Research into caveolae began with their morphological identification in 1953. By transmission electron microscopy, they appear as structures resembling 'little caves', which are vesicular invaginations of the plasma membrane of 50-100 nanometer (nm) in size 1. Caveolin-1 (also known as caveolin, Cav-1 or VIP21) was the first member of the caveolin family to be identified, and it was shown to be a structural component of caveolae and of transport vesicles derived from the trans-Golgi network 23. It was isolated as one of several proteins that became phosphorylated on tyrosine residues in chicken embryo fibroblasts transformed with the v-Src oncogene 4. Monoclonal antibodies directed against Cav-1 decorated the cytoplasmic protein coat of caveolae, making Cav-1 the first true protein marker of caveolae 2. Subsequent cloning of the Cav-1 cDNA revealed that it was identical to another protein, VIP21, which had been cloned almost simultaneously 35. Interestingly, VIP21 was isolated as an integral membrane protein component of transport vesicles derived from the trans-Golgi network in Madin-Darby canine kidney (MDCK) cells, suggesting that Cav-1/VTP21 may have a role in molecular trafficking as well as oncogenesis.

Caveolin-2 (Cav-2) and caveolin-3 (Cav-3) were identified in 1996 using different experimental methods. Cav-2 was discovered by the microsequencing of a 20 kDa protein that co-purified with adipocyte-derived caveolar membranes 6. Further characterization revealed that Cav-2 colocalizes with Cav-1 in caveolae, forms hetero-oligomers with Cav-1, is co-expressed in many of the same cells and tissues, and requires Cav-1 for proper membrane localization 78. Cav-3 (also known as M-caveolin) was identified through database searches and traditional cDNA library screening in an attempt to find Cav-1 homologs 910.

Caveolin sequences have been obtained from a range of vertebrates, including human, cow, mouse, Xenopus, and Fugu rubripes. A caveolin gene family has also been found in Caenorhabditis elegans 1112. The three mammalian genes encoding members of the caveolin family are similar in sequence (Table 1). A phylogenetic tree of all known caveolins shows that the C. elegans Cav-1 sequence is only distantly related to all the others (Figure 1). Interestingly, Cav-1 and Cav-2 are in very close proximity (about 19 kilobases (kb) apart) on human chromosome 7q31.1, while Cav-3 is located on a different chromosome (3p25) 1314. Although the evolutionary history of the caveolin genes has not been clearly defined, there are clues within their sequences and genomic organization to suggest possible mechanisms for their origin. For instance, although C. elegans Cav-1 has two exons, the region that is homologous to mammalian cave-olins is encoded by only a single exon, suggesting that mammalian caveolins are derived from this particular exon 11. Also, two observations derived from the human genomic sequence suggest that some family members may have arisen through gene duplication events: firstly, the exon-intron boundaries in the last exons of Cav-1, Cav-2, and Cav-3 are in analogous positions; and secondly, exon 2 of Cav-2 is divided into two parts (2a and 2b) by an intron, whereas the two homologous portions in the Cav-1 and Cav-3 sequences are fused together to form the final exon 1314. This second point may suggest that Cav-2 served as the genomic precursor of Cav-1 and Cav-3.

Currently, the structural features of this family of proteins are poorly defined, but information gleaned from the protein sequence has enabled some predictions of structure and motifs within the proteins. For instance, all three caveolins have an invariant 'FEDVIAEP' stretch (in the single-letter amino-acid code) within their hydrophilic amino-terminal domains that has come to be termed the 'caveolin signature motif' 69. The functional importance of this sequence or motif has yet to be determined, however.

Two Cav-1 isoforms (α and β) have been identified; the β isoform arises from an internal translational start site that gives a shorter amino terminus than that of the α form and is truncated by 31 residues 15. The predicted domains span almost the same number of residues in all three proteins: the amino-terminal domain comprises the first 101 residues in Cav-1α and the first 70-86 residues in Cav-1β, Cav-2, and Cav-3, with the putative transmembrane domain occupying 33 amino acids and the carboxy-terminal domain containing 43-44 amino acids (Figure 2).

Using a variety of experimental methods, it has been determined that the major sub-cellular location of Cav-1 is at the plasma membrane. From the primary sequence (hydrophilicity plots) and mutational analysis, Cav-1 is predicted to have a membrane-spanning hairpin-like structure, with both amino and carboxyl termini directed towards the cytoplasm (Figure 2). This atypical membrane-spanning model is supported by findings that antibodies directed against the Cav-1 amino or carboxyl terminus require cells to be permeabilized in order to bind Cav-1, that cell-surface biotinylation does not label Cav-1, and that there are known palmitoylation and tyrosine phosphorylation sites within both the amino- and the carboxy-terminal domains of the protein 16171819; palmitoylation and tyrosine phosphorylation are both cytoplasmically generated post-translational modifications.

Generally speaking, caveolins are small proteins (18-24 kDa). Structurally, however, perhaps one of the most interesting and significant findings about Cav-1 is that it forms an oligomeric complex comprised of approximately 14-16 monomers, as discovered through velocity gradient ultracentrifugation. In this assay system, Cav-1 was found to migrate as 200-400 kDa complexes 1620. Experiments with Cav-1 deletion mutants mapped the oligomerization domain to residues 61-101 16 (Figure 2). Interestingly, Cav-3 also forms large oligomeric complexes of approximately 350-400 kDa in vivo, whereas Cav-2 requires Cav-1 to participate in the formation of these high-molecular-mass complexes 789.

Caveolae are considered by many to be a subset of lipid rafts, which are highly-ordered microdomains residing within the plasma membrane that are enriched in certain lipids 212223; this may not be completely accurate, however, as some proteins are known to localize selectively to either lipid rafts or caveolae but not both 24. Caveolae-enriched membrane fractions can be purified efficiently on the basis of their buoyancy and resistance to solubilization by mild non-ionic detergents at 4°C 252627282930, but other types of membrane microdomains may also be enriched by this kind of purification.

Cav-1 localizes to plasma-membrane caveolae and also to the Golgi apparatus and trans-Golgi-derived transport vesicles 3531. Cav-1 may have a soluble cytoplasmic form, as well as a secreted form, depending on the cell type 32, and the first 31 amino acids may be important in selectively targeting isoforms of Cav-1 to different cellular compartments 33. Cav-1 is expressed ubiquitously, although at different levels in different tissues, with the highest levels in adipocytes, endothelial cells, fibroblasts, smooth-muscle cells, and a variety of epithelial cells. Cav-2 is tightly co-expressed with Cav-1, whereas Cav-3 is expressed predominantly in striated muscle cells 34. Interestingly, Cav-1 is required for the proper membrane localization of Cav-2.

Mice deficient in Cav-1, Cav-2, or Cav-3 are viable and fertile, but each has abnormal cellular and tissue-specific phenotypes peculiar to the specific ablated gene 353637383940414243. A mutant form found in up to 16% of human breast cancers, Cav-1 (P132L), does not localize properly to the plasma membrane and behaves in a dominant-negative manner, causing the mislocalization and intracellular retention of wild-type Cav-1 4445. An analogous P-to-L mutation in Cav-3 (P104L) has been detected in patients with autosomal dominant limb-girdle muscular dystrophy type-1C, and this mutation also behaves in a dominant-negative fashion 46474849. Wild-type Cav-3 localizes to caveolae and the plasma membrane and also associates with the T tubules that form from invaginations of the muscle membrane 5051.

Caveolin-related research has shown that caveolae function in vesicle trafficking 52, cholesterol homeostasis, signal transduction and tumor suppression. Endothelial caveolae may be involved in transcytosis; they have the molecular components used by other transport vesicles during vesicle formation, docking, and fusion 53, and the motor protein dynamin, which is important for vesicle fission, also localizes to caveolae 5455. Regarding endocytosis, it appears that certain ligands and extracellular molecules, such as cholera and tetanus toxins, are transported across the plasma membrane through caveolae, rather than via clathrin-dependent mechanisms 5657. Pathogens appear to have evolved mechanisms to gain entry into eukaryotic cells through caveolae, including simian virus 40 and certain strains of Escherichia coli 58.

Caveolae are enriched with cholesterol and Cav-1 is one of the few proteins that binds cholesterol tightly and specifically 5960; free cholesterol is required for the proper formation of caveolae 616263 and regulates the Cav-1 promoter 64. Intracellular cholesterol balance may be affected by caveolins 65, as a dominant-negative Cav-1 mutant causes intracellular retention of free cholesterol as well as a decrease in cholesterol synthesis and its efflux from the cell 66. Also, caveolae have been linked to the process of reverse cholesterol transport, during which excess free cholesterol is released into the blood plasma via uptake by high-density lipoprotein (HDL) particles 6768697071, and caveolae appear to be involved in the uptake of cholesterol esters from the plasma.

Caveolae appear to serve as signaling platforms by compartmentalizing and concentrating signaling molecules (this is referred to as the 'caveolae signaling hypothesis') 27. Various classes of signaling molecules, including G-protein subunits, receptor and non-receptor tyrosine kinases, endothelial nitric oxide synthase (eNOS), and small GTPases 2742, bind Cav-1 through its 'caveolin-scaffolding domain' (CSD) (Figure 2). Cav-1 also appears to inhibit the downstream activation and signaling of many proteins, including c-Src, H-Ras, mitogen-activated protein (MAP) kinases, and eNOS 72737475767778. The evidence that Cav-2 is a signaling modulator is less clear, partly perhaps because its CSD sequence is divergent from that of Cav-1. The Cav-3 CSD is very similar to the Cav-1 CSD, however, and Cav-3-generated caveolae have been shown to compartmentalize and modulate a number of signaling proteins, including eNOS, β-adrenergic receptors, protein kinase C isoforms, G proteins, Src-family kinases, and multiple components of the dystrophin-glycoprotein complex 3476777879.

Several lines of evidence have implicated Cav-1 in tumor suppression 18198081828384, and there is also accumulating evidence that Cav-1 has an anti-proliferative function. Furthermore, the Cav-1 and Cav-2 genes are close to the microsatellite marker D7S522 on human chromosome 7q31.1 1314, a region that is commonly deleted and implicated in the pathogenesis of many human epithelial-based cancers, including breast, colorectal, prostate, ovarian, and renal-cell carcinomas. Recent results 858687888990 strongly argue that Cav-1 functions either as a negative regulator of cell proliferation or as a tumor suppressor, in both cultured cells and whole animals.

The ubiquitous nature and diverse tissue expression of caveolin family members in mammals suggest that caveolins are indeed important for normal cellular and tissue physiology in highly evolved organisms. The discovery of a caveolin gene family in the invertebrate C. elegans 11 raises the questions of when caveolins joined the cellular repertoire and whether they are present in more primitive animals, plants or fungi. Another equally important area of research is deciphering the structure of caveolins, as such knowledge would greatly contribute to our understanding of how caveolins function. Recently, gene knockout and transgenic technology has facilitated the study of caveolins in mice, from a whole-organism point of view, allowing the generation of caveolin-deficient or caveolin-overexpressing transgenic mice. The molecular-genetic analysis of these caveolin-deficient mouse models, and cell lines derived from these animals, will greatly facilitate the progress of caveolae-related research into the next decade.