrosis factor (TNF)-α, or interleukin (IL)-6 did not result in any measurable caveolin-1 phosphorylation (122).

In addition to these findings, several investigators have used biochemical means to assess the contribution of caveolin/caveolae to insulin signaling. Because caveolae are highly enriched in cholesterol, agents such as β-methyl-cyclodextrin (βMCD), which bind and sequester membrane cholesterol, can be used to reversibly flatten caveolae and disrupt signaling through these microdomains. With the use of this methodology, it has been shown that βMCD-treated rat adipocytes have a significantly reduced response to insulin stimulation, as both ATP citrate lyase phosphorylation and glucose uptake were greatly decreased (84). Replenishment of plasma membrane cholesterol resulted in morphologically identifiable caveolae and a reversal of this phenotype, data suggesting that caveolae/caveolin are involved in insulin signaling. Further studies corroborated these findings, demonstrating that cholesterol depletion resulted in decreased activation of downstream insulin responsive elements, such as insulin receptor substrate (IRS)-1 and protein kinase B (PKB)/Akt (186).

Using cell fractionation schemes to biochemically purify caveolae and their resident proteins, several groups have investigated the caveolar localization of the insulin receptor with mixed results. Mastick and colleagues (31, 84, 163, 286) used two properties of caveolae and lipid rafts to isolate these domains from insulin stimulated 3T3-L1 adipocytes: 1) their relative resistance to extraction in Triton X-100 at 4°C and 2) their propensity to “float” on a discontinuous sucrose gradient. These authors found that while up to 2% of the total phosphorylated insulin receptor is insoluble in cold Triton X-100, only a very small percentage (0.3%) cosediments with caveolar resident proteins such as caveolin-1, suggesting that the insulin receptor does not localize to caveolae (31). In support of these findings, Muller et al. (163) prepared cellular extracts using either ice-cold Triton X-100 or sodium carbonate, followed by sucrose density gradient centrifugation. Western blot analysis of the subsequent fraction revealed that the majority of the insulin receptor was excluded from the “floating” caveolar pool (163).

While these results are indicative of the separation of insulin receptor from caveolae, at least two studies have presented clear contrary evidence (84, 286). In one of these studies, caveolar resident proteins were isolated by a detergent-free method followed by sucrose gradient centrifugation. Interestingly, this methodology yielded an ∼36-fold enrichment of the insulin receptor in the caveolar fraction, while excluding noncaveolar proteins (i.e., clathrin) (84). In addition, when caveolae were isolated based on their detergent resistance, it was found that the insulin receptor no longer cofractionated with caveolin-1, suggesting that caveolar resident proteins may have varied degrees of detergent resistance.

Perhaps the most convincing data regarding the role of caveolin-1 in insulin signaling and diabetes came with the development of the Cav-1 null mouse. Although this mouse was not found to be overtly diabetic, as might be predicted from previous experiments, Cav-1 null mice do develop an interesting metabolic phenotype. Preliminary analysis of this phenotype showed that Cav-1 null mice are resistant to diet-induced obesity and developed progressive adipose tissue atrophy (197). As seen in Figure 11, a variant of diet-induced obesity, i.e., interbreeding Cav-1 null mice with LepR (db/db) mice, leads to a very pronounced phenotype whereby LepR (db/db)/Cav-1 (−/−) null mice weigh approximately half that of LepR (db/db) Cav-1 (+/+) wild-type mice at 1 year of age [95.01 ± 2.08 vs. 48.5 ± 1.04 (SE) g; P < 0.00005].

FIG. 11.

Loss of caveolin-1 rescues the phenotype of LepR (db/db) mice. Cav-1 null mice were interbred with LepR (db/db) mice to produce the corresponding two cohorts: LepR (db/db)/Cav-1 (+/+) and LepR (db/db)/Cav-1 (−/−). The mass of 1-year-old male mice that have been fed a standard Chow diet are represented for each cohort, respectively [95.01 ± 2.08 vs. 48.5 ± 1.04 (SE) g; P < 0.00005].

Metabolically, these abnormalities in Cav-1 (−/−) mice are characterized by elevated free fatty acid and triglyceride levels, decreased leptin and Acrp30 levels, and no changes in plasma insulin or glucose levels (197). However, when these mice were later placed on a high-fat diet for 9 mo, they were found to develop postprandial hyperinsulinemia (30). Additionally, when young Cav-1 null mice were challenged with an insulin tolerance test (ITT), they showed markedly decreased glucose uptake compared with wild-type control animals. These metabolic derangements are similar to those seen in prediabetic individuals in the human population, suggesting that caveolin-1 does indeed play a critical role in insulin signaling in vivo.

Further analysis of these physiological abnormalities revealed that Cav-1 null mice have an ∼90% decrease in detectable insulin receptor levels, selectively in adipose tissue (30). Furthermore, insulin stimulation of these mice did not cause measurable activation of downstream targets, such as PKB/Akt and GSK-3β, indicating that caveolin-1 is necessary for insulin's action in adipocytes. In Cav-1 null MEFs, transfection of the full-length caveolin-1 cDNA restored insulin receptor levels, as seen by both Western blot and immunofluorescence microscopy (Fig. 12). This effect was attributed to a stabilizing effect of the caveolin-1 scaffolding domain, rescuing the insulin receptor from proteasomal degradation, thus providing a molecular mechanism to explain these findings (30).

FIG. 12.

Recombinant expression of full-length caveolin-1 in Cav-1 (−/−) null MEFs rescues insulin receptor protein expression. A: Western blot analysis of lysates from wild-type (WT) and Cav-1 (−/−) null mouse embryonic fibroblasts (MEFs) reveals that WT cells contain significantly more insulin receptor than those of the Cav-1 null cells. B and C: transient transfection of the full-length caveolin-1 cDNA (+), but not vector alone (−), into Cav-1 null MEFs (B) or HEK 293 cells (C) significantly increases the expression of the insulin receptor, as seen by Western blot. D: transfection of the full-length caveolin-1 cDNA rescues insulin receptor expression in Cav-1 null MEFs, as assessed by immunofluorescence microscopy. Under phase contrast, four different Cav-1 null MEFs can be seen, of which only one received the caveolin-1 cDNA (arrow). Immunofluorescence microscopy of these cells following colabeling with anti-caveolin-1 IgG (green) and anti-insulin receptor IgG (red) revealed that the cell that was transfected with the caveolin-1 cDNA (arrow) showed an increase in insulin receptor protein expression, while the other three cells (arrowheads) did not. [Adapted from Cohen et al. (30).]

Next, we directly investigated the role of the caveolin-1 scaffolding domain on insulin receptor stability by incubating cells with a plasma membrane-permeable derivative of the scaffolding domain peptide. When Cav-1 null MEFs were treated with the caveolin-scaffolding domain, expression levels of the insulin receptor were dramatically increased, as assessed by Western blot (27). Thus the scaffolding domain alone is sufficient to functionally stabilize the insulin receptor.

In light of current data, caveolin-1 can be thought of as a major player in insulin signaling in tissues where it is expressed; however, loss of caveolin-1 is not sufficient to produce fulminant diabetes. This is consistent with other mouse models, specifically the adipose tissue selective insulin receptor knockout mouse (FIRKO) (9). These mice phenotypically have relatively mild full-body insulin resistance despite a significant reduction in the insulin-responsiveness of isolated adipocytes.

Also of note are findings that a subset of patients with severe insulin resistance were found to have mutations in the caveolin-binding motif of the insulin receptor (Table 3). These mutations led to the rapid degradation of the insulin receptor when expressed in cultured cells, similar to the degradation observed in Cav-1 null animals (100, 101, 103, 157, 158). Taken together, these findings further support the idea that caveolin-1 stabilizes the insulin receptor against degradation in the human population and that perhaps some diabetic patients exist with caveolin-1 mutations.


Mutations of the caveolin-binding motif within the human insulin receptor are associated with insulin resistance and insulin receptor instability

3. Modulation of lipid storage and breakdown

Recently, it has been shown that caveolin-1 and -2 can be directed to lipid droplets, in a variety of cell types (64, 178, 195). Fujimoto et al. (64) found that the β-isoform of caveolin-2, an endogenous NH2-terminal truncation of the caveolin-2 molecule, is constitutively localized to the surface of lipid droplets, by immunofluorescence and immunoelectron microscopy. In addition, these authors found that disruption of vesicular transport with brefeldin-A caused the further accumulation of caveolin-2, as well as caveolin-1, in lipid droplets (64). A similar report found that caveolin-1 accumulated in lipid droplets when it was modified by the addition of an ER retrieval sequence, thus forcing its retention in the ER (178). Furthermore, truncation mutants of all three caveolin isoforms have been localized to lipid droplets (195). More recently, a proteomics screen of isolated lipid droplets derived from Chinese hamster ovary (CHO)-K2 cells identified that caveolin-1 is normally found as a component of lipid droplet membranes. Furthermore, the amount of caveolin-1 present in these droplets increased with lipid loading (138).

A latter study identified a short hydrophobic stretch of amino acids (residues 101–134) within caveolin-1 that was required for targeting to lipid droplets. Furthermore, mutational insertion of select leucine residues within this hydrophobic domain blocked functional lipid droplet targeting of caveolin-1 (179). While these studies suggest that caveolin-1 may modulate lipid droplet structure or function, further work is necessary to elucidate the contribution of caveolin-1 to this important process.

The abnormalities in lipid homeostasis presented by Cav-1 null mice suggest relevance to the pathogenesis of metabolic human diseases. Indeed, the metabolic derangements found in Cav-1 null mice are characterized by postprandial hypertriglyceridemia, elevated serum free fatty acids (FFA), and elevations in serum very-low-density lipoproteins (VLDL)/chylomicrons. These phenotypic characteristics parallel many of those associated with type II diabetes, suggesting the importance of caveolin-1 in normal lipid metabolism.

In a direct physiological assessment of this proposed function, our group explored the role of caveolin-1 in lipid droplet metabolism, i.e., formation and breakdown, using the Cav-1 null mouse as a model organism. Lipid breakdown, or lipolysis, occurs through stimulation of β-adrenergic receptors by any number of agonists, which essentially all lead to the activation of protein kinase A (PKA). The two major downstream lipolytic targets in the adipocyte of PKA-mediated phosphorylation are the neutral lipid lipase, hormone sensitive lipase (HSL), and the lipid droplet coat protein perilipin (85, 107, 143, 147, 226, 257).

Upon PKA-mediated phosphorylation, HSL translocates from the cytoplasm to the lipid droplet, via an interaction with perilipin, where it acts upon stored triglycerides (15, 2426, 43, 85, 249, 253, 254). However, the activation of lipolysis is also dependent on PKA-mediated phosphorylation of perilipin-A (143, 144). It is thought that perilipin-A functions as a protective coat (Fig. 13), surrounding the lipid droplet until phosphorylated by PKA, whereupon it undergoes a conformational change, leaving the lipid droplet an open target for HSL and breakdown of stored triglycerides (2426, 143, 254).

FIG. 13.

Immunofluorescence microscopy of lipid droplets in a 3T3-L1 adipocyte. Fully differentiated 3T3-L1 adipocytes were fixed in paraformaldehyde, permeablized with 0.02% Triton X-100, and colabeled with anti-perilipin IgG (red) and the neutral lipid-specific stain Bodipy 495/503 (green). Note that perilipin (red) forms a ring around the lipid droplets (green) in these cells. Original magnification, ×60.

In our recent study on this subject, we found that Cav-1 null mice have a severely blunted response to both pharmacological and physiological lipolytic agonists (29). Furthermore, isolated perigonadal adipocytes derived from Cav-1 null mice responded poorly to lipolytic stimuli compared with wild-type cells. Further studies revealed that PKA activity was dramatically increased in Cav-1 null fat pads; however, this increased activity did not translate into perilipin phosphorylation. In addition, we found that β-adrenergic receptor stimulation of 3T3-L1 adipocytes resulted in coimmunoprecipitation of perilipin, caveolin-1, and the catalytic subunit of PKA. This ligand-dependent complex formation between perilipin and PKA was clearly defective in Cav-1 null adipocytes. These results suggest that perilipin phosphorylation, and thus functional lipolysis, is dependent on complex formation between perilipin and the catalytic subunit of PKA, and that caveolin-1 facilitates this interaction.

In this same study, we also identified a role for caveolin-1 in de novo lipid accumulation and lipid droplet stabilization. Wild-type and Cav-1 null MEFs, into which the perilipin-A cDNA had been stably transfected (Peri-MEFs), were loaded for 24 h with oleic acid to promote lipid storage (29). Subsequent analysis of these cells revealed that caveolin-1 is normally necessary for lipid accumulation and lipid droplet stabilization, as Cav-1 null Peri-MEFs contained significantly less lipid than wild-type Peri-MEFs. Furthermore, the lipid droplets in these cells were generally smaller and significantly less abundant (Fig. 14A). In addition, we noticed an electron-dense area surrounding the lipid droplets of wild-type, but not Cav-1 null Peri-MEFs (Fig. 14B). Ultrastructural analysis of wild-type and Cav-1 null perigonadal fat pads, using a relatively new technique involving high-pressure freeze substitution, revealed a morphological correlate of the physiological differences we found in lipolysis and lipid droplet formation in the absence of caveolin-1. In wild-type but not Cav-1 null adipocytes, we observed an electron-dense band surrounding the lipid droplet that most likely represents a conglomeration of various proteins, similar to that observed in wild-type Peri-MEFs, but much more pronounced (29). Indeed, many recent studies have identified several proteins that specifically target to the lipid droplet, including TIP47, adipophilin, perilipin, and S3–12 (14, 82, 142, 156, 174, 258, 275, 276). These proteins all have certain structural similarities and are thought to be involved in modulating lipid storage. Some of these proteins, along with other yet unidentified proteins, may be components of the electron-dense band that we observe in wild-type adipocytes.

FIG. 14.

Ultrastructural analysis of perilipin-A expressing mouse embryonic fibroblasts (Peri-MEFs) after lipid loading. A: as seen by transmission electron microscopy, lipid droplets (arrows) of wild-type Peri-MEFs appear larger and more abundant than those of Cav-1 (−/−) null Peri-MEFs following incubation with oleic acid for 24 h (to promote lipid accumulation). Bar, 500 nm. B: densitometric quantification of the area directly adjacent to the lipid droplet reveals a region of significant electron density in wild-type Peri-MEFs, but not in Cav-1 null Peri-MEFs. This area most likely represents a conglomeration of lipid droplet structural proteins (i.e., perilipin, TIP47, and S3–12) as well as many unidentified proteins. Bar, 200 nm.

4. Vascular abnormalities: hypertrophic cardiomyopathy

Hypertrophic cardiomyopathy (HCM) affects nearly 1 in 500 individuals in the general population resulting in significant morbidity and mortality. A wide variety of genetic mutations have been identified that result in a heterogeneous clinical presentation and patient course. However, the one commonality that exists among most of the mutated genes is that they encode cardiac myocyte contractile proteins. Yet, regardless of the gene involved, hallmark findings in patients with HCM include cardiomyocyte hypertrophy, interstitial inflammation and fibrosis, myocyte degeneration, and cardiac dysfunction.

Because caveolin-1 expression is limited to the supporting cell types of the heart (i.e., fibroblasts and endothelial cells), it was somewhat surprising that Cav-1 null mice develop cardiac disease similar to HCM of humans. To date, two independent studies examining the effects of caveolin-1 gene ablation on cardiac structure and function have been published, with somewhat disparate results (28, 293). In examining our Cav-1 null mice at three time points, 2, 4, and 12 mo of age, by cardiac gated magnetic resonance imagining (MRI) and transthoracic echocardiography (TEE), our group found that loss of caveolin-1 leads to progressive concentric left ventricular hypertrophy and severe right ventricular dilation (28). Together, MRI and TEE produce the most reliable measure currently possible of several cardiac parameters, including end diastolic and systolic diameters, relative anterior, posterior and septal wall thicknesses, as well as fractional shortening (28).

In a similar study of this cardiac phenotype, Zhao et al. (293) used TEE to observe that their independently generated Cav-1 null mice developed a dilated cardiomyopathy, evident at 5 mo of age. The different phenotypes reported by these two groups may be due to several factors including the genetic background of the mice, the choice of anesthetic used, as well as the limitations of a single methodology (TEE) used by the later group for evaluating cardiac wall thickness. In our studies, we found that Cav-1 null mice are exquisitely susceptible to commonly used inhaled anesthetics, resulting in bradychardia and artificial chamber dilation at dosages that do not produce these effects in wild-type mice (unpublished observation). Because of this, we used the anesthetic chloral hydrate, which has very limited cardiac effects and produced data corroborating that obtained by MRI (28). It is unclear which anesthetic was used by Zhao et al. (293), but we suspect that differences in the anesthetic used may explain these apparently conflicting results (293).

The molecular mechanism responsible for the cardiac phenotype observed in Cav-1 null mice involves a reversion to a state of early/fetal gene transcription, as is often the case in cardiomyopathies. Both groups noted a marked increase in atrial natriuretic factor (ANF) in ventricular tissue from Cav-1 null mice (28, 293). In addition, histopathological examination of Cav-1 null mouse hearts revealed areas of myocyte necrosis and foci of interstitial inflammation, as well as prevalent fibrotic lesions. Mechanistically, these pathological abnormalities were accompanied by hyperactivation of the p42/44 MAP kinase cascade, specifically in areas of interstitial fibrosis (28). Isolation of primary cardiac fibroblasts demonstrated that the loss of caveolin-1 in this cell type leads to hyperactivation of the p42/424 MAP kinase cascade. It is conceivable that, consistent with the literature, elevations in the p42/44 MAP kinase cascade in the cardiac fibroblast result in the secretion of growth factor molecules, such as endothelin-1 and transforming growth factor-β, which then promote the hypertrophic response in neighboring cardiac myocytes (28).

These and other cardiovascular phenotypes in caveolin-knockout mice are summarized in Table 4

View this table:

Cardiovascular phenotypes in caveolin-knockout mice

5. Vascular abnormalities: increased nitric oxide production

Initial reports of Cav-1 null mice demonstrated the unequivocal role of this protein in the negative regulation of nitric oxide (NO) production. Experiments performed in isolated aortic rings, which measured both vasoconshowed and vasorelaxation responses of the tissue, showed that caveolin-1 is essential for regulating vascular tone via modulation of eNOS activity.

Isolated aortic rings from Cav-1 null animals were unable to establish a steady-state contractile tone, oscillating at a frequency of once per minute (40). When acetylcholine was added to the media to stimulate vasorelaxation, the aortic rings from Cav-1 null mice showed a significantly greater (maximum 100%) relaxation response than wild-type control rings (maximum 45%) (198). Additionally, tension measurements showed that Cav-1 null aortic rings failed to contract in response to the α1-adrenergic agonist phenylephrine to the same extent as wild-type aortic rings (198). These effects could be abrogated by the addition of the potent eNOS inhibitor NG-nitro-l-arginine methyl ester (l-NAME), thus implicating increased NO production in this phenotypic characteristic (198).

6. Vascular abnormalities: impaired angiogenesis

Angiogenesis is a process that describes the formation of new blood vessels from the preexisting vasculature, through the proliferation of endothelial and smooth muscle cells. Angiogenesis is a normally quiescent process that is activated during such conditions as wound healing or changes in female reproductive tissues, associated with the preparation and maintenance of pregnancy. In addition, angiogenesis represents a hallmark feature of tumorigenesis, an essential component for tumor growth. Previous in vitro studies have implicated caveolin-1 in this process, demonstrating that caveolin-1 expression positively correlates with capillary tubule formation (83, 135). In accordance with these findings, Woodman et al. (277) demonstrated that Cav-1 null mice implanted with Matrigel plugs (a complex mixture of extracellular matrix components) combined with the angiogenic agent fibroblast growth factor-2 show a significant reduction in blood vessel infiltration and vessel density compared with wild-type mice (277). In addition, these authors found that the subcutaneous injection of the melanoma cell line B16-F10 resulted in significantly fewer and smaller primary tumors in Cav-1 null mice than in their wild-type counterparts. Histopathological and ultrastructural examination of these tumors showed that there was a decrease in blood vessel density and an incomplete formation of capillaries, also lacking identifiable endothelial cell caveolae. These data provide the first in vivo evidence indicating that caveolin-1-deficient endothelial cells have a diminished response to angiogenic growth factors and support the notion that loss of caveolin-1 can retard tumor growth via a diminished angiogenic response (277).

7. Vascular abnormalities: atheroprotection

Atherosclerosis is the most prevalent contributing factor to coronary heart disease, the leading cause of death in the Western world. The pathogenesis of atherosclerosis is most often initiated by injury to the vascular endothelium resulting in the accumulation of macrophages and LDL at the site of injury. Recruited macrophages will engulf and oxidize the accumulating lipids, resulting in 1) foam cell formation, 2) the release of various degradative enzymes and cytotoxic substances, 3) the subsequent denudation of the overlying endothelium, and 4) the accumulation of platelets that combine to form a plaque (168). Over time, this plaque increases in size and can eventually rupture producing vaso-occulsion and local ischemia, leading to a myocardial infarction or a cerebral vascular accident (CVA; stroke).

Caveolin-1 and endothelial cell caveolae have been postulated to play an important role in the development of atherosclerotic lesions, based on studies showing that oxidized LDL particles are taken up by caveolae and that the LDL scavenger receptor CD36 localizes to caveolae and interacts with caveolin-1 (63, 111, 132, 265). To directly address this issue, Cav-1 null mice were interbred with the atherosclerosis-prone apolipoprotein E null mice (194) to generate ApoE/Cav-1 double-knockout (dKO) mice. While Cav-1 null mice exhibit normal baseline plasma cholesterol levels and elevations in plasma triglyceride levels, interbreeding with ApoE null mice results in an approximately twofold increase in plasma cholesterol levels (62, 197). However, in spite of this normally “proatherogenic” hypercholesterolemia, loss of caveolin-1 conferred dramatic protection against atheroma formation (62). Gross examination of aortic preparations demonstrated that ApoE/Cav-1 dKO mice have an ∼70% reduction in lesion area, compared with ApoE null mice (Fig. 15). Mechanistically, these changes were associated with alterations in several proatherogenic endothelial cell molecules, such as CD36 and vascular cell adhesion molecule (VCAM)-1 (62). These alterations may affect the recruitment and migration of monocytes/macrophages into areas of endothelial lesions, as well as decrease the uptake and deposition of certain lipoproteins.

FIG. 15.

Loss of caveolin-1 is protective against atherosclerosis. Whole-mount en face visualization of Sudan IV-stained aortas demonstrates a significant reduction (∼65%) in the gross appearance of atheromatous lesions (red) in 6-mo-old male ApoE/Cav-1 dKO mice compared with matched ApoE KO mice, both fed a Western-type diet for 5 mo. [Adapted from Frank et al. (62).]

Cav-1 null mice also show other vascular defects, such as increased neointimal hyperplasia (smooth muscle cell proliferation) in response to vascular injury of the carotid artery (87). Interestingly, this phenotype appears to be due to hyperactivation of the Ras-p42/44 MAP kinase cascade and upregulation of Cyclin D1 in Cav-1 null vascular smooth muscle cells.

8. Abnormalities of the urogenital system

Cav-1 null male mice have recently been shown to exhibit severely compromised renal calcium reabsorption, leading to hypercalciuria and urolithiasis (17). Cao et al. (17) found that, by 5 mo of age, 67% of male Cav-1 null mice have developed noticeable early urinary bladder calculi, compared with 19% of wild-type male mice. At this same age, frank urinary bladder stone formation was observed in as many as 13% of Cav-1 null male mice, while no stones were detected in wild-type male mice of the same age. Additionally, neither stones nor calculi were observed in female Cav-1 and wild-type mice, indicating a multigenetic gender-specific cause potentiating stone formation (17). Investigating the mechanisms behind this phenomenon, it was shown that male Cav-1 null mice excrete significantly higher amounts of calcium in their urine, compared with wild-type male mice. Immunohistochemical analysis of the kidney revealed that caveolin-1 is normally expressed in the epithelial cells of the distal convoluted tubule and that a lack of caveolin-1 expression in these cells resulted in mislocalization of an important calcium transporter. Further analysis indicated that Cav-1 null mice show compromised renal calcium reabsorption, while maintaining otherwise normal kidney function. Thus this study illustrates the essential role of caveolin-1 in normally regulating urine calcium homeostasis (in males) and suggests that impaired caveolin-1 function may contribute to human calculi or stone formation.

Loss of caveolin-1 in male mice has also been found to lead to a syndrome similar to that of lower urinary tract dysfunction (LUTD) in elderly adult male humans. Interestingly, only the bladders of male Cav-1 null mice, but not Cav-2, -3, or -1/3 null mice, show pathological changes, despite the fact that the bladder is composed primarily of smooth muscle cells that express all three caveolin proteins (278). The loss of caveolin-1, but not -2 or -3, resulted in the absence of identifiable caveolae in the bladder, indicating that it is caveolin-1 that is necessary for caveolar biogenesis in smooth muscle cells. By 12 mo of age, male Cav-1 null mice exhibited significant bladder hypertrophy and several pathophysiological abnormalities, including increased baseline and spontaneous pressures of micturation and decreased agonist-induced bladder contraction. Furthermore, these mice exhibited significant fluid accumulation in the prostate, seminal vesicles, and kidneys, which eventually results in marked dilation and even necrosis of these organs. In addition, Cav-1 null prostates from 12-mo-old mice show a hypercellular and thickened fibromuscular stroma (Fig. 16) (278). In contrast to the study mentioned above, these authors did not observe any urinary bladder calculi, perhaps due to the differing genetic backgrounds of the mice used in each study (17, 278).

FIG. 16.

Histopathological examination of prostate specimens from 12-mo-old wild-type and Cav-1 null male mice. Microscopic evaluation of hematoxylin and eosin-stained prostate sections reveals severe thickening of the smooth muscle layer in Cav-1 null mice compared with wild-type mice (arrowheads) at 12 mo of age. However, there appears to be no histological change in the prostate epithelial lining (arrows). Medium: reduced from ×16; High: reduced from ×31.5. [Adapted from Woodman et al. (278).]

9. Ocular disease

The role of caveolin-1 in ocular disease has been proposed based in part on the observation that caveolin-1 is a specific component of rod cell outer segments disk membranes (10, 44). This pattern of caveolin-1 expression suggests a possible role in retinal degenerative diseases, such as retinitis pigmentosa.

Retinitis pigmentosa (RP) refers to a group of genetically heterogeneous progressive inherited retinopathies, characterized clinically by gradual visual field narrowing and classical pigmented lesions on fundoscopic exam. RP is generally the result of a single gene defect (monogentic); however, in at least one case, inheritance of defects in two unrelated genes (peripherin/rds and rom-1) appears necessary to cause RP (41, 93, 106, 141, 180, 205). While a multitude of genes have now been implicated in the pathogenesis of RP, the role of caveolin-1 is only now being elucidated (93).

The rod cell is a highly complex, specialized cell type composed of an inner and outer segment connected by a thin cilium. The inner segment is responsible for most of the cells metabolic needs, while the outer segment, composed of stacked membrane-bound disks, is responsible for light sensation. The aforementioned peripherin/rds and rom-1 proteins form oligomeric transmembrane complexes that localize to rod outer segment disks (141). In a recent study, rom-1 was found to be resistant to extraction with Triton X-100 at 4°C and shown to associate with lipid rafts and caveolin-1 (10). Because of this association, it is possible that defects in the Cav-1 gene may lead to a similar RP phenotype via mislocalization of rom-1. To address this possibility, we examined the eyes of 1-yr-old Cav-1 null mice for any histopathological changes consistent with RP. However, there were no significant differences between the retinal architecture of Cav-1 null mice and wild-type mice, in hematoxylin and eosin-stained sections (Fig. 17). Similarly, fundoscopic examination and fluorescein angiography revealed no retinal pigmentation or arteriolar narrowing consistent with RP (Fig. 17). Although these findings suggest that loss of caveolin-1 does not contribute to the pathogenesis of RP, it is possible that like rom-1 and peripherin/rds, mutations in Cav-1 constitute a yet unidentified cause of digenic RP.

FIG. 17.

Histological and fundoscopic evaluation of wild-type and Cav-1 null mouse eyes. Top: paraffin-embedded hematoxylin and eosin (H&E)-stained sections of 12-mo-old wild-type and Cav-1 null mouse retinas reveal no difference between the two genotypes and no pathological changes, consistent with an absence of retinal degeneration in Cav-1 null mice. Original magnification, ×20. Middle: visualization of the fundus and vasculature does not reveal any noticeable differences between wild-type and Cav-1 null mice at 12 mo of age. Bottom: similarly, fluorescein angiography of mouse retinal vessels does not show any changes in Cav-1 null mice, compared with wild-type mice. Note the web of interlaced capillaries observed by this method.

10. Reductions in life span

On the basis of the myriad of phenotypes described above, it was not surprising to find that loss of caveolin-1 conferred a significant survival disadvantage (182). Our group followed a large cohort of male and female wild-type, Cav-1 heterozygous, and Cav-1 null mice over a period of 2 years. There was an ∼50% reduction in life span, most evident between 27 and 65 wk of age, in Cav-1 null mice with no change in heterozygote animals (Fig. 18). Interestingly, Cav-1 male and female mice experienced similar declines in life span, thus abolishing the normal sexual dimorphism observed in the murine species. Histopathological examination of several organ systems indicated that the most likely cause of increased mortality in Cav-1 null animals was a combination of pulmonary fibrosis and cardiac hypertrophy. We found that 1-yr-old Cav-1 null animals showed progressive thickening of the left ventricular and intraventricular septal walls, as assessed by cardiac gated MRI and TEE (182). In addition, the lungs of Cav-1 null animals were severely fibrotic and hyperplastic, leading to increased right ventricular strain and most likely cardiac failure. We speculate that these mice die from an acute arrythmia secondary to hypertrophic cardiomyopathy.

FIG. 18.

Survival curves for wild-type, heterozygous, and Cav-1 null mice. Following a cohort of 180 mice for 2 years revealed that loss of caveolin-1 confers a significant survival disadvantage (an ∼50% reduction in life span) for both male and female Cav-1 null (KO) mice, compared with wild-type (WT) mice. This difference was most noticeable between 27 and 65 wk of age. Note that complete absence of caveolin-1 is necessary for this phenotype, as heterozygous (Het) mice did not show any changes in longevity. [Adapted from Park et al. (182).]

B. Caveolin-2

Caveolin-2 null mice were generated by targeted disruption of exons 1 and 2 of the murine Cav-2 gene (199). Initial characterization of these mice demonstrated unequivocally that caveolin-2 is not necessary for caveolae formation or the proper membrane localization of caveolin-1. In addition, analysis of several phenotypes described in the Cav-1 null mouse (i.e., vascular dysfunction and lipid imbalance) revealed that caveolin-2 plays no observable role in the pathogenesis of these abnormalities. However, Cav-2 null mice do display marked lung pathology, similar to that described in Cav-1 null mice (40, 198, 199). Because caveolin-2 expression is drastically reduced in Cav-1 null mice, the identification of this phenotype in Cav-2 null animals directly implicates a selective loss of caveolin-2 as the primary cause of these abnormalities.

1. Interstitial lung disease

A thorough histopathological examination of Cav-2 null lung specimens revealed significant abnormalities with marked alveolar hypercellularity (∼2-fold increase in nuclei) and septal thickening, compared with wild-type alveoli (199). Further analysis revealed that the lung hypercellularity appeared to be due to an increase in the number of vascular endothelial growth factor-receptor (VEGF-R) positively stained endothelial cells, corresponding to an approximately three- to fivefold increase in endothelial cell number. Also contributing to the noticeable septal thickening observed in these mice was as increase in extracellular matrix deposition as revealed by reticulin staining of basement membrane components (199). It should be noted that the exact composition of this matrix is variable, as Drab et al. (40) found significant fibrosis (as identified by trichrome staining) in Cav-1 null lungs during their description of this phenotype. Regardless, one measurable physiological consequence of these pulmonary abnormalities was severe exercise intolerance, as manifested by the early onset of exhaustion in Cav-1 null mice during a swimming test.

These pathological abnormalities are reminiscent of the interstitial lung diseases (ILDs) of humans, which, although due to a multitude of disorders, are all similarly characterized by progressive, irreversible fibrosis and severely compromised gas exchange (150). Most ILDs are caused by the inhalation of known agents, such as silica or coal dust pneumoconiosis, or occur secondary to systemic disorders, such as collagen vascular disease or sarcoidosis (23, 61, 248). However, a subset of ILDs remain idiopathic, or of unknown origin, and even a few of these have been found to cluster in families, suggestive of a genetic defect (283). The clinical course of idiopathic pulmonary fibrosis (IPF) is relatively variable; however, most patients progress to end-stage severe pulmonary fibrosis and death despite current therapy (54). Thus the Cav-2 null mouse may serve as a useful model organism to enhance further study of this disease process.

C. Caveolin-3

1. Specialized function: a muscle-specific isoform

Following the initial identification of caveolae in epithelial cells, several ultrastructural studies quickly identified abundant caveolae in striated muscle tissue, where they were originally thought to play a role in the formation of the T-tubule system (interconnected transverse membrane pockets that penetrate into the muscle fibers) (11, 102). With the discovery of the muscle-specific caveolin family member caveolin-3, the role of this protein in muscle function could finally be evaluated (256, 272).

Caveolin-3 is most closely related to caveolin-1, sharing 65% identity and 85% similarity. However, unlike caveolin-1, the expression of caveolin-3 is restricted to muscular tissue where it appears to be the exclusive caveolin member present (except in smooth muscle where all three caveolin family members are present) (256, 278). Similarly to caveolin-1, the sole expression of caveolin-3 has been shown to be sufficient to drive the formation of morphologically identifiable caveolae in cells (125). During the differentiation of skeletal myoblasts in culture, the levels of caveolin-3 mRNA and protein increase dramatically (242). Furthermore, anti-sense-mediated downregulation of caveolin-3 is sufficient to inhibit myotube fusion in vitro (69), suggesting that caveolin-3 expression is intimately linked to proper myocyte development. Caveolin-3 is transiently associated with the T-tubule system during early development, whereas in mature myocytes it is localized to caveolae of the muscle cell plasma membrane (sarcolemma) (189, 242).

Numerous studies have shown that the functional role of caveolin-3 in muscle cells is similar to that of caveolin-1 in other cell types. For instance, with regard to signal transduction, an analogous cohort of signaling molecules are known to localize to caveolin-3 generated caveolae, including NOS isoforms, various protein kinase C isoforms, α- and β-adrenergic receptors, and Src-family tyrosine kinases (55, 207, 208, 266). In addition to their role in sequestering signaling molecules, it also appears that muscle cell caveolae and caveolin-3 are important modulators of dystrophin-glycoprotein complex function, abnormalities of which are associated with a variety of human muscle diseases (242).

2. Pathogenesis of muscular dystrophy

The term muscular dystrophy (MD) refers to a broad range of phenotypically similar inherited myopathies characterized by progressive muscle degeneration and replacement with fibrous connective tissue. The prognosis of the disease is determined by the type of MD, as is the progression and severity of muscle weakening. The most common and severe form of MD is Duchenne (DMD), an X-linked (Xp21) recessive disease afflicting 1 in 3,500 young males, resulting in morbidity by early adulthood from respiratory failure (45). DMD is caused by a variety of mutations in the open reading frame of the dystrophin gene, resulting in a deficiency of the protein product and loss of function of the dystrophin-glycoprotein complex (DGC) (96, 97). The DGC is composed of several protein components, including the dystroglycans and the sarcoglycans, and spans the muscle sarcolemma, linking the cortical cytoskeleton with the extracellular matrix (96, 97, 227).

Dystrophin, as well as several members of the DGC including α-sarcoglycan and β-dystroglycan, cofractionate with caveolin-3 in cultured mouse C2C12 myocytes (242). In addition, coimmunoprecipitation experiments have shown that dystrophin forms a stable complex with caveolin-3 and that dystrophin and caveolin-3 colocalize by immunofluorescence microscopy in myocytes. The WW-like domain within caveolin-3 directly interacts with the COOH terminus of β-dystroglycan, a region containing a PPXY motif (245). Because dystrophin also contains a WW domain, it is thought that caveolin-3 can competitively inhibit dystrophin binding and can consequently alter the recruitment of dystrophin to the sarcolemma. In patients suffering from DMD, muscle biopsies show that there is both an upregulation of caveolin-3 expression as well as an increase in the number and size of caveolae at the sarcolemma (11, 201). This is consistent with the dystrophin-deficient mdx mouse, which has a similar phenotypic increase in the number of caveolae with elevated levels of the caveolin-3 protein in skeletal muscle (262).

While these studies provided preliminary suggestive evidence that changes in caveolin-3 expression might be linked to the pathogenesis of DMD, it remained unknown whether this was indeed the case. To address this issue directly, Galbiati et al. (68) used a transgenic approach to broadly overexpress caveolin-3 in mice. Pathophysiological examination of these mice revealed that they exhibit a DMD-like phenotype as early as 3 wk of age, characterized by 1) a dramatic increase in the number of skeletal muscle caveolae; 2) numerous hypertrophic, immature, and necrotic muscle fibers; 3) a preponderance of connective tissue infiltrates in skeletal muscle; 4) a near-compete loss of skeletal muscle dystrophin; and 5) a dramatic (∼4-fold) reduction in the expression of skeletal muscle β-dystroglycan. These changes were accompanied by elevations in serum creatine kinase levels, indicative of and consistent with the observed myonecrosis (68). Although this study provided strong support for the role of strictly regulated caveolin-3 expression in the pathogenesis of MD, it was not until the identification of two mutations in the human CAV-3 gene resulting in an autosomal-dominant form of limb-girdle MD (LGMD-1C), that the specific role of caveolin-3 in muscular pathogenesis emerged (152, 155).

Limb-girdle MD (LGMD) is a group of hereditary myopathies, including autosomal dominant and recessive forms that are clinically and genetically heterogeneous. Several autosomal dominant forms of LGMD have been recognized, including 1) LGMD-1A, linked to chromosome 5q; 2) LGMD-1B, associated with cardiac defects and linked to chromosome 1q11–21; and 3) LGMD-1C, linked to a missense mutation (P104L) in the membrane-spanning region and a 9-bp in-frame deletion in the CAV-3 gene that removes residues 63–65 (ΔTFT) in the scaffolding domain. These two mutations result in similar histopathological abnormalities, including a severe (∼95%) reduction of caveolin-3 expression in muscle tissue and a concomitant loss of sarcolemmal caveolae (154, 155). The LGMD-1C caveolin-3 mutant protein acts in a dominant-negative manner by forming unstable aggregates with the wild-type caveolin-3 protein. This results in the intracellular retention of both wild-type and mutant caveolin-3 proteins at the level of the Golgi and leads to their ubiquitination and proteasomal degradation (71, 72). More recently, a novel missense mutation (A45T) in the NH2-terminal domain of caveolin-3 has been identified in a patient with LGMD-1C (92). In contrast to the mutations discussed above, which occur within a highly conserved portion of caveolin-3, the A45T mutation does not. However, this mutation is also thought to result in the production of a dominant negative caveolin-3 protein, leading to similar retention and degradation of wild-type caveolin-3. Of interest, loss of caveolin-3 expression in skeletal muscle leads to a decrease in the expression of nNOS as well as α-dystroglycan, while β-dystroglycan levels remain unaffected (92).

Additional studies have shown that dysferlin, a skeletal muscle membrane protein whose deficiency causes distal and proximal recessively inherited forms of Miyoshi myopathy (MM) and limb-girdle muscular dystrophy type 2B (LGMD-2B), is mislocalized in skeletal muscle biopsies taken from LGMD-1C patients. Dysferlin contains caveolin-3 binding motifs and coimmunoprecipitates with caveolin-3 from normal skeletal muscle, suggesting that its interaction with caveolin-3 may serve an important function that remains undetermined (148).

Patients with LGMD-1C present with moderate proximal muscle weakness, with muscle cramps, calf-hypertrophy, and elevated serum creatine kinase (CK) levels, which are typical of muscle pathology (154). Elevated levels of serum CK (hyperCKemia) without muscle weakness have been linked to sporadic CAV-3 gene mutations that result in reduced caveolin-3 protein levels in skeletal muscle. Therefore, idiopathic hyperCKemia may be indicative of a partial caveolin-3 deficiency in skeletal muscle (19, 153).

Rippling muscle disease (RMD) is a relatively benign myopathy, first described in 1975, characterized by stretch-induced muscle contractions which spread to neighboring muscle fibers and give the appearance of ripples moving over the muscle (261, 284). Generally showing an autosomal dominant pattern of inheritance, RMD was originally thought to be caused by alterations in sarcoplasmic reticulum calcium homeostasis (261). However, using linkage analysis and positional cloning to identify the gene responsible for the pathogenesis of RMD in a large cohort of German families, Betz et al. (5) discovered that four previously identified missense mutations (R26Q, A45T, A45V, and P104L) in the CAV-3 gene on chromosome 3p25 were responsible for this phenotype. Specifically, the R26Q has been identified in patients with asymptomatic hyperCKemia, while the other mutations are associated with LGMD-1C (19).

Mutations in the CAV-3 gene provide an explanation for the allelism identified in the dystrophic (LGMD-1C) and nondystrophic (hyperCKemia and RMD) muscle pathologies (Table 5). Recently, Sotgia and co-workers (244, 246) addressed this issue directly, examining the behavior of three mutants (P104L, ΔTFT, and R26Q) in cultured cells (see Fig. 19). In these studies, it was shown that both the P104L and ΔTFT mutants result in the formation of unstable aggregates between wild-type and mutant caveolin-3 proteins that are targeted for proteasomal degradation. This leads to an ∼95% reduction in caveolin-3 protein levels. On the other hand, overexpression of the R26Q mutant did not reduce wild-type caveolin-3 levels and did not effect the cellular distribution of the wild-type protein (246). Thus the R26Q mutant does not behave in a dominant negative fashion.

FIG. 19.

In L6 skeletal muscle myoblasts, pathogenic mutants of caveolin-3 (P104L, ΔTFT, and R26Q) are all retained in a perinuclear/Golgi-like compartment. Caveolin-3 wild-type (WT) or mutant cDNAs (R26Q, P104L, or ΔTFT mutant) were used to transiently transfect L6 cells. At 36 h posttransfection, the cells were formaldehyde-fixed and immunostained with antibodies directed against caveolin-3. Unlike caveolin-3 WT, which localizes to the plasma membrane as expected, the R26Q, P104L, and ΔTFT mutants were all retained intracellularly, showing a perinuclear distribution pattern (see arrows). It is important to note that L6 myoblasts do not express endogenous caveolin-3 (not shown). N, nucleus. [Adapted from Sotgia et al. (246).]

View this table:

Disease-related mutations in the human caveolin-3 (CAV-3) gene

Studies of the P104L mutant in a transgenic mouse model have confirmed that this mutant results in a severe myopathic phenotype, loss of caveolin-3 expression, and identifiable sarcolemmal caveolae (251). In addition, these mice show significant increases in skeletal NOS activity, suggesting that nitric oxide may play a role in muscle fiber degeneration.

3. Caveolin-3 null mice

The generation of Cav-3 null mice provided a means by which to assess the functional role of genetic ablation of this gene in a model organism. Characterization of these mice by two independent groups showed that they lack muscle cell caveolae yet maintain normal levels of caveolin-1 and -2 expression as well as normal caveolae in nonmuscle tissue (67, 86). Both groups found that Cav-3 null mice exhibit mild myopathic changes, similar to those seen in patients with LGMD-1C. Muscle degeneration in Cav-3 null mice was recognized in the soleus muscle as early as 8 wk of age, while degeneration was noted in the diaphragm between 8 and 30 wk (67, 86). However, no growth or movement differences were identifiable in Cav-3 null mice, even at the extremes of age. In addition, heterozygous Cav-3 (+/−) mice did not demonstrate any noticeable muscle pathology, indicating that this phenotype was inherited in a recessive manner (86). These findings stand in contrast to those observed clinically, where LGMD-1C has been shown to be inherited in a dominant-negative manner (155).

Biochemical and ultrastructural analysis of skeletal muscle samples from Cav-3 null mice showed that the dystrophin-glycoprotein complex no longer associated with lipid rafts and that the T-tubule system was markedly disorganized, containing dilated and abnormally oriented tubules (67). Thus, unlike the severely myopathic transgenic P104L mutant mouse discussed above, a complete ablation of caveolin-3 produces a much milder phenotype, suggesting that an expressed dominant negative caveolin-3 mutant effects more than just caveolin-3 protein levels (251).

4. Cardiomyopathy

Further characterization of Cav-3 null mice revealed that they also develop a cardiomyopathic phenotype similar to that described above in Cav-1 null animals. Using cardiac gated MRI and transthoracic echocardiography, Woodman et al. (279) found that, at 4 mo of age, Cav-3 null mice showed significant cardiac hypertrophy and reduced fractional shortening (279). Histopathological examination of cardiac sections revealed perivascular fibrosis, myocyte hypertrophy, and cellular infiltration. Mechanistically, these changes were characterized by an exclusion of the dystrophin-glycoprotein complex from cardiac myocyte lipid rafts and hyperactivation of Ras-p42/44 MAP kinase cascade in Cav-3 null cardiac tissue. These changes are consistent with the known role of p42/44 MAP kinase activation in cardiac myocyte hypertrophy and the role of caveolin-3 as a negative regulator of the Ras-p42/44 MAP kinase cascade, similar to that of caveolin-1 (28, 279).

Interestingly, in screening patients with hypertrophic and dilated cardiomyopathies, Hayashi et al. (89) found that a caveolin-3 mutation (T63S) was associated with siblings with hypertrophic cardiomyopathy. This mutation occurs in the same position as the ΔTFT deletion associated with LGMD-1C (71). However, when expressed in cell culture, the T63S mutant does not reduce the expression of wild-type caveolin-3 to the same extent as the ΔTFT mutant, suggesting that subtle changes in caveolin-3 expression may result in a myriad of clinically relevant phenotypes.

D. Caveolin-1/3

The generation of truly caveolae-deficient animals was accomplished by interbreeding Cav-1 and Cav-3 null mice, to produce caveolin-1/3 double knockout mice (Cav-1/3 dKO). Surprisingly, although caveolae have been implicated in a plethora of cellular functions, Cav-1/3 dKO mice are viable and fertile, despite a complete absence of morphologically identifiable caveolae in muscle and nonmuscle cells (184). Additionally, these mice are deficient in the expression of all three caveolin family members, as caveolin-2 is unstable and degraded in the absence of caveolin-1. General phenotypic evaluation of these mice revealed that a combined loss of caveolin protein expression did not produce any alterations in the phenotypes previously identified in single knockout animals, with the exception of the cardiac defects.

These mice exhibit a combined phenotype of the individual caveolin null mice and present with a severe cardiomyopathy. At 2 mo of age, these mice show more pronounced increases in several parameters, including left ventricular wall thickness and septal thickness, than either the Cav-1 and Cav-3 null animals individually (184). Histopathologically, cardiac tissue appears hypertrophic, with areas of interstitial inflammation, perivascular fibrosis, and myocyte necrosis. Thus a combined loss of caveolin-1 and -3 has profound effects on cardiac structure and function, a tissue that normally expresses an abundance of both caveolin proteins, but in different cell types (cardiac fibroblasts vs. cardiac myocytes).


Curious plasma membrane-associated “little caves” were first recognized over 50 years ago by electron microscopists and dubbed “caveolae.” Since this time, the importance of these abundant organelles has proven both thought provoking and elusive. Caveolae have remained the focus of numerous research efforts to probe their biochemical properties and to explore their functional roles in various cellular processes.

With the cloning of the structural subunits of caveolae in the 1990s, the caveolin gene family emerged, finally allowing researchers a biochemical tool through which they could dissect caveolar function. A rapid succession of publications implicated caveolae and caveolins in a variety of important cellular processes, including vesicular trafficking, cholesterol homeostasis, and signal transduction. The identification of caveolins as oligomeric multivalent scaffolding proteins, responsible for the subcellular regulation of numerous signaling molecules, led to the proposal of the “caveolae signaling hypothesis” and implicated caveolins and caveolae in numerous disease processes. Ultrastructural, genetic, and molecular biological analysis of caveolae and caveolins in cell culture systems provided convincing evidence that these components were indeed involved in the pathogenesis of diseases, as disparate as muscular dystrophy, cancer, and diabetes. Thus it was with great enthusiasm that the development of caveolin-deficient mice was received, as these whole animal models provide the first opportunity to examine the role of caveolins and caveolae in an intact living organism. To date, the Cav-null mice have proven powerful tools in our understanding of caveolae/caveolins and have both provided fundamental support for previous hypotheses as well as given cause to reassess prior suppositions.

Despite the ablation of an entire cellular organelle following the deletion of either caveolin-1 or caveolin-3, it is rather surprising that all of the Cav-null mice are viable and fertile, including the complete “caveolae-less mouse” (Cav-1/3 dKO). These findings suggest that, like so many other disease-related genes, caveolins are not essential for life and that, indeed, mutations in the genes encoding the caveolins may be relevant to the pathogenesis of human disease.

It was similarly surprising to find that, although experiments in cell culture have demonstrated an intricate role for caveolae and caveolins in cholesterol trafficking, no abnormalities in plasma cholesterol levels were observed in Cav-1 null mice (197). However, when Cav-1 null mice were interbred with ApoE null mice, creating ApoE/Cav-1 dKO mice, Frank et al. (62) found noticeably higher plasma cholesterol levels in these mice than in the single null ApoE mice. These findings illustrate certain discrepancies observed between the properties of a protein in vitro compared with that observed in the context of a whole animal model. For instance, techniques involving ablation of a gene in vivo are confounded by inherent variables such as unknown compensatory mechanisms that exist in the setting of an entire organism but may not exist in cell culture models. In the case of cholesterol homeostasis, it may be that cholesterol and fatty acid transport proteins are upregulated in response to caveolin-1 ablation, thus accounting for the limited changes observed in Cav-1 null mice in this regard. Furthermore, these findings exemplify the unpredictable nature of physiological ablation of a gene and the subtle phenotypic variations that can only be deciphered when the model is challenged with a particular stressor.

Indeed, while a plethora of cell culture data indicate a predominant role for caveolin-1 in a variety of cancers, initial evaluation of Cav-1 null mice showed no increased incidence of spontaneous tumor formation. Subsequent challenges of these mice with further carcinogenic “stressors” (i.e., crossing the mice with MMTV-PyMT mice or by applying the carcinogen DMBA) increased their tumor incidence and showed that disruption of caveolin-1, in the context of additional tumor-promoting perturbations, facilitates the progression of cancer (18, 274). Thus it appears that loss of caveolin-1, like many other cancer-related genes, works in concert with coincident genetic mutations to enhance cell survival and accelerate tumor growth (a phenomenon known as cooperativity).

Analysis of Cav-1 null mice has also helped to further our understanding of insulin signaling and the pathogenesis of type II diabetes. Whether functional insulin signaling is truly dependent on intact caveolae has been a hotly debated topic; however, findings in Cav-1 null mice strongly indicate that insulin signaling is directly relayed via caveolae. Furthermore, the now dubious role of adipose tissue in whole body insulin resistance is supported by findings in Cav-1 null mice showing a ∼90% reduction in insulin receptor levels in adipocytes, without severe whole body insulin resistance, similar to findings reported in an adipose-tissue specific insulin receptor knockout (FIRKO) mouse (9, 30). It is through such studies that the true contribution of each insulin-responsive tissue to the diabetic phenotype can be assessed.

In addition, analysis of the various Cav-null mice has greatly expanded the current repertoire of diseases thought to be influenced by these proteins. Indeed, caveolin-1 has now been implicated in urinary tract function, a potential pathogenic component of both urinary calculi formation and LUTD, diseases that cause immense morbidity in the human population (17, 278). Similarly, caveolin-2 may be involved in the pathogenesis of idiopathic pulmonary fibrosis (IPF), and thus the Cav-2 null mouse may serve as an important model for this disease (199). Cav-3 null mice have served as a “proof of principle” showing that caveolin-3 is indeed important for muscular dystrophy, while also enhancing our understanding of the mechanisms by which mutations in CAV-3 may lead to muscle-related diseases (67, 86).

Undoubtedly, the next step in the process of understanding the role of the various caveolin family members in human disease is to apply what we have learned from the murine models to the human population. This has already occurred to some extent, as identification of the hypertrophic cardiac phenotype in Cav-3 null mice was rapidly followed by the identification of a mutation in CAV-3 causing hypertrophic cardiomyopathy in humans (89, 175, 279). In terms of caveolin-1, mutations in this gene have been found in a variety of human cancers, while no reports have been published identifying caveolin-1 mutations in other human diseases, such as diabetes and cardiomyopathy. Therefore, it will be necessary to examine other relevant human populations, to identify genetic mutations in caveolin family members, and thus potentially provide a molecular basis for the many human diseases that still remain idiopathic in nature.

Additionally, while the generation of the various caveolin-deficient mice has proven extremely useful in furthering our understanding of the role of these proteins in a variety of diseases, these model systems are inherently complex, in part, because the gene of interest is eliminated in the entire organism. This fact may partially obscure the nature of a given phenotype and the particular contribution of the cell type(s) involved. For instance, in studies regarding the role of caveolin-1 in the pathogenesis of breast cancer, it is difficult to distinguish the contribution of mammary epithelial cells from that of mammary adipocytes, both cell types that normally express caveolin-1. Therefore, it would be highly beneficial to create mice with a tissue or cell type specific and/or inducible targeted downregulation of the various caveolins, i.e., tissue-specific knockout animals. With such techniques, it would be possible to directly address the tumor-promoting effects of caveolin-1 specifically within mammary epithelial cells versus adipocytes in vivo. Perhaps mice such as these would unmask yet unidentified phenotypes, allowing for further characterization and understanding of the cellular roles of the various caveolins.

Numerous hypotheses exist regarding the structure and composition of the various subclasses of lipid rafts, of which caveolae are just one example (for review, see Ref. 149). In addition to the caveolins, several other proteins have been identified that may be enriched in lipid rafts resulting in dramatic changes in raft morphology. These proteins, referred to as MORFs (modifiers of raft function; see Ref. 200), include the caveolins, the flotillins (FLO-1 and -2), stomatins, LAT/PAG, VIP36, and MAL/BENE (Table 6). Like the caveolins, which when inserted into lipid rafts cause the invagination of the otherwise flat plasma membrane, these proteins may similarly lead to morphological and/or biochemical partitioning of lipid rafts. Future research will undoubtedly be necessary to address the function of these various proteins and their contribution to human disease.

View this table:

Modifiers of raft function (MORFs)


We are gratefully indebted to Dr. Paul Latkany (Assistant Professor, New York Medical College, Codirector of Uveitis, New York Eye and Ear Hospital, New York, NY) for help in evaluating the eyes of these mice.

This work was supported by grants from the National Institutes of Health and the Susan G. Komen Breast Cancer Foundation (to M. P. Lisanti). A. W. Cohen was supported by National Institutes of Health Medical Scientist Training Grant T32-GM-07288. M. P. Lisanti is the recipient of a Hirschl/Weil-Caulier Career Scientist Award.


View Abstract