Differential Targeting of β-Adrenergic Receptor Subtypes and Adenylyl Cyclase to Cardiomyocyte Caveolae

Differential modes for β1- and β2-adrenergic receptor (AR) regulation of adenylyl cyclase in cardiomyocytes is most consistent with spatial regulation in microdomains of the plasma membrane. This study examines whether caveolae represent specialized subdomains that concentrate and organize these moieties in cardiomyocytes. Caveolae from quiescent rat ventricular cardiomyocytes are highly enriched in β2-ARs, Gαi, protein kinase A RIIα subunits, caveolin-3, and flotillins (caveolin functional homologues); β1-ARs, m2-muscarinic cholinergic receptors, Gαs, and cardiac types V/VI adenylyl cyclase distribute between caveolae and other cell fractions, whereas protein kinase A RIα subunits, G protein-coupled receptor kinase-2, and clathrin are largely excluded from caveolae. Cell surface β2-ARs localize to caveolae in cardiomyocytes and cardiac fibroblasts (with markedly different β2-AR expression levels), indicating that the fidelity of β2-AR targeting to caveolae is maintained over a physiologic range of β2-AR expression. In cardiomyocytes, agonist stimulation leads to a marked decline in the abundance of β2-ARs (but not β1-ARs) in caveolae. Other studies show co-immunoprecipitation of cardiomyocytes adenylyl cyclase V/VI and caveolin-3, suggesting their in vivo association. However, caveolin is not required for adenylyl cyclase targeting to low density membranes, since adenylyl cyclase targets to low buoyant density membrane fractions of HEK cells that lack prototypical caveolins. Nevertheless, cholesterol depletion with cyclodextrin augments agonist-stimulated cAMP accumulation, indicating that caveolae function as negative regulators of cAMP accumulation. The inhibitory interaction between caveolae and the cAMP signaling pathway as well as domain-specific differences in the stoichiometry of individual elements in the β-AR signaling cascade represent important modifiers of cAMP-dependent signaling in the heart.

Catecholamines act through cardiac ␤-adrenergic receptors (␤-ARs) 1 to influence the contractile state of the heart. The direct inotropic and chronotropic support provided by cardiac ␤-ARs represents a critical compensatory mechanism to preserve cardiac function during stress and/or states associated with circulatory compromise. In the hearts of most mammalian species, the physiologic effects of catecholamines are mediated by the predominant ␤ 1 -AR subtype (75-80% of the total ␤-ARs), which activates a signaling pathway involving the G sdependent stimulation of adenylyl cyclase leading to the accumulation of cAMP and protein kinase A-dependent phosphorylation of key target proteins. Cardiomyocytes also express ␤ 2 -ARs that support contractile function. Until quite recently, most studies of ␤ 2 -AR signaling in cardiomyocytes were wedded to the concept that ␤ 2 -ARs signal to the G s /cAMP pathway in a manner that is essentially equivalent to the pathway activated by ␤ 1 -ARs. However, there is evidence that ␤ 2 -ARs are not functionally redundant, including the findings that ␤ 2 -ARs couple to the G s /cAMP pathway more efficiently than ␤ 1 -ARs and that the pathway for ␤ 1 -AR activation of adenylyl cyclase is susceptible to inhibitory modulation by m 2 -muscarinic cholinergic receptors (m 2 -mAChRs), whereas the pathway for ␤ 2 -AR activation of adenylyl cyclase is not (1,2). This suggests that individual ␤-AR subtypes might fulfill distinct roles in transmembrane signaling due to spatially or developmentally regulated patterns of expression (3), but a specific molecular mechanism to adequately explain all of the distinct signaling properties of individual ␤-AR subtypes has not been identified.
Until quite recently, the prevailing concept was that components of the ␤-AR complex are freely mobile in the plasma membrane and that the specificity of molecular interactions is dictated entirely by information encoded in the three-dimensional structures and recognition surfaces of individual moieties. However, a simple "random collision coupling" model is inadequate to explain all of the experimental data, including the aforementioned evidence that ␤ 1 -and ␤ 2 -ARs display distinct susceptibility to inhibitory modulation by the m 2 -mAChR (2)(3)(4). One potential molecular mechanism that could impart this type of specificity to ␤-AR subtype signaling is compartmentalization to membrane subdomains such as caveolae. Caveolae were first identified as flask-shaped uncoated invaginations on the surface of highly differentiated cells. Caveolae are now recognized to be plasma membrane compartments with distinct lipid and protein composition that sequester and regulate the function of cytoplasmically oriented signal transduction molecules (5). In particular, there is evidence that ␤ 2 -ARs and m 2 -mAChRs (6 -10), their associated G protein ␣ and ␤ subunits (11)(12)(13)(14), certain adenylyl cyclase isoforms (15)(16)(17)(18), one member of the G protein-coupled receptor kinase family (GRK2, which phosphorylates agonist-activated ␤-ARs (19)), and the catalytic subunit of protein kinase A (PKA (20)) accumulate in caveolae at steady state and/or following ligandinduced activation. Localization of these diverse signaling molecules to caveolae suggests that this structure can serve as a scaffold to preassemble membrane-bound oligomeric complexes and thereby facilitate efficient and rapid coupling of agonistoccupied receptors to effectors. Such a mechanism could be particularly pertinent for sympathetic regulation of cardiomyocyte function, where rapid and productive signaling from receptor-activated G protein subunits to the adenylyl cyclase enzyme could be restricted by low (potentially limiting) levels of the adenylyl cyclase enzyme. Caveolae also may act to dampen signaling as a result of the properties of their principle structural protein, caveolin. The mammalian caveolin gene family consists of caveolin-1, caveolin-2, and the muscle-specific caveolin-3 (5). Domain-mapping studies identify a cytosolic membrane-proximal region (designated the "caveolin scaffolding domain") in caveolin-1 (as well as the structurally homologous caveolin-3) that interacts with putative caveolin-binding motifs in a wide range of signaling molecules (including G protein ␣ subunits and the catalytic domains of certain adenylyl cyclase isoforms, GRK2, and the catalytic subunit of PKA (5, 19 -21)). In vitro studies suggest that caveolin negatively regulates the activation state of heterotrimeric G proteins and functions as a "general kinase inhibitor" for many signaling enzymes. Accordingly, the goal of the present study was to determine whether caveolae form a signaling module for components of the ␤-AR complex in cardiomyocytes.
Cell Culture-Cardiac myocytes were isolated from hearts of 2-dayold Wistar rats by a trypsin dispersion procedure according to a protocol that incorporates a differential attachment procedure to enrich for cardiac myocytes as described previously (22). The yield of myocytes typically is 2.5-3 ϫ 10 6 cells/neonatal heart. Cells were plated at a density of 0.5 ϫ 10 6 cells/ml on protamine sulfate-coated 100-mm culture dishes. Although the preplating step effectively decreases fibroblast contamination, a small number of cells with proliferative capability such as cardiac fibroblasts persist in myocardial cell cultures. Proliferation of these cells was further curtailed with an irradiation protocol (22). Experiments were performed following 5-6 days of culture in minimal essential medium (Life Technologies, Inc.) with 10% fetal calf serum, 5 ϫ 10 Ϫ6 M hypoxanthine, and 12 mM NaHCO 3 . HEK293 cells were obtained from Dr. Jonathan Javitch and were propagated in Dulbecco's modified Eagle's medium supplemented with Geneticin and 10% fetal bovine serum. Cardiac fibroblast cultures were obtained from cells adherent to culture dishes during preplating and cultured according to standard methods (23).
Purification of Caveolin-rich Membrane Fractions-Fractions enriched in the muscle-specific caveolin-3 isoform were prepared by two methods. Most of the isolations were performed according to the detergent-free purification scheme of Song et al. (24) essentially as described previously. All steps were carried out at 4°C. Briefly, cells from five 100-mm diameter dishes were washed twice with ice-cold phosphatebuffered saline and then scraped into 0.5 M sodium carbonate, pH 11.0 (0.5 ml/dish; total volume ϳ2.5 ml for each preparation). To disrupt cellular membranes, homogenization was carried out sequentially with a loose fitting Dounce homogenizer (10 strokes), a Polytron tissue grinder (three 10-s bursts), and a tip sonicator (three 20-s bursts). The homogenate was then adjusted to 40% sucrose by adding an equal volume of 80% sucrose prepared in Mes-buffered saline (25 mM Mes, pH 6.5, 0.15 M NaCl), placed on the bottom of an ultracentrifuge tube, overlaid with a 5-35% discontinuous sucrose gradient (4 ml of 5% sucrose, 3 ml of 35% sucrose; both in Mes-buffered saline containing 250 mM sodium carbonate), and centrifuged at 38,000 rpm for 16 -18 h in an SW40 rotor (Beckman). After centrifugation, 12 1-ml gradient fractions were concentrated by precipitation with trichloroacetic acid as follows. Fractions were mixed with 7.92% trichloroacetic acid (1:10; v/v) and incubated for 30 min on ice. Precipitated proteins were pelleted by centrifugation at 3700 ϫ rpm for 15 min at 4°C (IEC; Centra-8R centrifuge). The pellet was washed once with 5 ml of ethyl ether and dissolved in SDS-PAGE sample buffer (generally without boiling, to prevent aberrant migration during immunodetection of receptors and adenylyl cyclase). In some experiments, the entire 40% lower sucrose layer (fractions 8 -12) was pooled for immunoblot analysis.
Some studies also were performed with caveolin-3-enriched fractions prepared according to the method of Smart et al. (13). In this case, plasma membranes were isolated by scraping cells from 10 100-mm plates into 30 ml of ice-cold buffer A (0.25 M sucrose, 1 mM EDTA, 20 mM Tricine, pH 7.8) followed by centrifugation at 1,000 ϫ g for 5 min. Cells were then resuspended in 1 ml of buffer A, placed in a 2-ml Potter-Elvehjem tissue grinder (Wheaton; catalog no. 358003), and homogenized (20 strokes). The suspension was centrifuged at 1000 ϫ g for 10 min to yield a postnuclear supernatant fraction, which was removed and stored on ice. The nuclear pellet was resuspended in 1 ml of buffer A, homogenized, and centrifuged at 1000 ϫ g for 10 min. The supernatant was combined with the previous postnuclear supernatant, layered on top of 23 ml of 30% Percoll in buffer A, and centrifuged at 65,000 ϫ g for 30 min (Beckman Ti60 rotor). The plasma membrane fraction, a visible band ϳ5.7 cm from the bottom of the centrifuge tube, was collected, adjusted to 2 ml with buffer A, and sonicated six times. Intracellular membranes (IM) were obtained by centrifuging fractions below the plasma membrane at 360,000 ϫ g for 1 h. The IM fraction pellet was mixed with 4ϫ SDS-PAGE sample buffer (3:1, v/v). A 0.2 ml aliquot of plasma membrane sonicate was diluted 60 times with buffer B (1 mM EDTA, 20 mM Tricine, pH 7.8, 150 mM NaCl), centrifuged at 360,000 ϫ g for 1 h, and then dissolved in SDS-PAGE sample buffer. The remaining plasma membrane sonicate (1.8 ml) was mixed with 2.39 ml of 50% OptiPrep prepared in buffer C (0.25 M sucrose, 6 mM EDTA, 120 mM Tricine, pH 7.8) plus 1.01 ml of buffer A to make a 23% OptiPrep solution. This was placed on the bottom of an ultracentrifuge tube, a linear 20 to 10% OptiPrep gradient (prepared by diluting 50% OptiPrep in buffer C with buffer A) was layered on top, and the sample was centrifuged at 57,600 ϫ g for 90 min in an SW40 rotor. The bottom fractions were collected and designated noncaveolae membrane (NCM). The top 6.5 ml of the gradient was collected and mixed with 5.38 ml of 50% OptiPrep in buffer C plus 0.12 ml of buffer A to make an ϳ30% OptiPrep solution. This was overlaid with 0.5 ml of 15% OptiPrep and then with 0.5 ml of 5% OptiPrep (both in buffer A) and centrifuged at 57,600 ϫ g for 90 min. An opaque band located at the 5/15% OptiPrep interface was collected and designated caveolae fraction. Caveolae fraction was diluted 10 -13 times with buffer B followed by centrifugation at 360,000 ϫ g for 1 h. The resulting caveolae (yield typically 12 g of protein) was dissolved in SDS-PAGE sample buffer. NCM was diluted four times with buffer B followed by centrifugation at 360,000 g for 1 h and solubilization in SDS-PAGE sample buffer.
Electrophoresis and Immunoblotting-Samples were separated by SDS-PAGE (10% acrylamide) and transferred to nitrocellulose, which, in general, was cut into longitudinal strips for incubation with various primary antibodies. Sample boiling was avoided to prevent aberrant migration during immunodetection of receptors and adenylyl cyclase. Immunoblotting was performed with antibodies against the ␤ 1 -AR, ␤ 2 -AR, G␣ s , type V/VI adenylyl cyclase, GRK2, RI␣, RII␣, ␣-catalytic subunit of protein kinase A, clathrin, flotillin, and ESA (diluted in 50 mM Tris, pH 7.5, 0.2 M NaCl containing 5% non-fat dry milk, 0.1% Tween 20, and 0.02% NaN 3 ); anti-caveolin-1 and caveolin-3 antibodies (diluted in 50 mM Tris, pH 7.5, 0.2 M NaCl containing 1% nonfat dry milk, 0.5% Tween 20, and 0.02% NaN 3 ); anti-G␣ i1/2 , -G␣ o/i3 , and -G protein ␤ 35/36 subunit antibodies (diluted in 50 mM Tris, pH 7.5, 0.2 M NaCl containing 5% bovine serum albumin, 0.05% Tween 20, and 0.02% NaN 3 ); and anti-m 2 -mAChR-antibodies (diluted in 50 mM Tris, pH 7.5, 0.2 M NaCl containing 1% nonfat dry milk, 3% bovine serum albumin, and 0.02% NaN 3 ). Primary antibodies were used at final dilutions of 1:5000 (ESA), 1:1000 (caveolin-3, G␣ i1/2 , G␣ i3/o ), 1:500 (G␣ s , GRK2, G protein ␤ 35/36 subunit), 1:250 (RI␣, RII␣, ␣-catalytic subunit of protein kinase A, flotillin, m 2 -mAChR), 1:200 (clathrin), or 1:100 (␤ 1 -AR, ␤ 2 -AR, type V/VI adenylyl cyclase), and bound primary antibodies were visualized with enhanced chemiluminescence according to the manufacturer's instructions. Each of the antibodies was initially screened with total cell lysates to ensure that it reacted with a band (or bands) of the appropriate molecular weight in cardiomyocytes. To establish immunospecificity, polyclonal antibodies were also subjected to preblocking with their respective immunogen peptide as follows: ␤ 1 -AR antibody and a peptide corresponding to the C terminus of the mouse ␤ 1 -AR; ␤ 2 -AR antibody and a peptide corresponding to the C terminus of the mouse ␤ 2 -AR, G␣ s/olf antibody and a peptide corresponding to the C terminus of rat G␣ s ; type V/VI adenylyl cyclase antibody and a peptide corresponding to the C terminus of human adenylyl cyclase V; GRK2 antibody and a peptide corresponding to the C terminus of human GRK2 (amino acids 675-689); RI␣ antibody and a peptide corresponding to the C terminus of human RI␣ (amino acids 343-361); RII␣ antibody and a peptide corresponding to the C terminus of human RII␣ (amino acids 385-404); ␣-catalytic subunit of PKA antibody and a peptide corresponding to the C terminus of human PKA␣; clathrin antibody and a peptide corresponding to the N terminus of human clathrin heavy chain; m 2 -mAChR antibody and a fusion protein of glutathione S-transferase and amino acids 225-356 of the i3 intracellular loop of the human m 2 -mAChR. For all but the anti-m 2 -mAChRantibody, 20 g of antibody was preincubated with 100 g of peptide in a final volume of 0.6 ml for 2 h at room temperature. 6 g of the anti-m 2 -mAChR antibody was preincubated with 30 g of the respective protein in a final volume of 80 l for 1 h at room temperature. After preincubation, antibodies were brought to working concentration (see above).
Immunoprecipitation-Cardiomyocytes from two 100-mm diameter dishes (ϳ2 mg of protein) were rinsed with ice-cold phosphate-buffered saline and harvested by the addition of 1.8 ml of extraction buffer (10 mM Tris-Cl, pH 8, 150 mM NaCl, 2 mM phenylmethylsulfonyl fluoride, 60 mM octyl glucoside). The cells were scraped and sonicated. Lysates were centrifuged at 4°C for 15 min at maximal speed in a microcentrifuge, and the supernatant was removed. For immunoprecipitation, supernatant was incubated with anti-caveolin-3 antibodies or irrelevant mouse IgG1 for 1 h at 4°C followed by the addition of 120 l of a 1:1 slurry of protein G-Sepharose beads (Sigma) and incubation overnight at 4°C. The beads were washed three times with extraction buffer, and bound proteins were eluted with 140 l of SDS-PAGE sample buffer and boiled for 5 min. Samples were subjected to SDS-PAGE and immunoblotting with caveolin-3 and adenylyl cyclase type V/VI-specific antibodies.
FIG. 1. Subcellular distribution of components of the ␤-AR signaling complex in cardiomyocytes. A, cardiomyocyte cultures were lysed in sodium carbonate followed by subcellular fractionation using a 5-35% discontinuous sucrose gradient as described under "Experimental Procedures." 1-ml fractions collected from the top of the gradient were concentrated (see "Experimental Procedures"). Aliquots (22 g of protein or a volume equal to that of the fraction with the least amount of detectable protein for "protein-free" fractions 1-3) were subjected to SDS-PAGE, electrophoretically transferred to nitrocellulose, and probed with the indicated antibodies. Fractions 4 and 5 represent the 5-35% sucrose interface, while the fractions at the bottom of the gradient are the 40% sucrose cushion, and P represents the insoluble pellet. This scheme typically accomplishes an ϳ500-fold purification of caveolin-3 relative to total cellular protein; ϳ15 g of caveolin-3-enriched domains are purified from ϳ8 mg of total cellular protein. Cholesterol Depletion-Cholesterol depletion was accomplished by incubating cells in the presence of the cholesterol-binding agent 2-hydroxypropyl-␤-cyclodextrin (2% for 1 h at 37°C). Preliminary experiments established that cyclodextrin caused a 66.3 Ϯ 3.2% decrease in total cellular cholesterol (n ϭ 3). The profound reduction in cholesterol was largely reversed within 1 h of subsequent incubation with cholesterol-cyclodextrin complexes at 37°C. For these experiments, a stock solution of 0.4 mg/ml cholesterol and 10% cyclodextrin was prepared by adding 200 l of cholesterol (20 mg/ml in ethanol) to 10% cyclodextrin at 40°C. The solution was filtered through a 0.2-m filter prior to use.
cAMP Measurements-Intracellular cAMP was measured according to standard methods essentially as described previously (22). Briefly, neonatal myocytes grown in 22.1-mm multiwell dishes were preincubated for 60 min at room temperature with 10 mM theophylline. Assays were performed for the indicated time intervals at room temperature and were terminated by removal of the incubation buffer and the addition of 1 ml of ethanol. Each condition was performed on three wells and was assayed for cAMP in quadruplicate. Aliquots of the alcoholfixed cell extract were dried under a stream of nitrogen, and cAMP in the residue was determined using a radioimmunoassay (PerkinElmer Life Sciences).

Selective Association of Individual ␤-AR Subtypes with
Caveolae-The cAMP pathway in cardiac cells was among the first and most rigorously studied signal transduction pathways, but there has been surprisingly little scrutiny of the subcellular localization of individual ␤-ARs and their downstream effector molecules. Studies of ␤-AR targeting in particular have been confined to ␤-ARs overexpressed at high levels in heterologous expression systems (10,17). There is no information on the targeting of native ␤-ARs expressed at physiologic concentrations in cardiomyocytes. Neonatal rat ventricular myocytes represent an optimal model to compare the subcellular targeting of individual ␤-AR subtypes and their downstream effector components as they co-express ␤ 1 -and ␤ 2 -ARs; both ␤ 1 -and ␤ 2 -ARs couple to an increase in intracellular calcium and enhanced contractility via a cAMP-dependent pathway in this cell type (3).
Caveolae were prepared according to two methods for these studies. First, membranes enriched in caveolin were separated from the bulk of cellular membranes and soluble proteins by extraction in detergent-free alkaline sodium carbonate buffer followed by isopycnic centrifugation on a sucrose gradient. We previously demonstrated that the light scattering band that forms at the 5/35% sucrose interface (fractions 4 and 5) of bottom-loaded discontinuous sucrose gradients contains ϳ50 -100-nm vesicular structures by transmission electron microscopy and the bulk of the cellular caveolin-3 (but excludes Ͼ99% of other cellular proteins including markers of the Golgi (25)). The results of immunoblot analyses examining the partitioning of selected components of the ␤-AR signaling cascade across these gradient fractions are shown in Fig. 1. This isolation method identifies resident caveolae proteins, but it provides no information as to whether proteins that are excluded from caveolae reside on the remainder of the cell surface plasma membrane or on intracellular membranes. Therefore, the partitioning of selected components of the ␤-AR signaling complex between caveolae, the remainder of the cell surface plasma membrane, and intracellular membranes was compared according to a second method (13). Here, plasma membranes were purified from the remainder of the intracellular proteins by Percoll gradient centrifugation. Membranes were then subjected to two OptiPrep density gradient centrifugations to separate and concentrate light caveolae vesicles from noncaveolae plasma membranes. Immunoblot analysis on these fractions is shown in Fig. 2.
The predicted molecular mass of the ␤ 1 -AR is ϳ51 kDa (26). However, an antibody directed against the C-terminal domain of the ␤ 1 -AR specifically recognizes multiple protein species, including bands that migrate at approximately 45, 51, 65, 96, and 150 kDa (shown in Fig. 1 and in more detail in Fig. 3). Immunoreactivity for all of these species is blocked by preincubation of antiserum with peptide antigen (Fig. 3A). Similar molecular heterogeneity of ␤ 1 -ARs was detected previously in the hearts of wild type mice but not ␤ 1 -AR null mutants (27), suggesting that each of these immunoreactive species represents a bona fide ␤ 1 -AR gene product. Fig. 3B shows that the 96-and 65-kDa species undergo appreciable mobility shifts upon treatment with PNG-F (to remove glycosyl moieties). Hence, even the 96-kDa species, which migrates as a relatively sharp band, is a glycosylated protein. Surprisingly, the diffuse migration of the smaller 51-kDa species is not appreciably increased by PNG-F. Of note, treatment with PNG-F does not eliminate the molecular heterogeneity of the ␤ 1 -AR (even when incubations are extended for longer intervals and samples are treated with more enzyme). These results indicate that factors other than protein N-glycosylation contribute to the differences in ␤ 1 -AR migration. Although caveolae are isolated in buffers that do not contain protease inhibitors, the multiple bands are not likely to result from proteolytic degradation during sample preparation, since identical results are obtained in particulate fractions prepared according to a protocol that uses protease inhibitors liberally (Fig. 3A). Although the 96-and 150-kDa species are detected in reducing SDS-PAGE sample buffer (in some experiments even with 10% ␤-mercaptoethanol, data not FIG. 2. OptiPrep gradients reveal the differential targeting of ␤-AR subtypes, G protein ␣ subunits, and adenylyl cyclase V/VI between caveolae and the remainder of the cardiomyocyte plasma membrane. Caveolae were prepared from quiescent cardiomyocyte cultures by sequential OptiPrep gradient centrifugation according to the method of Smart et al. as described under "Experimental Procedures" (13). Approximately 26 Ϯ 5% of the total plasma membrane of cardiomyocytes is recovered as light membrane vesicles according to this method (n ϭ 3, consistent with previous evidence that a large proportion of the total plasma membrane surface area of endothelial cells, smooth muscle cells, and adipocytes is represented by caveolaederived membranes (11,52)). Samples ( shown), they correspond to the approximate predicted size of receptor dimers/trimers or nonspecific aggregates. A number of G protein-coupled receptors form disulfide-linked SDS-resistant dimers (28,29). To determine whether these bands represent ␤ 1 -AR complexes, cells were treated with 5 mM N-ethyl-maleimide or iodoacetamide prior to cell lysis and then maintained in buffers supplemented with these SH group blocking agents throughout membrane isolation. Fig. 3A shows that by preventing disulfide bond exchange reactions during the preparation of samples for SDS-PAGE, the 96-and 150-kDa species are converted to 65-and 51-kDa forms of the ␤ 1 -AR. These smaller immunoreactive species represent bona fide ␤ 1 -AR gene products; their mobility precisely matches the mobility of the species expressed by cardiomyocytes transfected with a ␤ 1 -AR receptor expression plasmid (using the adenoviral component system described by Kohout et al.; Fig. 3C (30)). Hence, the 96-kDa immunoreactive species is a ␤ 1 -AR, formed as a result of dimerization or nonspecific aggregation of 65-and 51-kDa species of ␤ 1 -AR. Of importance to the studies at hand, Fig. 1 shows that ␤ 1 -ARs are detected across the sucrose gradient; they reside in caveolae but also are abundant in heavy fractions. The OptiPrep gradient fractionation method shows that ␤ 1 -ARs distribute between caveolae, noncaveolae cell surface plasma membrane fractions, and internal membranes (Fig. 2). Caveolae prepared according to both methods appear to be enriched in the most rapidly migrating ␤ 1 -AR species, but the interpretation of this finding is uncertain. It is possible that incomplete glycosylation or other factors contributes to a true preferential targeting of the 51-kDa species to caveolae. Alternatively, an apparent increase in the abundance of the 51-kDa ␤ 1 -AR species in caveolae could be an artifact, due to a diminished propensity of free thiol groups on this form of the native ␤ 1 -AR to undergo disulfide bond exchange reactions and form larger species in the local caveolae microenvironment.
Immunoblot analysis of cardiomyocyte ␤ 2 -ARs was more straightforward. Although ␤ 2 -ARs are the minor ␤-AR subtype in cardiomyocytes (representing only approximately 16% of total ␤-ARs), they are readily detected as a broad ϳ66-kDa species in caveolin-enriched fractions of resting cardiomyocytes (obtained either by the alkaline sodium carbonate extraction and sucrose gradient flotation scheme (Fig. 1) or sonication and flotation on OptiPrep gradients (Fig. 2)). Specific ␤ 2 -AR immunoreactivity (blocked by competing antigen peptide) is detected in caveolae and not in any other fractions (even with 10-fold heavier protein loading and long exposures of the gel; Figs. 1, 2, and 4A). Indeed, ␤ 2 -ARs are enriched in caveolae relative to the total plasma membrane fraction, consistent with the conclusion that cell surface ␤ 2 -ARs are confined to caveolae (Fig. 2). The diffuse appearance of ␤ 2 -ARs in caveolae can be attributed to glycosylation, since this receptor migrates as a distinct ϳ47-kDa band (in close agreement with its calculated molecular mass deduced from the gene sequence) following treatment with PNG-F (data not shown). Collectively, these studies indicate that in resting cardiomyocytes the cell surface distribution of ␤ 1 -and ␤ 2 -ARs is quite different. Although ␤ 1 -ARs are readily detected in all membrane fractions, the vast majority of the ␤ 1 -AR subtype is recovered in noncaveolar fractions along with Ͼ99% of total cell protein. In contrast, ␤ 2 -ARs are confined to caveolae.
To address the possibility that ␤ 2 -ARs localize to caveolae as a result of the low ␤ 2 -AR expression level in cardiomyocytes, the analysis was extended to cardiac fibroblasts. Fig. 5A shows that ␤ 2 -ARs migrate as an approximately 66-kDa band in cardiac fibroblasts. ␤ 2 -ARs are abundant in the caveolar fraction, but they also are detected in the heavy sucrose gradient fractions (F8 -12). To determine whether the ␤ 2 -ARs in the heavy fraction reside in noncaveolae plasma membranes versus internal membrane, caveolae were purified from the remainder of the plasma membrane by flotation of sonicated plasma membranes on OptiPrep gradients. Fig. 5B shows that fibroblast ␤ 2 -ARs are recovered entirely from caveolae and internal mem- Immunoblot analysis was with an antibody that recognizes the C terminus of the ␤ 1 -AR (lanes 1-3) or following antibody preblocking with antigen peptide (lane 4). Epitope-specific ␤ 1 -AR immunoreactivity is denoted by the arrowheads. B, the ␤ 1 -AR is expressed as multiple glycosylated forms in cardiomyocytes. Caveolae (lanes 1, 2, and 7), heavy sucrose fractions (lanes 3, 4, and 8), and the insoluble pellet fraction (lanes 5, 6, and 9) were prepared, and aliquots of protein pooled from these fractions were treated with PNG-F as described under "Experimental Procedures." Samples (25 g/lane) were then subject to SDS-PAGE and immunoblot analysis with the ␤ 1 -AR antibody (left) or following preblocking of this antibody with antigen peptide (right). Panel C, detection of heterologously expressed ␤ 1 -ARs in cardiomyocytes. Samples were prepared at 24, 48, or 96 h after transfection with a ␤ 1 -AR expression vector or 24 h following transfection with the vector control (Ϫ) using the "adenoviral component system" described previously (30). Immunoblot analysis was with the ␤ 1 -AR antibody and 10 g of particulate protein fraction. branes; ␤ 2 -AR are excluded from noncaveolae cell surface plasma membranes. This result suggests that the strict localization of cell surface ␤ 2 -ARs to caveolae is a generalized phenomenon and not a unique property of cardiomyocytes.
A previous report indicated that m 2 -mAChRs are excluded from cardiomyocyte caveolae under basal conditions but traffic to caveolae following stimulation by agonist (9). m 2 -mAChR localization was identified by receptor binding techniques in that previous study. When the subcellular localization of m 2 -mAChRs is examined by immunoblot analysis with a highly sensitive/specific m 2 -mAChR antibody (and sensitive enhanced chemiluminescence detection), m 2 -mAChRs are readily detected as a broad 66 -67-kDa band in caveolae as well as in heavy sucrose gradient fractions (obtained by alkaline sodium carbonate extraction and sucrose gradient flotation; Fig. 4A). m 2 -mAChRs are detected in caveolae, noncaveolae cell surface plasma membranes, and internal membranes when the Opti-Prep gradient fractionation method is used (Fig. 4B). Although these sensitive techniques identify m 2 -mAChRs in caveolae, the vast majority of total m 2 -mAChRs are excluded from caveolae (along with Ͼ99% of total cell protein). Hence, the data essentially concur with the result published previously (9). Of note, these studies identify similar membrane distributions for m 2 -mAChRs and ␤ 1 -ARs in resting cardiomyocytes. The spatial co-localization of these receptors would be permissive for interactions at the level of cAMP formation. In contrast, ␤ 2 -ARs and m 2 -mAChRs largely segregate to separate membrane subdomains, providing a potential explanation for the previous observation that ␤ 2 -AR-dependent cAMP formation is refractory to inhibitory modulation by mAChRs.
Recent studies identify ␤ 2 -AR trafficking to clathrin-coated vesicles as part of a dual process to terminate activation of the G s -adenylyl cyclase pathway as well as to initiate mitogenic signaling (31,32). To determine whether agonist stimulation promotes the egress of ␤-ARs from caveolae to permit their trafficking to other cellular compartments, caveolae were prepared by sequential OptiPrep gradient centrifugation from quiescent cells and following stimulation with 10 Ϫ7 M isoproterenol for 30 min; this concentration of agonist maximally activates both ␤ 1 -and ␤ 2 -ARs. Fig. 6 shows that agonist stimulation leads to a dramatic decrease in the abundance of ␤ 2 -ARs in caveolae, with no change in the recovery of caveolin-3 (or caveolae protein). This is associated with at best a trivial  1 and 4) or 80 g of the pooled heavy sucrose fraction (F8 -12, lanes 2, 3, and 5) were probed with the indicated antibodies. Specificity for ␤ 2 -AR and m 2 -mAChR immunoreactivity was established by the parallel immunoblot analyses with antibody preblocked with antigen peptide (lanes 4 and 5). Lane 3 represents a 5 times longer exposure time for lane 2. Under these conditions, there is abundant m 2 -mAChR immunoreactivity, but no specific ␤ 2 -AR immunoreactivity is detected. The relatively sharp band that migrates just ahead of the 66-kDa molecular mass marker is nonspecific. B, caveolae were separated from NCMs and internal membranes (IM) by OptiPrep gradient fractionation, and samples (10 g/lane) were probed with the m 2 -mAChR antibody.
FIG. 5. Cell surface ␤ 2 -ARs are confined to caveolae in cardiac fibroblasts. A, caveolae were prepared by alkaline sodium carbonate extraction and isopycnic centrifugation on a sucrose gradient. Caveolae (CAV, 23 g), pooled heavy sucrose fraction (F8 -12, 100 g) or insoluble pellet (P, 100 g) were probed for ␤ 2 -AR immunoreactivity. B, caveolae were prepared by OptiPrep gradient fractionation, and samples were probed for ␤ 2 -AR immunoreactivity. The individual fractions are as described in Fig. 4. The sharp band that co-migrates with the 66-kDa molecular mass marker is nonspecific immunoreactivity.
FIG. 6. ␤ 2 -ARs translocate out of caveolae following stimulation by isoproterenol. Cardiomyocytes were incubated in the absence or presence of 10 Ϫ7 M isoproterenol for 30 min at 37°C. Caveolae (CAV) and the remainder of the plasma membrane (NCM) were separated by sequential OptiPrep gradient centrifugation. Samples (12 g each) were subjected to immunoblot analysis with the indicated antibodies. The arrowheads denote the multiple immunoreactive species detected with the antibody for the ␤ 1 -AR and single diffuse immunoreactive band detected with the antibody for the ␤ 2 -AR (see Figs. 3 and 4). For the ␤ 2 -AR, the relatively sharp band that migrates just ahead of the 66-kDa molecular mass marker (particularly in the NCM fraction) is nonspecific immunoreactivity (see Fig. 4). The experiment was performed three times on separate cardiomyocyte culture preparations with similar results.
increase in the abundance of ␤ 2 -AR in the noncaveolae surface plasma membrane fraction. Based upon models of agonistinduced trafficking of ␤ 2 -ARs to clathrin-coated vesicles, a commensurate increase in ␤ 2 -ARs in the noncaveolae cell surface membrane would not be expected, since the bulk of the cellular clathrin is excluded from this fraction. Agonist-induced changes in ␤ 2 -AR abundance in caveolae are specific (blocked by preincubation with propranolol to prevent receptor activation; data not shown). Agonist-stimulated trafficking is confined to the ␤ 2 -AR; agonist-dependent changes in abundance (or mobility) of any molecular form of the ␤ 1 -AR was not detected under these assay conditions.
Selective Association of Downstream Components of the Cardiomyocyte cAMP Signaling Pathway with Caveolae-G protein subunit partitioning between caveolae and the remainder of the cell is shown in Fig. 1. Individual G protein subunits distributed quite differently across the sucrose gradient. Neonatal cardiomyocytes express both short and long splice variants of G␣ s (33); these proteins, as well as ␤ subunits are recovered in both caveolae and the heavy sucrose fractions. Cultured neonatal rat ventricular myocytes also express three pertussis toxin-sensitive G␣ subunits: G␣ i2 , G␣ i3 , and G␣ o . Fig.  1 shows that G␣ i2 is highly localized to caveolae. Similar results were obtained with an antibody that identifies G␣ i3 /G␣ o (without discriminating between these proteins). Fig. 2 shows that when cell surface plasma membranes are partitioned into caveolae and noncaveolae fractions by the OptiPrep-based cell fractionation method, the caveolae are particularly enriched in G␣ i3/o , whereas long and short splice variants of G␣ s distribute between caveolae and the remainder of the cell surface membrane. G␣ s is detectable in internal membranes, whereas G␣ i is not. Quantitative analysis of a series of immunoblots (adjusted to linear range) reveals that the vast majority of G␣ i subunits, but only approximately ϳ50% of G␣ s and ␤ subunits (expressed relative to total immunoreactivity in the cell), are recovered in the caveolin-enriched fraction.
The cardiac adenylyl cyclase enzymes (types V and VI) are the predominant adenylyl cyclase isoforms detected in cardiomyocytes (34). Nevertheless, they are scarce membrane proteins and generally difficult to detect by immunoblot analysis. Fig. 1 shows that caveolae are markedly enriched in cardiac adenylyl cyclase isoforms (detected with an antibody that does not discriminate between the two cardiac adenylyl cyclase isoforms), with as much as 50% of the total enzyme in this fraction. Although the mobility of adenylyl cyclase under reducing conditions in SDS-PAGE is considerably slower than would be expected based upon the calculated molecular mass of this protein, this is due to glycosylation of the protein; adenylyl cyclase migrates as a 120-kDa protein following treatment with PNG-F (Fig. 7). It also is pertinent to note that adenylyl cyclase is detected as a diffuse band with a tail of immunoreactivity only in caveolae preparations; in all other preparations, adenylyl cyclase runs as a single tight immunoreactive band. This provided the first clue that adenylyl cyclase might interact with caveolin-3 oligomeric complexes that are easily detected when this region of the gel is stripped and reprobed with the caveolin-3 antibody (data not shown).
cAMP actions in the heart largely result from cAMP binding to the regulatory subunits of the dormant PKA heterotetrameric complex, causing the release of the catalytic subunits and the subsequent phosphorylation of target proteins. PKA enzymes are classified according to their regulatory subunit (RI or RII), which display known differences in molecular weight, affinity for cAMP, phosphorylation state, and subcellular localization (35). Fig. 1 shows that RI and RII display totally different distribution patterns. RI is detected in only trace amounts in caveolae; it is primarily recovered in the heavy fractions. Similarly, the ␣-catalytic subunit of PKA (whose signaling function is subject to inhibitory modulation through an interaction with caveolin (20)) is largely excluded from the caveolinenriched fraction of cardiomyocytes. In contrast, caveolae are highly enriched in RII.
GRKs phosphorylate agonist-activated G protein-coupled receptors. This generally is viewed as a mechanism to promote high affinity binding of arrestins, which acts to uncouple the receptor from G proteins and target the receptor for internalization via clathrin-coated pits. Nevertheless, there is recent evidence that the catalytic activity of GRK is inhibited through an interaction with caveolin (19). In resting cells, GRK2 primarily fractionates as a cytoplasmic protein. Nevertheless, a small fraction of GRK2 (5-15%) is reported to associate with the membrane fraction; in A431 cells, a large portion of this membrane-associated GRK2 is in the caveolin-enriched fraction (19). Fig. 1 shows that GRK2 is primarily a soluble protein in cardiomyocytes (recovered in the heavy sucrose fractions), but a minor component of GRK2 co-fractionates with caveolin-3.
Finally, flotillin-1 and flotillin-2/ESA are two newly described caveolae-associated proteins that are believed to act as functional homologous of caveolins (36,37). Although both were detected previously in murine heart muscle (37), immu- Cardiomyocyte lysates were subject to immunoprecipitation with caveolin-3-specific IgG1 or irrelevant mouse IgG1 as described under "Experimental Procedures." Washed immunoprecipitates were subject to SDS-PAGE and immunoblot analysis using adenylyl cyclase V/VI (top) or caveolin-3 (bottom) antibodies. The experiment was performed three times with similar results. noreactivity in intact heart tissue preparations could result from flotillin expression in cardiomyocytes or noncardiomyocyte-contaminating cellular elements, since these proteins (particularly flotillin-2/ESA) display a rather wide tissue distribution. Fig. 1B shows that flotillins are expressed by cardiomyocytes and that they co-fractionate with caveolae. Caveolae prepared according to the OptiPrep method also are enriched in flotillin-1 relative to the remainder of the plasma membrane (data not shown). Importantly, clathrin is not detected in the caveolin-3-enriched fractions (i.e. these fractionation methods exclude clathrin-coated vesicles).
Caveolin-3 Interacts in Vivo with Adenylyl Cyclase, but This Is Not Required for the Targeting of Adenylyl Cyclase to Caveolae-The diffuse mobility of the adenylyl cyclase enzyme in the caveolae fraction suggested an interaction with homo-oligomers of caveolin-3. To determine whether there is an in vivo binding interaction between type V/VI adenylyl cyclase and caveolin-3, cells were solubilized in Triton X-100-and octyl glucoside-containing buffer. Cleared lysates were subjected to immunoprecipitation with caveolin-3 or an irrelevant monoclonal antibody; bound proteins were eluted with SDS-PAGE sample buffer and subjected to electrophoresis and immunoblotting with adenylyl cyclase-and caveolin-3-specific antibodies. Fig. 8 shows that adenylyl cyclase co-immunoprecipitates with caveolin-3 antibodies; caveolin-3, but not adenylyl cyclase V/VI, is completely cleared from the postimmune supernatant (data not shown). Adenylyl cyclase V/VI does not co-immunoprecipitate with irrelevant antibody. These results indicate that the adenylyl cyclase V/VI enzyme specifically associates with caveolin-3 in intact cardiomyocytes.
Further studies indicate that an interaction with caveolin is not required for adenylyl cyclase V/VI (or ␤-AR subtype) localization to light vesicular membranes. These studies were performed in HEK293 cells, which lack caveolin-1 or -3 proteins but endogenously express type V adenylyl cyclase. We reasoned that HEK293 cells might represent an informative model to examine the caveolin-3 requirement for adenylyl cyclase targeting to low buoyant density membrane subdomains, since such fractions are readily isolated by equilibrium sucrose density gradient centrifugation from these cell extracts. Fig. 9A shows that a vesicular preparation that is highly enriched in the caveolin functional homologues, flotillin-1 and flotillin-2/ ESA, can be floated from HEK293 cell extracts; flotillins are completely excluded from the heavy fractions. Flotillin purification is over 1000-fold relative to total cell lysates (10 g of caveolae, which contain greater than 99% of total cell flotillins, was isolated from 10 mg of starting cell protein). Immunoreactivity for caveolin-1 and caveolin-3 is not detected in HEK293 cells, and clathrin is excluded from the light vesicular fraction prepared by this method. Consistent with the notion that flotillins drive formation of a low density vesicular structure, this preparation appears as 50 -200-nm vesicles as well as curved membrane fragments by transmission EM (Fig. 9B). Using the position of flotillins as a marker to track the "caveolae-related membrane" domains, Fig. 9A shows the co-fractionation of the endogenous type V/VI adenylyl cyclase enzyme and both ␤-AR subtypes. These results argue that neither caveolin-1 nor caveolin-3 is required for cardiac adenylyl cyclase or ␤-AR FIG. 9. ␤ 1 -ARs, ␤ 2 -ARs, and adenylyl cyclase V/VI are detectable in light membrane vesicles from HEK293 cells that lack caveolin expression. A, HEK293 cell derivatives that stably overexpress the ␤ 2 -AR were lysed in sodium carbonate followed by subcellular fractionation using a 5-35% discontinuous sucrose gradient as in Fig. 1 localization to low buoyant density membrane domains.
Functional Significance of Caveolin-3/Adenylyl Cyclase Interactions-To determine whether localization to caveolae constitutes a mechanism to regulate the functional activity of the adenylyl cyclase enzyme, cardiomyocytes were treated with cholesterol-binding drugs to disrupt the functional integrity of the very cholesterol-enriched caveolae membranes. Initial experiments indicated that invasive/membrane-permeable cholesterol-binding agents such as filipin and nystatin induce gross cardiomyocyte toxicity (at 5 g/ml for 20 -30 min) before any changes in cellular cholesterol (or cAMP) can be detected. This precluded their use in these studies. In contrast, 2% cyclodextrin (a membrane-impermeable cholesterol-binding drug) for 1 h extracts 2 ⁄3 of total cell cholesterol, without inducing any gross morphological toxicity or major changes in spontaneous automaticity (i.e. both cholesterol depletion and the cyclodextrin treatment appear to be well tolerated). Fig. 10 shows that treatment with cyclodextrin results in an obvious redistribution of caveolin-3 and adenylyl cyclase from the caveolae fraction to the heavy sucrose layer; there also is some shift in ␤ 1 -and ␤ 2 -ARs, whereas the movement of G␣ s is inconsistent. The functional consequences of cyclodextrin treatment were assessed by comparing the time course of cAMP accumulation in response to 10 Ϫ9 M isoproterenol (a low concentration, which primarily activates the predominant ␤ 1 -AR subtype) and 10 Ϫ7 M zinterol (a ␤ 2 -AR-selective agonist). However, given recent evidence that cholesterol depletion with cyclodextrin can interfere with clathrin-coated pit internalization (38,39) (which in theory could provide an alternate explanation for changes in receptor-dependent cAMP accumulation), cells also were challenged with forskolin, a direct activator of adenylyl cyclase. Fig. 11 shows that cAMP accumulation is markedly increased in cells that are cholesterol-depleted with cyclodextrin. In separate experiments, cholesterol depletion with cyclodextrin was followed by an additional incubation for 1 h with cyclodextrin-cholesterol complexes to replete cholesterol levels. This returned cAMP responses back to base-line values (data not shown), indicating that changes in cAMP accumulation result from a loss of cholesterol (and diminished caveolae integrity) rather than any direct effect of cyclodextrin.

DISCUSSION
This study provides direct evidence that conventional paradigms for ␤-AR-dependent cAMP signaling in cardiomyocytes must be extended to incorporate the concept of localization to membrane subdomains. The major findings of this study are that individual ␤-AR subtypes display markedly distinct subcellular targeting (both between intracellular and surface membrane compartments and between caveolae and noncaveolae compartments on the surface membrane), that ␤ 2 -ARs are confined to caveolae in the basal state and egress from caveolae upon activation, and that adenylyl cyclase V/VI is highly localized to cardiomyocyte caveolae, where it interacts with caveolin-3. The experiments in HEK cells establish that the prototypical caveolins are not required for ␤-AR and adenylyl cyclase V/VI targeting to low buoyant density membranes. Nevertheless, the experiments on cyclodextrin-treated (cholesteroldepleted) cardiomyocytes indicate that targeting to caveolae represents a mechanism to negatively regulate cAMP accumulation.
There is a substantial body of literature describing the properties of ␤ 2 -ARs in heterologous expression systems. Here, most studies describe ␤ 2 -AR redistribution to clathrin-coated vesicles following agonist stimulation (although two laboratories have presented immunocytochemical evidence that ␤ 2 -ARs segregate to non-clathrin-coated invaginations following incubation with a monoclonal antibody to the receptor and antimouse IgG-gold in epidermoid A431 cells (8,40)). There is little to no information on the subcellular distribution of native ␤-ARs in a physiologically relevant cell type such as a cardiomyocyte. In particular, few studies have examined the subcellular localization of ␤ 2 -ARs in quiescent cells or extended the analysis to consider ␤ 1 -ARs. While Ostrom et al. recently reported that epitope-tagged ␤ 1 -ARs heterologously expressed in cardiomyocytes are detected as a single 72-kDa band in caveolae (17), the interpretation of this finding is uncertain in the context of results reported herein (and in a previous study (27)), which identify the endogenous cardiomyocyte ␤ 1 -AR as multiple prominent immunoreactive species. This study identifies the endogenous cardiomyocyte ␤ 1 -AR as multiple molecular species that distribute between caveolae, noncaveolae cell surface plasma membranes, and internal membranes. It is FIG. 10. Cholesterol depletion with cyclodextrin promotes the egress of adenylyl cyclase V/VI and caveolin-3 from caveolae. Cardiomyocytes were incubated with vehicle (Ϫ) or 2% cyclodextrin (ϩ) for 1 h at 37°C and then subjected to lysis in sodium carbonate and subcellular fractionation using a 5-35% discontinuous sucrose gradient as in Fig. 1. Aliquots of protein (30 g/lane) pooled from fractions 4 and 5 (Caveolae) or 8-12 (Heavy Fractions) were probed for immunoreactivity for ␤ 1 -ARs, ␤ 2 -ARs, G␣ s , adenylyl cyclase V/VI, or caveolin-3. Results are representative of the results obtained in three separate preparations.
FIG. 11. Cholesterol depletion with cyclodextrin augments isoproterenol-, zinterol-, and forskolin-dependent accumulation of cAMP in cardiomyocytes. Following preincubation with 10 mM theophylline plus vehicle (open symbols) or 2% cyclodextrin (filled symbols) for 1 h at 37°C, cardiomyocytes were challenged for the indicated intervals with 10 Ϫ9 M isoproterenol, 10 Ϫ7 M zinterol, or 10 Ϫ5 M forskolin, and cAMP accumulation was measured as described under "Experimental Procedures." Cyclodextrin induced a trivial increase in basal cAMP accumulation over 30 min, but this was not statistically significant (control: 22 Ϯ 4.5 pmol/dish; cyclodextrin 32 Ϯ 4.8 pmol/dish; mean Ϯ S.E. of triplicate determinations from three separate experiments). important to recognize that Ͻ1% of total cellular protein is recovered in caveolae. Hence, the vast majority of total cellular ␤ 1 -ARs are excluded from caveolae in resting cardiomyocytes. The subcellular distribution of ␤ 1 -ARs is not grossly altered by ligand binding. In contrast, ␤ 2 -ARs partition exclusively to caveolae in resting cardiomyocytes and egress from caveolae upon ligand binding, a conclusion that fully accommodates the prevailing paradigm for clathrin-dependent endocytosis of ␤ 2 -AR following activation by agonist. ␤ 2 -AR localization in caveolae was detected with two distinct purification techniques (flotation of sodium carbonate-insoluble membranes on sucrose gradients and flotation of sonicated plasma membranes on OptiPrep gradients), lessening the likelihood that ␤ 2 -ARs are in contaminating noncaveolae membrane fragments with sufficiently similar physical properties that they are not resolved by centrifugation in density gradients. The high degree of localization of cell surface ␤ 2 -AR to caveolae is not unique to cardiomyocytes (which contain only a minor population of ␤ 2 -ARs), since cell surface ␤ 2 -ARs also are confined to caveolae in quiescent cardiac fibroblasts. These results suggest that changes in ␤ 2 -AR expression over the physiologic range do not alter the fidelity of ␤ 2 -AR targeting to this structure. Whether targeting fidelity is retained in the context of very high levels of (epitope-tagged) ␤ 2 -AR overexpression must be determined in future studies, since prevailing concepts regarding ␤-AR function rely heavily on the results of studies in overexpression systems.
Differences in the cell surface partitioning of ␤ 1 -AR versus ␤ 2 -AR provide a rational explanation for previously identified differences in m 2 -mAChR-␤-AR subtype interactions at the level of cAMP accumulation (2). The similar cell surface distributions for m 2 -mAChRs and ␤ 1 -ARs would be permissive for interactions on the cell surface. In contrast, the targeting of ␤ 2 -AR to caveolae, a compartment that is distant from most of the cell surface m 2 -mAChRs, represents a very plausible mechanism to explain the absence of any ␤ 2 -AR-mAChR interaction at the level of cAMP formation. However, targeting to caveolae also may constitute a mechanism to facilitate adenylyl cyclase activation by the more minor ␤ 2 -AR population. The current dogma is that ␤ 2 -ARs inherently couple to the activation of adenylyl cyclase better than ␤ 1 -ARs (1,41,42). While this has been attributed to structural differences between ␤ 1 -and ␤ 2 -ARs (43), the results of this study suggest that the spatial proximity of ␤ 2 -ARs with the adenylyl cyclase enzyme in caveolae may selectively enhance the efficiency of signal transduction from this ␤-AR subtype to the adenylyl cyclase enzyme. In this regard, the drastic differences in the relative density of ␤ 1versus ␤ 2 -ARs in caveolae (which contain all of the ␤ 2 -ARs and only a fraction of ␤ 1 -ARs) and the remainder of the plasma membrane (which only has ␤ 1 -ARs) also are worth emphasis. Domain-specific differences in the stoichiometry of various elements in this signaling cascade are predicted to impact significantly on the efficiency of signal transduction.
In addition to ␤-AR subtypes, this study describes marked differences in the extent to which other components of the signaling machinery required to generate, propagate, or downregulate the cAMP signal target to caveolae. For example, G␣ s and G␣ i differ markedly in the extent to which they localize to cardiomyocyte caveolae. This could provide a mechanism to generate a gradient in cAMP levels in cells exposed to catecholamines and is of interest, given the early evidence that local pools of cAMP may differentially activate PKA-dependent functions (44,45). Similarly, the functional consequences of the cAMP signal may be modulated by local differences in the PKA enzyme. This study demonstrates that the RII subunit is abundant in caveolae; high concentrations of RII at this site would effectively increase local concentrations of the PKA II holoenzyme and could serve to promote the phosphorylation of proteins in the vicinity of this structure. In contrast, RI is excluded from this site, supporting the provocative (but as yet unproved) hypothesis that PKA I can phosphorylate a distinct spectrum of target proteins and subserve functions that are distinct from PKAII in the heart. Finally, this study demonstrates that GRK2 can be detected in caveolae. While GRK2 is reported to be negatively regulated by caveolin (19), the functional significance of this process in cardiomyocytes is uncertain, since only a minor component of GRK2 is detected in caveolae (at least under resting conditions).
On the basis of the numerous regulatory functions that have been attributed to caveolae, the integrated functional consequences of targeting to this microdomain are not entirely predictable. If one assumes that caveolin oligomers form a scaffold on the cytoplasmic surface of caveolae to sequester and organize signaling molecules, it might be predicted that disruption of this microdomain would impair cAMP-dependent signaling. Alternatively, the functional activity of multiple components of the ␤-AR complex (including such as G protein ␣ subunits and the adenylyl cyclase enzyme) are reported to be dampened through interactions with caveolin. According to this formulation, signaling would be enhanced by the removal of this negative regulatory function for caveolae. Results reported herein show that disassembly of caveolae with cyclodextrin (and dispersion of caveolae proteins) leads to enhanced cAMP accumulation in response to ␤ 1 -AR agonists, ␤ 2 -AR agonists, and forskolin. This favors a formulation in which caveolae act to repress cAMP formation, with the inhibitory control at the level of the adenylyl cyclase enzyme. There are at least two potential mechanisms that could account for local negative regulation of adenylyl cyclase in caveolae. First, based upon previous evidence that the activity of cardiac adenylyl cyclase isoforms is suppressed by caveolin-3-based scaffolding domain peptides (16) and the evidence reported herein that adenylyl cyclase and caveolin-3 co-immunoprecipitate, it is possible that in vivo adenylyl cyclase-caveolin-3 interactions tonically inhibit enzyme activity. Alternatively, there is recent evidence that the calcium-sensitive adenylyl cyclase VI isoform is susceptible to inhibitory regulation by calcium signals generated by capacitative calcium entry channels that co-localize to caveolae (47). A potential role for capacitative calcium entry channels in the regulation of adenylyl cyclase catalytic activity in electrically excitable cardiomyocytes, with spontaneous calcium cycling function, requires further study.
The identification of caveolae as a site to assemble functionally active ␤-AR complexes in cardiomyocytes is likely to be associated with important clinical implications. For example, there are almost decade-old reports that treatment of cardiac myocytes with cholesterol synthesis inhibitors in the presence of lipoprotein-depleted serum (which lowers plasma membrane cholesterol content and presumably disrupts the normal caveolar structure) leads to profound changes in ␤-AR responses in cardiac myocytes (48,49). These early studies suggest that caveolae play a critically important role in calibrating ␤-AR signaling to adenylyl cyclase. There also is recent evidence that caveolin-3 expression is regulated by the cAMP pathway and in heart failure syndromes (50,51). These studies suggest that changes in caveolin-3 expression and the associated alterations in the structural integrity of cardiomyocyte caveolae constitute an additional pathophysiologic mechanism that contributes to the altered autonomic regulation of heart.