© ANDREIUC88/SHUTTERSTOCK.COMIn 1953, cell biologist George Palade used the recently developed technique of electron microscopy to examine the surface of the endothelial cells that form blood vessels. He saw in these mammalian cells that the plasma membrane, which forms the outer barrier of all cell types, was riddled with invaginations.1 The appearance of these folded structures was remarkable: uniformly flask-shape, they self-associated to form intricate, interconnected arrays.

Two years later, Japanese electron microscopist Eichi Yamada coined the term “caveolae,” from the Latin for “little caves,” to describe these invaginations. Caveolae have since turned up in several cell types, accounting for nearly half of all the plasma membrane surface of fat cells as well as endothelial cells. Yet more than half a century after Palade’s discovery, a complete understanding of the cellular function of caveolae remains frustratingly elusive.

One thing that is clear is that genetic data show...

Several models to explain exactly what caveolae are doing have been proposed: caveolae may pinch off from the surface of cells to act as transport vesicles, they may organize the surface of cells to facilitate signalling across the plasma membrane, and they may protect cells from mechanical damage. These different models are not mutually exclusive, and the task of unravelling the molecular mechanisms that support such diverse functions is ongoing. Given the amount of research on caveolae, it is surprising that there is not more conclusive experimental evidence of their cellular activities; relative to other subcellular compartments and structures, caveolae are still enigmatic.

Recent investigations into the formation of caveolae have provided a much more complete picture of the protein complexes that sculpt the plasma membrane into such distinctive shapes. One important implication of the new data is that the shape of caveolae is not fixed. Rather, the structures can undergo dynamic transitions between flat and invaginated states, and some of the recently identified protein components of caveolae may well be important for regulating these changes. This does not completely resolve the debate on the function of caveolae, but it does provide considerable support for the idea that caveolae act as buffers within the membrane to stop stretch forces from rupturing cells. In motile, multicellular organisms with a closed and pressurized vascular system, mechanical stretching is an important part of the functional environment of many different cell types, and defining a role for caveolae in withstanding stretch forces would open up this little-understood area of cell biology.

Four eras of discovery

We can divide the history of research on caveolae into four eras, each of which yielded different hypotheses for what these structures might be doing in cells. From Palade’s initial description in 1953 until the early 1990s, observations were driven by electron microscopy, and caveolae were defined by their characteristic shape and tendency to cluster in complex arrays.

The era of the molecular biology of caveolae began in 1992, when a landmark paper from the laboratory of Richard Anderson at the University of Texas Southwestern Medical School identified the first caveolae-associated protein, which the investigators called caveolin-1.2 This discovery of a defining component present specifically in caveolae triggered a series of biochemical studies linking the protein to a number of potential binding partners involved in cell signalling.

Genetics entered the fray with the 2001 development of mice lacking the gene for caveolin-1, CAV-1. These mice surprised the field by being relatively healthy and developmentally normal.3,4 Most recently, close molecular structural scrutiny of caveolae has revealed new clues to their function. In 2008, the laboratories of Rob Parton at the University of Queensland in Australia and Paul Pilch at Boston University School of Medicine showed that proteins now called cavins are important structural elements of caveolae.5,6 These findings both led to the beginnings of a detailed structural model for how caveolae assemble, and paved the way for experiments showing that caveolae are unexpectedly flexible.

The era of electron microscopy

In the initial era of caveolae research, much of the early discussion of function was informed by the abundance of these structures, particularly in endothelial cells. Why are there so many?

Controlled transport of albumin and other macromolecules between blood and tissue interstitial fluid is a crucial part of mammalian physiology, not least because it generates a concentration gradient of albumin to maintain fluid balance in the pressurized vascular system. Palade and others proposed that caveolae act as transport containers, or vesicles, carrying albumin from one side of the endothelium to the other. Many caveolae are needed to support the high flux of albumin and fluid, they reasoned.
CAPTURING CAVEOLAE: The structures were first imaged on endothelial cells by electron microscopy in 1953, but their function is still largely mysterious. IMAGES BY GILLIAN HOWARD, COURTESY OF BEN NICHOLS, MRC
Even at this early stage of caveolae exploration, additional functions were posited. In the 1970s Angela Dulhunty of Australian National University and Clara Franzini-Armstrong of the University of Pennsylvania, both then at the University of Rochester, carried out beautifully detailed measurements of the prevalence of caveolae in frog muscle tissue and showed that, as muscles are stretched, caveolae become less abundant.7 The researchers suggested that caveolae flatten out to give the inelastic plasma membrane some capacity to extend as muscle cells stretch, and pointed out that this buffering action could explain both the abundance of caveolae and their propensity to form interlinked clusters.

Enter molecular biology

The discovery of caveolin-1 in 1992 as a defining protein component of caveolae was a technical tour de force, and the same paper documented striations on the surface of caveolae suggestive of a protein coat. Caveolin-1 is a membrane protein with both N and C termini in the cytoplasm and stretches of hydrophobic amino acids embedded within the lipid bilayer. Caveolin-1’s discovery initially raised more questions than it answered, however.

Co-immunoprecipitation experiments, which use detergents to disrupt the membranes of cultured cells and antibodies immobilized on beads, have revealed associations between caveolin-1 and a large array of different plasma membrane receptors, leading to the idea that caveolae play a role in signal transduction. However, because caveolin-1 forms highly stable oligomers that resist extraction by commonly used detergents, co-immunoprecipitation experiments can yield false-positive interactions if solubilization is incomplete. Indeed, later research examining the structures of several signaling proteins thought to interact with caveolin-1 clearly showed that their putative caveolin binding domain is, in fact, unlikely to be accessible for interactions with caveolin, undermining the idea that caveolae act as hubs for signal transduction.8 Much more recently, the use of chemical cross-linkers has allowed for more-stringent detergents to be used to isolate caveolin’s binding partners without triggering the disassembly of protein complexes, and, tellingly, no signaling receptors were detected.9

Another confounding issue surrounding the initial attempts to determine binding partners was that the identification of caveolin-1 coincided with a rise in interest in plasma membrane structures known as lipid rafts. The fact that many membrane lipids and proteins, including caveolin-1, are tough to extract with detergents was interpreted as evidence that these components reside in the same region, or microdomain. This may or may not be the case, and the literature contains different views and findings addressing the issue. But it is now abundantly clear that caveolae do not contain enriched populations of many of the molecules thought to be present in lipid rafts, and are thus likely to be entirely distinct from rafts.

Genetic data surprise the field

Early hypotheses about caveolar function were challenged when, in 2001, the groups of Michael Lisanti, then at the Albert Einstein College of Medicine in New York, and Teymuras Kurzchalia at the Max Planck Institute of Molecular Cell Biology and Genetics in Germany independently reported that mice lacking caveolin-1 do not have caveolae but are apparently healthy.3,4 Despite showing some phenotypic anomalies, the animals had no major developmental defects, and ran around their cages eating, reproducing, and generally acting normally.

Genetic data show caveolae to be important for the normal function of blood vessels, muscle, and fat tissue.

There were already hints that mice without caveolae do in fact have an array of less obvious problems, which we now know to include vascular abnormalities, lipodystrophy, muscular dystrophy, and other fat-related metabolic dysfunctions. These phenotypes show that adipose tissue, endothelium, and muscles—the very tissues where caveolae are super-abundant—do not function correctly without caveolae. Nevertheless, given the near-normalcy of the CAV-1 knockout mice, one cannot escape the conclusion that either caveolae are not crucial components of cell signaling pathways and the machinery for regulating endothelial permeability, or that complex compensatory mechanisms exist to allow such key systems to function in the absence of caveolae. This conundrum can perhaps best be resolved from the bottom up. A detailed knowledge of the molecular mechanisms mediating the assembly of caveolae may lead to further details on functionally significant interaction partners, and thereby towards specific information on what caveolae actually do in cells.

Caveolae and membrane stretching

Over the past decade, a series of papers from several laboratories has transformed our picture of the structure of caveolae. There was already evidence for a protein coat around the bulb of caveolae, and we now know that the components of the coat are caveolins and cavins. (See illustration below.)

There are three caveolin isoforms in mammals. Caveolin-1 is crucial for forming caveolae in tissues other than striated muscle, while caveolin-3 has a similar importance in muscle. Caveolin-2 appears to be less essential for forming caveolae.

There are four cavins: cavin-1 is needed for forming caveolae in all tissues, cavin-2 and cavin-3 have variable abundance across different tissues, and cavin-4 is muscle-specific. The cavins contain extended regions that are likely to join together cavins of the same or different varieties into coiled oligomers. My colleagues and I examined the purified complex of cavins and caveolins by electron microscopy, revealing it to have the size and shape of the membrane bulb of caveolae.9 It is likely, therefore, that this caveolar coat complex (CCC) is what generates the distinctive shape of caveolae.Caveolae form when caveolin oligomerizes in the membrane before cavin proteins associate. Caveolae then fold inward and can form clusters.
See full infographic: WEB | PDF

The discovery of cavins and the CCC has generated tools to study the assembly and disassembly of caveolae, as the dissociation of key components from the CCC can be used as a proxy for changes in the functional state of caveolae. This approach forms the core of a 2011 landmark paper published by Christophe Lamaze and colleagues at the Institut Curie in Paris.10 In this study, the authors showed that caveolae disassemble or flatten out under increased membrane tension. These data support the idea first promulgated by Dulhunty and Franzini-Armstrong back in 1975 that caveolae can buffer mechanical tension within the plasma membrane, and thereby prevent membranes from breaking under stress forces. Lamaze and colleagues went on to show that cells with compromised caveolin function are more likely to rupture when stretched.

Since then, further studies have provided clear in vivo confirmation that at least partial disassembly of the CCC and flattening out of caveolar membranes is caused by increases in plasma membrane tension and that both muscle cells and endothelial cells under physiologically relevant stretch forces are more likely to suffer membrane rupture if they lack caveolae.

These and other experiments have generated a consensus within the field that caveolae protect cells from mechanical stresses that could otherwise lead to breaks in the plasma membrane. The precise mechanism of this protective effect is less clear. The simplest idea is that the folds in the plasma membrane introduced by caveolae flatten out when the cell needs to stretch, and this stops the membrane from breaking under tension. But other mechanisms for maintaining the integrity of the membrane are also possible. In 2013, for example, Matthias Corrotte and Norma Andrews of the University of Maryland, College Park used a pore-forming bacterial toxin to introduce small holes or breaks in the plasma membrane, and found that this induces local internalization of caveolae, potentially repairing membrane breaks by removing the damaged region of membrane from the cell surface.11 And there is still a large literature linking caveolae to signal transduction, so it is possible that caveolae protect cells from membrane damage indirectly, by inducing cellular responses such as cytoskeletal rearrangement or transcriptional changes. Once again, we are left with the hope that further molecular details will generate insights into these other proposed functions.

Neck vs. bulb

While the CCC around the bulb of caveolae is increasingly well characterized, the nature and precise function of proteins found around the constricted neck region is less well understood. There are three protein families that may be relevant: dynamins, pacsins, and EHDs.
In 1998, two papers that appeared back-to-back in the Journal of Cell Biology claimed that dynamin-2, involved in the budding of clathrin-coated vesicles, is similarly involved in the budding of caveolae.12,13 More than 20 years later, however, our knowledge of how dynamin might act in budding remains limited.CAVEOLAE DYNAMICS: Data from recent studies have led scientists to suspect that caveolae may have a role in buffering changes in cell membrane tension by changing their conformation. Indeed, membrane tension causes flattening of caveolae and loss of clusters of caveolae. It is probable, but not fully proven, that cavin and caveolin proteins can switch between flat and invaginated complexes on the membrane, but other proteins known as EHDs resist these processes.
See full infographic: WEB | PDF

Some publications have argued that caveolae are unlikely to bud from the membrane at all, while my group recently used genome editing to express GFP-tagged caveolin-1 at endogenous levels and showed that caveolae are internalized at a very slow rate.14,15 The basic problem here is that there is still no extracellular cargo known to be internalized specifically by caveolae. Unless and until such cargo is identified, it remains possible that internalization of caveolae occurs not so much to deliver material (such as signaling receptors, for example) from the cell surface to the cell interior, as to control the turnover or distribution of caveolae themselves.

EHD proteins and pacsins are also likely present at the neck of caveolae. EHD proteins are ATPases that, like dynamin, use energy from nucleotide hydrolysis to alter the curvature of the membrane to which they are bound. Only in the past year has it become clear that multiple members of the EHD protein family are recruited to the caveolae neck, where they appear to have two distinct but possibly related functions.16 In cells lacking the EHD proteins, caveolae are much less likely to form interlinked clusters. Also, although cells lacking EHDs clearly still have caveolae, the number of caveolae drops markedly when the cells are stretched. These observations suggest that EHDs help link caveolae together in higher-order arrays and allow caveolae to maintain their shape in the face of repeated membrane stretching.

The shape of caveolae is not fixed. Rather, they can undergo dynamic transitions between flat and invaginated states.

Less is known about how pacsins function in caveolae. They are not present in all of these structures, but the depletion of pacsins can cause a reduction in the number of caveolae. We speculate that the proteins at the neck control the dynamic distribution and potentially the reversible changes in membrane shape that are emerging properties of caveolae, while the basic membrane shape of the caveolar bulb is determined by the CCC.

Plenty of missing pieces

In the last few years our understanding of the parts list for caveolae has been transformed. Nevertheless, many unknowns remain, such as how cavins and caveolins fit together to make a stable, bulb-shape protein lattice. More high-resolution structural information on the CCC will be invaluable in answering this question. It will also be important, though difficult, to better understand the molecular underpinnings of the transitions in membrane shape that we infer caveolae to undergo. This will likely require establishing assays to determine the kinetics with which these morphological changes occur.

Further gaps in the molecular picture of caveolae include how cells regulate their polarized, nonrandom distribution, which suggests that they are likely linked to the cortical cytoskeleton. Why there are four different cavins and three different EHDs recruited to caveolae is also not at all clear, though it seems likely that this variety allows functional specialization of some kind.

Finally, as has been the case for the 65 years since their discovery, perhaps the biggest question pertains to the cellular function of caveolae. While we are increasingly certain that caveolae protect cells from mechanical damage, both how they do this, and what else they do, are still unresolved. More time, and more data, will be needed to unravel the mystery of caveolae.

Ben Nichols is a group leader at the Medical Research Council Laboratory of Molecular Biology in Cambridge, U.K. He uses genetic, imaging, and biochemical approaches to study caveolae and endocytosis.


  1. G.E. Palade, “Fine structure of blood capillaries,” J Appl Phys, 24:1424, 1953.
  2. K.G. Rothberg et al., “Caveolin, a protein component of caveolae membrane coats,” Cell, 68:673-82, 1992.
  3. B. Razani et al., “Caveolin-1 null mice are viable but show evidence of hyperproliferative and vascular abnormalities,” J Biol Chem, 276:38121-38, 2001.
  4. M. Drab et al., “Loss of caveolae, vascular dysfunction, and pulmonary defects in caveolin-1 gene-disrupted mice,” Science, 293:2449-52, 2001.
  5.  L. Liu et al., “Deletion of Cavin/PTRF causes global loss of caveolae, dyslipidemia, and glucose intolerance,” Cell Metab, 8:310-17, 2008.
  6. M.M. Hill et al., “PTRF-Cavin, a conserved cytoplasmic protein required for caveola formation and function,” Cell, 132:113-24, 2008.
  7. A.F. Dulhunty, C. Franzini-Armstrong, “The relative contributions of the folds and caveolae to the surface membrane of frog skeletal muscle fibres at different sarcomere lengths,” J Physiol, 250:513-39, 1975.
  8. B.M. Collins et al., “Structure-based reassessment of the caveolin signaling model: do caveolae regulate signaling through caveolin-protein interactions?” Dev Cell, 23:11-20, 2012.
  9. A. Ludwig et al., “Architecture of the caveolar coat complex,” J Cell Sci, 129:3077-83, 2016.
  10. B. Sinha et al., “Cells respond to mechanical stress by rapid disassembly of caveolae,” Cell, 144:402-13, 2011.
  11. M. Corrotte et al., “Caveolae internalization repairs wounded cells and muscle fibers,” eLife, 2:e00926, 2013.
  12. P. Oh et al., “Dynamin at the neck of caveolae mediates their budding to form transport vesicles by GTP-driven fission from the plasma membrane of endothelium,” J Cell Biol, 141:101-14, 1998.
  13. J.R. Henley et al., “Dynamin-mediated internalization of caveolae,” J Cell Biol, 141:85-99, 1998.
  14. P. Thomsen et al., “Caveolae are highly immobile plasma membrane microdomains, which are not involved in constitutive endocytic trafficking,” Mol Biol Cell, 13:238-50, 2002.
  15. E. Shvets et al., “Dynamic caveolae exclude bulk membrane proteins and are required for sorting of excess glycosphingolipids,” Nat Commun, 6:6867, 2015.
  16. I. Yeow et al., “EHD proteins cooperate to generate caveolar clusters and to maintain caveolae during repeated mechanical stress,” Curr Biol, 27:2951-62.e2955, 2017.

Interested in reading more?

Magaizne Cover

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?