Sites of Ca²⁺ wave initiation move with caveolae to the trailing edge of migrating cells

The original proposal that caveolae were involved in signal transduction was based on studies that suggested this membrane domain was well suited for compartmentalizing and processing information at the cell surface(Anderson, 1993;Anderson, 1998). One of the design features that seemed most attractive was that caveolae could carry signaling machinery to different locations in the cell, thereby contributing to the spatial organization of signal transduction. Regulating the location of a particular signaling event is critical for efficient signal processing. For example, mating pheromone receptors stimulate the polarization of Gβγ, Ste5 and Fus3 in yeast and disruption of polarity establishment genes such as FAR1 impair the ability of cells to respond to signals that transmit information through these signaling intermediates(Leeuw et al., 1995;Nern and Arkowitz, 2000;O'Shea and Herskowitz, 2000;Shimada et al., 2000). Another example is the loss of function mutations in the Drosophila cell polarity regulator genes scribble, discs large and lethal giant larvae. Mutations in these genes lead to proliferation of epithelial cells (Bilder et al., 2000;Bilder and Perrimon, 2000),indicating that cell cycle progression is linked to establishment of cell polarity.

A variety of studies now support the proposal that caveolae compartmentalize signal transduction(Anderson, 1998). Caveolae are enriched in a variety of signal transducing molecules; complex signaling events that use these molecules have been localized to this domain, and the caveolin family of lipid binding proteins have been identified as molecules involved in organizing the activity of various signaling molecules in caveolae(Smart et al., 1999). There remains, however, the question of whether or not caveolae can spatially organize signal transduction. Immunofluorescence localization of caveolin-1 in fibroblasts and endothelial cells show that this caveolae marker protein collects at the margins of cells and in patches that are aligned with actin-rich stress fibers (Isshiki et al.,1998; Rothberg et al.,1992). Electron microscopic thin section and enface(Rothberg et al., 1992) images of plasma membranes have confirmed that both invaginated and flat caveolae are densely packed in specific regions of the membrane, but devoid from other regions. GFP-tagged caveolin-1 has been used to monitor the distribution of caveolin-1 in response to different growth conditions and at various stages of the cell cycle. GFP-caveolin collects at sites of cell-cell contact in contact inhibited cells (Volonte et al.,1999) and concentrates at the cleavage furrow during cytokinesis(Kogo and Fujimoto, 2000). Assuming that GFP caveolin-1 is a reliable marker for caveolae, these studies suggest that the distribution of caveolae on the cell surface is regulated. It is not known if resident signaling molecules move with caveolae when they are induced to relocate.

A signaling event that has been localized to caveolin-rich regions of the endothelial cell surface is the ATP-stimulated initiation of Ca²⁺waves (Isshiki et al., 1998). Changes in intracellular Ca²⁺ concentration can be detected with the confocal microscope using the Ca²⁺ sensing dye Indo-1. The initial phase of Ca²⁺ release from ER was found to occur at the edge of cells in juxtaposition to caveolin-rich areas of membrane. A wave of Ca²⁺ release was then propagated from this site across the cell. Interestingly, initiation of Ca²⁺ waves occurred only at a subset of the caveolin-rich regions, suggesting that not all caveolae in a cell are equipped to stimulate Ca²⁺ release from ER. These results suggest that caveolae spatially organize ATP-induced wave initiation. If so, then conditions that cause signal competent caveolae to relocate on the cell surface should shift sites of Ca²⁺ wave initiation to these regions. Here we demonstrate that when endothelial cells are exposed to shear stress the caveolae accumulate in the region of the cell nearest the direction of media flow (designated the `upstream edge' of the cell). Polarization of caveolae is dependent on the amount of flow force and the time that it is applied. Once polarized, G_αq, the heterotrimeric G protein that links ATP receptors to ER Ca²⁺ release, and sites of Ca²⁺ wave initiation both co-localize with caveolae.

Hanks balanced salt solution (HBSS), ATP, vinculin mAb, and clathrinβ ₁- and β₂-adaptin mAb were from Sigma (St Louis, MO). Indo 1-AM was from Dojindo (Kumamoto, Japan). Medium-199 (M199)was from ICN Biomedicals (Tokyo, Japan). Fetal bovine serum (FBS) was from Gibco (Grand Island, NY). Texas Red-X phalloidin was from Molecular Probes,(Eugene, OR). Caveolin pAb and mAb (2234) was from Transduction Laboratories(Lexington, KY). G_αq/11 pAb (C-19) was from Santa Cruz Biochemicals (Santa Cruz, CA.). Cy5-goat anti-rabbit IgG, FITC-sheep anti-mouse IgG, and α-tubulin mAb were from Amersham (Buckinghamshire,UK). Alexa Fluor 488 goat anti-mouse IgG and Alexa Fluor 568 goat anti-rabbit IgG were from Molecular Probes (Eugene, OR).

Primary cultures of endothelial cells (ECs) were derived from the descending thoracic aorta of a bovine fetus by brief collagenase digestion of the intimal lining. Cells were grown in M199 supplemented with 20% FBS (v/v),2 mM L-glutamine, 50 IU/ml of penicillin and 50 μg/ml of streptomycin(designated, standard medium) and routinely passed by trypsinization in a 0.25% trypsin/1 mM EDTA solution before reaching confluence. Cumulative population doublings (CPD) of ECs were calculated using a Coulter Counter(Model ZM system, Coulter Electronics Ltd., Luton, UK). Sub-confluent ECs with a CPD under 40 were used in each experiment.

On day zero, 3.0×10⁵ endothelial cells were seeded onto a 60 mm culture dish containing a 25.8×45×0.2 mm glass coverslip(Matsunami, Osaka, Japan) and grown for 3 days in 5 ml of standard medium. The coverslip was removed from the dish and placed in a flow-loading chamber(Yasuhisa Biomechanics, Tokyo, Japan) for each experiment. A parallel-plate-type flow chamber was used to load a laminar flow on ECs as previously described (Ando et al.,1993). Briefly, one side of the chamber was the coverslip(25.8×45×0.2 mm) on which the cells were cultured. The other side was a polymethacrylate plate. These two flat surfaces were held approximately 200 μm apart by a silicone rubber gasket. Culture medium maintained at 37°C and 5% CO₂ was re-circulated through the chamber using a silicone tube connected to a reservoir by a roller/tube pump (ATTO Co., Tokyo,Japan). The amount of shear stress τ (dyn/cm²) applied to the cells was calculated using the formula τ =6μQ/a²b, whereμ is the viscosity of the perfusate (0.0094 poise at 37°C), Q is the volume flow (ml/s), and `a' (0.02 cm) and `b' (1.6 cm) are the cross-sectional dimensions of the flow path.

On day zero, 3.0×10⁵ endothelial cells were seeded onto a 60 mm culture dish containing a 25.8×45×0.2 mm coverslip and grown for 3 days in 5 ml of standard medium. One half of the cell monolayer was sharply scraped with a sterilized cell scraper (Corning Inc, Corning, NY) and the scraped cells were removed by washing the culture dish with fresh medium. The cells were then allowed to grow for the indicated time and processed for indirect immunofluorescence.

Loading of cells with the Ca²⁺ indicator dye, image acquisition and image processing was carried out as previously described(Isshiki et al., 1998).

After the application of shear stress and the recording of[Ca²⁺] images, the cells were washed three times with PBS, fixed for 30 minutes at room temperature with 3% (w/v) paraformaldehyde in PBS and processed for fluorescence localization of the indicated proteins as previously described (Isshiki et al.,1998). To co-localize caveolin-1 with other cellular proteins,fixed cells were incubated in the presence of a 1/120 dilution of caveolin-1 pAb plus either 4 U/ml Texas Red-X phalloidin (actin), a 1/80 dilution of vinculin mAb, a 1/50 dilution of α tubulin mAb or a 1/50 dilution of AP-1/2 mAb. For co-localization of G_αq/11, cells were stained with mAb caveolin-1 (1/20 dilution) and pAb G_αq/11 (1/50 dilution) and visualized with Alexa Fluor 488 goat anti-mouse IgG and Alexa Fluor 568 goat anti-rabbit IgG, respectively.

After the cells were exposed to the indicated amount of shear-stress, the coverslip was removed from the chamber, washed with PBS, fixed with 3%paraformaldehyde in PBS and processed for indirect immunofluorescence localization of caveolin-1. Images of 300-500 cells were recorded and each cell was divided into eight equal sectors (45° each) using the nucleus as the pivot point for sectioning (Fig. 4). Each sector was designated as belonging to either region A, B,C, D or E as shown in Fig. 4. Each cell was then scored and assigned to one of these regions according to where the heaviest caveolin staining occurred on the cell margins. Cells that exhibited caveolin-1 staining in more than one region were not classified. Approximately 10-15% of the cells were not classified.

Cells that had been subjected to shear stress as described above were washed with ice-cold PBS and solubilized in 500 μl RIPA buffer (1% Nonidet P-40, 20 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 0.5% sodium deoxycholate, 2 mM EDTA, 2 mM EGTA, 0.1% SDS, 0.2 mM Na₂MoO₄, 10 mM NaF, 1 mM Na₃VO₄, 1 mM phenylmethylsulfonyl fluoride, 5μg/ml leupeptin, 5 μg/ml antipain, 5 μg/ml pepstatin A, 0.2 unit/ml aprotinin). Lysates were centrifuged at 26,000 g for 30 minutes. The supernatants were mixed with SDS sample buffer and separated using a 15% PAGE. Gels were transferred to Immobilon^TMpolyvinylidene difluoride membranes (Millipore, Bedford, MA). Membranes were blocked in TBS (20 mM Tris-HCL, 137 mM NaCl, pH 7.6) plus 5% nonfat milk and 0.1% Tween-20, and then incubated for 1 hour in the presence of caveolin pAb diluted in blocking solution. Membranes were washed in PBS and incubated with anti-rabbit IgG horseradish peroxidase-conjugated antibody. The blots were developed using Enhanced Chemiluminescence kit (Amersham, Piscateway, NJ) and analyzed by a GS363 Molecular Imager System (Bio-Rad, Hercules, CA).

PT/PCR was performed to quantify the mRNA levels of caveolin as previously described (Ando et al., 1994). The cDNA samples were co-amplified by PCR with primer pairs for caveolin-1 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primer pairs for caveolin-1 were 5′-CAACAAGGCTATGGCAGAGG-3′ and 5′-CGTAGATGGAATAGACACGGC-3′, and for GAPDH were 5′-GGAAGCTCGTCATCAATGG-3′ and 5′-AGGAGGCATTGCTGACAATC-3′. 10 μl of amplified product was sampled every other cycle, and separated on a 5% polyacrylamide gel. To quantify the PCR products, the radioactivity of each band was measured with a GS363 Molecular Imager, and plotted against the number of PCR cycles on a semi-logarithmic scale. From the sigmoid curve, the cycle in which the operating range of the PCR was linear was selected and the ratio of radioactivity caveolin to GAPDH in that cycle was designated as the relative amount of caveolin mRNA.

Endothelial cell cultures were prepared as described above except that a 10.5×22 mm Thermanox coverslip (Electron Microscopy Sciences, Ft Washington, PA) was substituted for the glass coverslip. Cells were exposed to 20 dynes/cm² for 24 hours in a flow chamber customized for the smaller coverslip. The coverslip was removed and marked to indicate the orientation of laminar flow. The cells were fixed with 2% glutaraldehyde in 0.1 M phosphate buffer pH 7.6 for 30 minutes at room temperature followed by 1% OsO₄ in 0.1 M phosphate buffer for 1 hour at room temperature. Multiple coverslips with the same orientation were stacked together and embedded in Epon 812. The embedded coverslip stacks were cut out of the Epon,marked and then oriented in flat embedding molds so that the coverslips were parallel with the bottom of the mold. The molds were infiltrated with Epon and polymerized. These blocks were removed from the mold, positioned so that sections would be cut perpendicular to the coverslip and oriented with respect to the direction of laminar flow. These sections were mounted on slot grids and viewed with a JEOL 1200 electron microscope.

We were interested in whether the distribution of caveolae changed when endothelial cells were induced to migrate on a substratum. To induce migration, primary bovine endothelial cells were grown to confluence on coverslips and the cells on one half of the coverslip were removed by scraping(Fig. 1, below the yellow line). The coverslip was put back in the media and cultured for 0, 4 or 24 hours before processing each sample for indirect immunofluorescence co-staining of caveolin-1 (left panel) and actin (right). The cells that remained after scraping the coverslip (0 hr, above line) displayed a typical clustered pattern of caveolin-1 staining but there was no polarity to the location of these caveolin-1 patches. Actin staining showed numerous stress fibers. Some cells began to migrate across the scrape boundary (yellow line)by 4 hours and in these cells the heaviest caveolin-1 staining was on the trailing edge of the cell (yellow arrows). The repositioning of the caveolin-1 patches to the trailing edge was even more pronounced in cells that had been migrating for 24 hours. Fewer actin stress fibers were evident in migrating cells (Fig. 1, compare 4 hr and 24 hr with 0 hr), but the distribution of filaments was unchanged.

Endothelial cells in situ are continually exposed to laminar shear stress,which is a mechanical activity that has been found to stimulate multiple signaling pathways (Davies,1995; Traub and Berk,1998). We used immunofluorescence to determine whether shear stress, like cell migration, also caused the rearrangement of caveolin-1 patches on the cell surface (Fig. 2). Cells grown on coverslips were either examined directly(unstressed) or placed in a parallel-plate flow chamber and subjected to a laminar shear stress of 20 dynes/μm² for 24 hours(Fig. 2, stressed, arrows indicate direction of flow). The cells were fixed and processed for indirect immunofluorescence co-localization of caveolin-1, actin and vinculin. The normal, patchy distribution of caveolin-1 was evident in unstressed cells, and these patches could be found in all regions of the cell(Fig. 2A). In stressed cells,by contrast, virtually all of the caveolin-1 staining was confined to the upstream edge of the cell (Fig. 2E, white arrows). Some cells had long extensions that protruded from the upstream portion of the cell in the direction of fluid flow(Fig. 2, yellow asterisk). The entire length of these extensions was stained heavily with the caveolin-1 pAb. The stressed cells appeared to have an increased number of actin stress fibers(compare Fig. 2B with F). The number of vinculin positive spots was not changed (compareFig. 2C with G) in shear stressed cells.

In separate experiments (Fig. 3), we compared the distribution of caveolin-1(Fig. 3A,E) with either actin(Fig. 3B), tubulin(Fig. 3C) or clathrin AP-1/2(Fig. 3F) in shear stressed cells. Whereas the caveolin-1 patches redistributed to the upstream edge of cells (white arrows), there was no change in the distribution of either AP-1/2 or tubulin. The number of stress fibers was increased and in many cells they tended to be oriented parallel to the direction of fluid flow. The polarization of caveolin-1 appeared to occur before stress fiber reorganization (data not shown).

The redistribution of the caveolin-1 patches was dependent on both the strength of the shear stress and the time of exposure to stress(Fig. 4). Cells were either incubated in the presence of an increasing shear stress force for 44 hours(Fig. 4, right panel) or exposed to a constant force (20 dynes/cm²) for increasing times up to 68 hours (Fig. 4, left panel). The number of cells that had the majority of the caveolin-1 staining in either region A, B, C, D, or E was tabulated(Fig. 4, top-left). At the beginning of the experiment, equal numbers of cells were found with caveolin-1 patches in the five regions. With either increasing time or increasing shear stress force, however, the number of cells with caveolin-1 patches in region A increased. The increase in the number of region A cells was matched by a corresponding decline in cells with caveolin-1 in regions C, D or E. Little change was seen in the number of cells with caveolin-1 patches in region B. After 68 hours of incubation, >80% of the cells were found to have caveolin-1 patches in region A. As little as 1.5 dynes/cm² caused an increase in the number of cells with caveolin-1 patches in region A, but it took 20 dynes/cm² to cause the number of cells with patches in this region to reach a maximum (∼58% of the cells).

A change in the distribution of caveolin-1 patches suggests that these conditions cause a redistribution of caveolae. Numerous studies have shown that caveolin-1 is a marker for caveolae in endothelial cells(Schnitzer et al., 1995). Nevertheless, we used thin section electron microscopy to confirm that the relocation of caveolae accounted for the redistribution of caveolin-1 staining. Cells were grown on plastic coverslips and subjected to 20 dynes/cm² of stress for 24 hours. The coverslip was fixed and positioned during embedding so that individual cells oriented in the direction of media flow could be viewed throughout their entire length.Fig. 5A shows at low magnification a longitudinal view of a typical cell with the upstream region of the cell towards the right (arrow indicates direction of flow). A higher magnification of this region of the cell shows an anastomosing network of flask-shaped, tubular and vesicular caveolae(Fig. 5B, arrows). Many of the caveolae vesicles were tightly associated with smooth ER(Fig. 5B, asterisk). Progressing downstream of this region, the number of recognizable caveolae declined. Very few caveolae could be identified in the body of the cell or at its downstream edge (data not shown). Therefore, caveolin-1 staining at the trailing edge of shear-stressed cells corresponds to sites of massive caveolae accumulation. Since 20 dynes/cm² of shear stress for up to 24 hours did not change the level of either caveolin-1 protein(Fig. 6A) or mRNA(Fig. 6B), the caveolae that have accumulated at the upstream region probably came from other locations in the cell.

Previously we have shown that agonists such as ATP and bradykinin stimulate Ca²⁺ wave initiation in caveolin-rich regions of the cell(Isshiki et al., 1998). These results suggest that caveolae contain machinery that links G-protein-coupled receptors to the release of Ca²⁺ from ER stores. A key molecular component of this machinery is the heterotrimeric G protein subunit G_αq G_αq has been localized to caveolae(Oh and Schnitzer, 2001), so we used the same antibody to determine the distribution of G_αq in shear-stressed cells(Fig. 7). As reported previously, we found good colocalization(Fig. 7). of caveolin-1(Fig. 7A) and G_αq (Fig. 7B)in unstressed cells, although not all of the caveolin-rich areas were positive for G_αq. The application of 20 dynes/cm² of fluid shear to the cells in the direction of the arrow caused the polarization of caveolin-1 to the upstream edge of the cell(Fig. 7D). Large portions of these caveolin-1 patches were also positive for G_αq(Fig. 7E,F). Importantly, not all of the caveolin-1 patches were positive for G_αq,suggesting that not all caveolae are equipped with this signaling molecule.

We next examined whether or not Ca²⁺ wave initiation migrated with caveolae to the upstream edge of shear stressed cells(Fig. 8). Unstressed cells(A-D) were analyzed first. Endothelial cells were grown on coverslips for 2 days before being loaded with the Ca²⁺ indicator dye Indo-1 AM. Ca²⁺ waves were initiated by stimulating cells with 0.5 μM ATP.Fig. 8D shows images of cells taken at 0.38 second intervals after the addition of ATP. Within 10 seconds after ATP stimulation, focal sites of Ca²⁺ release were seen(Fig. 8, white arrows). Ca²⁺ release spread from these sites throughout the cell. Following the Ca²⁺ imaging phase of the experiment, the cells were fixed and processed for immunofluorescence detection of caveolin-1(Fig. 8A-C). Each site of Ca²⁺ wave initiation corresponded to a caveolin-1 patch(Fig. 8A,D, compare arrows). Not all caveolin-rich regions, however, were associated with sites of Ca²⁺ release.

Sites of Ca²⁺ wave initiation were next recorded in cells that had been exposed to a laminar shear stress of 20 dynes/cm² for 24 hours (Fig. 8E-H). Cells were loaded with Indo-1 AM and stimulated with 2 μM ATP before taking images of the cells at 0.38 second intervals (H). Ca²⁺ wave initiation occurred exclusively at the upstream edge of cells (H, arrow). Importantly,the initial site coincided with a cell extension that was oriented upstream of the cell. With time, Ca²⁺ release spread progressively towards the downstream end of the cell. Staining of cells with caveolin-1 pAb showed that the upstream edges and the extension were covered with caveolin-1. We conclude that sites of Ca²⁺ wave initiation follow the caveolae as they become polarized on the cell surface in response to shear stress.

In addition to repositioning sites of Ca²⁺ wave initiation,shear stress also caused a marked desensitization to ATP(Fig. 9). Sequential addition of increasing concentrations of ATP to either stressed (bottom panel) or unstressed (upper panel) cells showed that it took a tenfold higher concentration of ATP to initiate a Ca²⁺ wave in stressed cells than it did in unstressed cells.

We have identified two different conditions that cause the polarization of caveolae on the cell surface. Polarization was found to occur after either wounding a cell monolayer or exposing cells to laminar shear stress. In as much as both conditions stimulate cells to migrate(Ando et al., 1987), cell migration is most probably the underlying cause of caveolae redistribution in these cells. Caveolae, therefore, appear to accumulate at the trailing edge of migrating cells and concentrate over retraction fibers in this region. Clathrincoated pits did not become polarized under these conditions nor did we see much change in the distribution of vinculin-rich focal adhesion sites or the organization of the actin cytoskeleton. We do not believe these conditions stimulate the formation of new caveolae because we did not see an increase in the amount of caveolin-1 mRNA or protein and qualitatively the treated and untreated cells appeared to have similar levels of caveolin-1 staining.

Our observations stand in contrast to a recent report that growth factor-stimulated chemotaxis causes the relocation of raft markers to the leading edge of MCF-7 adenocarcinoma cells(Manes et al., 1999). This study did not determine if caveolin-1 moved to the leading edge, so it is not clear whether they were observing the behavior of caveolae or non-caveolae rafts. The possibility remains, therefore, that caveolae move to the trailing end of a migrating cell when it is moving in response to a mechanical stimulus and to the leading end when migrating in response to a chemical stimulus. Alternatively, the polarization of caveolae during cell migration is cell-type specific.

Several hours were required to achieve complete polarization of caveolae in migrating endothelial cells. However, Kogo and Fujimoto saw migration of GFP-caveolin-1 to the cleavage furrow within minutes after cells enter cytokinesis (Kogo and Fujimoto,2000). GFP-caveolin-1 also accumulates at sites of cell-cell contact during contact inhibited cell growth(Volonte et al., 1999). Assuming that GFP-caveolin-1 is a reliable marker for caveolae in these cells,then caveolae must be mobile and be able to assume rapidly different spatial arrangements in response to specific stimuli. The finding that a number of different conditions cause caveolae relocation suggests that this is a general mechanism cells use for spatially organizing specific activities at the cell surface.

An important future area of investigation is to determine the mechanism of caveolae polarization. One possibility is that recycling caveolae vesicles collect at specific regions of the cell in response to the different stimuli. Another is that caveolae behave like actin patches in yeast cellsWaddle et al., 1996) and move in the plane of the membrane to the trailing edge of the cell. Finally,caveolae might not move at all but accumulate at the end of a migrating cell as the result of differential movement of other organelles. A number of standard inhibitors such as staurosporin, herbimycin A, C3 toxin and the MEK-1/2 inhibitor PD98059 had no affect on caveolae polarization (data not shown).

Previous studies have shown that Ca²⁺ wave formation in the non-migrating endothelial cell occurs at caveolin-rich regions of the cell surface (Isshiki et al.,1998). Importantly, not all caveolin-rich regions in the same cell are competent to initiate waves. Tests to probe the functionality of these regions in individual cells showed that wave formation occurred at the same caveolin-1 patch in cells repeatedly exposed to the same stimuli or sequentially exposed to different stimuli. Therefore, only subsets of caveolae appear to contain the signaling machinery that regulates Ca²⁺release from the ER. This is in agreement with our observation that not all caveolin-1 patches were positive for G_αq. The simultaneous relocation of G_αq, sites of Ca²⁺ wave formation and caveolae to the trailing edge of migrating cells is strong evidence that caveolae can function as containers and carry signaling machinery to specific locations in the cell.

We cannot rule out the possibility that Ca²⁺ waves originate from non-caveolae membrane that has accumulated at the trailing edge of migrating cells. Nevertheless, we found that numerous invaginated caveolae and caveolae vesicles at the trailing edge were in close apposition to elements of smooth ER (Fig. 5, asterisk). These images give the impression that caveolae specifically interact with ER. Caveolae can deliver molecules to the ER(Benlimame et al., 1998), so it is possible that caveolae/ER contact sites are locations where Ca²⁺release is stimulated (Isshiki and Anderson, 1999).

Genetic studies of budding and fission yeast have identified a variety of molecules that control cell polarity in these organisms(Chant, 1999). Interestingly,the same molecules seem to be important for both budding and fission. Four of the critical regulatory molecules are the MAP kinase Mpk1, Rho1, PKC1 and phosphoinositides. Functional homologues of these molecules, such as ERK1 and ERK2) (Liu et al., 1997), Rho A (Michaely et al., 1999),multiple PKC isoforms (Mineo et al.,1998) and phosphoinositides(Pike and Casey, 1996), are concentrated in caveolae. Caveolae also contain G-proteincoupled receptors(Feron et al., 1997), are enriched in ERM proteins (Michaely et al.,1999) that recruit Rho A(Tsukita et al., 1994), and contain various receptor and non-receptor tyrosine kinases(Ko et al., 1998;Liu et al., 1996). If migrating caveolae can carry these signaling molecules to different locations in the cell, then they should be found wherever caveolae accumulate during cell migration and cytokinesis. Immunofluorescence has shown that ERM proteins collect both at the trailing edge of migrating cells(Serrador et al., 1997) and in the cleavage furrow during cytokinesis(Kosako et al., 1999). Transfected Rho A is found concentrated at the cleavage furrow and sites of cell-cell contact (Kosako et al.,1999). Endogenous urho1, the sea urchin Rho homologue, has been localized to the cleavage furrow of dividing eggs(Nishimura et al., 1998). Finally, PKC isoforms translocate to the uropod of lymphocytes in response to whole body hyperthermia (Wang et al.,1999). If caveolae are the vehicle for moving signaling molecules to these locations, then they may play a crucial role in establishing cell polarity.

We thank Brenda Pallares for her excellent administrative assistance. This work was supported by grants from the National Institutes of Health, HL 20948,GM 52016, the Perot Family Foundation, the Japanese Heart Foundation Grant for Research on Hypertension and Vascular Metabolism, and the Banyu Fellowship Awards in Cardiovascular Medicine, which is sponsored by Banyu Pharmaceutical Co., Ltd. and The Merck Company Foundation.