Cholesterol oxidation switches the internalization pathway of endothelin receptor type A from caveolae to clathrin-coated pits in Chinese hamster ovary cells.

We investigated the mechanism of endothelin receptor type A (ETA) internalization in Chinese hamster ovary cells using two assays; flow cytometric quantification of cell surface myc-ETA and in situ localization of Cy5-labeled ET-1. In both assays, agonist-dependent internalization of myc-ETA was inhibited by nystatin and filipin, both of which disrupt internalization via caveolae, whereas it was barely affected by chlorpromazine and hypertonic sucrose, both of which disrupt internalization via clathrin-coated pits. In addition to myc-ETA, ET-1 caused intracellular translocation of caveolin-1 and this translocation was also blocked by nystatin but not by chlorpromazine. These results strongly argue that ETA is internalized via caveolae but not clathrin-coated pits. Treatment of the cells with cholesterol oxidase reduced cellular cholesterol and caused intracellular translocation of caveolin-1 but did not affect cell surface localization of myc-ETA. In cholesterol oxidase-treated cells, however, both chlorpromazine and hypertonic sucrose effectively blocked ET-1-induced myc-ETA internalization and nystatin was less effective than in untreated cells. Accordingly, expression of a dominant negative form of beta-arrestin blocked myc-ETA internalization in cholesterol oxidase-treated cells but not in untreated cells. These results suggest that, in Chinese hamster ovary cells, 1) agonist-occupied ETA can be internalized either via caveolae or clathrin-coated pits; 2) of the two, the former is the default pathway; and 3) the oxidative state of cell surface cholesterol is one of the factors involved in the pathway selection.

via caveolae (3,4). The molecular mechanism for sequestration of agonist-occupied GPCRs to clathrin-coated pits has been best studied for ␤ 2 -adrenergic receptors (␤ 2 AR) (2,5). In brief, upon binding of an agonist, ␤ 2 AR is phosphorylated by a GPCR kinase (GRK) and the phosphorylated receptors bind ␤-arrestin, which then targets the receptors to clathrin-coated pits. Various GPCRs other than ␤ 2 AR have been shown to be phosphorylated by GRK and thought to be targeted to clathrincoated pits in a similar manner (6 -9).
Caveolae are small flask-shaped invaginations of the plasma membrane, characterized by high levels of cholesterol and glycosphingolipids and also by the presence of caveolin, a 20 -24-kDa integral membrane protein (2,3). Although much less is known about the molecular mechanism involved, sequestration of agonist-occupied GPCRs to this microdomain has been demonstrated both by morphological and biochemical methods. For example, electron microscopic studies showed that both agonist-occupied ␤ 2 AR and bradykinin B2 receptors were localized in caveolae (10 -12). For other GPCRs such as angiotensin II type 1 and m2 muscarinic acetylcholine receptors, agonist-dependent sequestration in this microdomain was demonstrated by the recovery of receptor proteins in caveolin-rich fractions (13)(14)(15)(16).
Apart from the molecular mechanisms involved in the individual internalization pathways, the presence of the two distinct pathways raises a problem of pathway selection, at least for some types of GPCRs. A typical example is again ␤ 2 AR; some types of cells internalize agonist-occupied ␤ 2 AR via clathrin-coated pits (2,5), and others via caveolae (10,11), and it remains unknown how the distinct pathway is utilized in the individual cell type.
The endothelins (ETs) have a wide variety of biological effects in various tissues and cell types, which are mediated by specific GPCR subtypes, ETA and ETB (17). Like other GPCRs, agonist-occupied ET receptors are rapidly internalized and the same problem of pathway selection can be addressed at least for ETA, because previous studies have provided evidence both for the clathrin-coated pits and caveolae as an internalization pathway of this receptor subtype. ETA can be phosphorylated by a subtype of GRK (18). On the other hand, when expressed in COS cells, ETA colocalizes with caveolin-1 on the cell surface and binds caveolin-1 either in a direct or indirect way (19). Thus, it is unknown whether ETA can be internalized both via caveolae and clathrin-coated pits in a given cell type, or alternatively, the internalization pathway may differ between the cell types.
To address the question of pathway selection, we asked whether ETA is internalized via clathrin-coated pits and/or caveolae in CHO cells. CHO cells were used because of the presence of both internalization pathways (20) and also be-cause of our previous characterization of agonist binding and signaling of ETA expressed in this cell type (21). By applying both pharmacological and molecular manipulations to selectively disrupt the internalization pathways, we found that the internalization pathway of agonist-occupied ETA can be switched from caveolae to clathrin-coated pits by oxidation of cell surface cholesterol.
Plasmid Construction-For epitope-tagging of ETA, a double-strand oligonucleotide that encodes a myc epitope (EQKLISEEDL) was inserted in frame by PCR between nucleotides 69 and 70 (amino acids 23 and 24) of the human ETA cDNA (22). The insertion site was chosen to place the epitope at the N terminus of expressed receptors without interfering with signal peptide cleavage between amino acids 21 and 22. The myc-tagged ETA cDNA was subcloned into a mammalian expression vector pME18Sf to give pME/myc-ETA (23), and the insertion was verified by direct sequencing. A cDNA fragment that corresponds to amino acids 319 -418 of human ␤-arrestin (24) was obtained from a brain cDNA library using PCR. The PCR primers were designed to place a methionine residue on the N terminus, and a stop codon on the C terminus of the expressed protein. The PCR product was verified by sequencing and subcloned into pME18Sf to give pME/arrestin-(319 -418).
Cell Culture and Transfection-CHO cells were routinely maintained in Ham's F-12/10% fetal calf serum (FCS) at 37°C in a humidified atmosphere containing 5% CO 2 . Cells were transfected with pME/myc-ETA together with pSVbsr using LipofectAMINE. Cell populations expressing bsr gene product were selected by blasticidin (10 g/ml), and clonal cell lines were isolated by colony lifting. The density of receptors expressed in each clone was determined by saturation isotherms of 125 I-ET-1 binding to crude membrane preparations as described (21). Transient expression of pME/arrestin-(319 -418) and pEGFP (CLON-TECH) was also done with LipofectAMINE.
Drug Treatments-Prior to ET-1 or 125 I-transferrin application, cells were incubated at 37°C for 30 min in Krebs-Hepes buffer (KHB: 140 mM NaCl, 4 mM KCl, 1 mM CaCl 2 , 1 mM Na 2 HPO 4 , 1 mM MgCl 2 , 5 mM HEPES, pH 7.4, 11.7 mM glucose, 0.2% bovine serum albumin) containing nystatin (50 g/ml), filipin (5 g/ml), chlorpromazine (25 g/ml), or sucrose (0.4 M). Every agent was included in the medium throughout the subsequent incubations. None of them was toxic to the cells at the concentrations used as determined by a lactic dehydrogenase cytotoxic assay (data not shown). Stock solutions of nystatin and filipin were made in Me 2 SO. Me 2 SO at the final concentration of 0.1% affected neither the binding nor internalization of ET-1 and 125 I-transferrin (data not shown).
Flow Cytometry of Cell Surface myc-ETA-Cells in six-well plates were washed and incubated at 37°C in KHB containing ET-1 (100 nM) or BQ123 (1 M). At the indicated time, the cells were rapidly cooled on ice and collected by incubation in PBS, 5 mM EGTA followed by centrifugation. They were stained with anti-myc (4 g/ml) in PBS, 3% FCS, followed by Cy2-labeled anti-mouse IgG, each at 4°C for 1 h. They were resuspended in ice-cold PBS, 3% FCS containing propidium iodide (1 g/ml) and analyzed by a FACScan flow cytometer (Becton Dickinson Co., Mansfield, MA). Cy2 fluorescence was acquired from 5,000 -10,000 live cells gated by exclusion of propidium iodide. The specific Cy2 fluorescence was calculated as the difference between the mean fluorescence from cells stained by both primary and secondary antibodies and that from those stained without the primary antibody.
In Situ Binding and Internalization of Cy5-labeled ET-1-ET-1 was labeled with Cy5 using a reactive dye pack (Amersham Pharmacia Biotech) exactly as described (25). The specificity of Cy5-ET-1 binding to CHO/myc-ETA cells was verified by its absence in native CHO cells and by displacement by unlabeled ET-1 (data not shown). Cells on poly-Llysine-coated glass coverslips were washed with ice-cold KHB and incubated at 4°C for 1 h in the same buffer containing 10 nM Cy5-ET-1. To facilitate internalization of the bound ligand, the cells were washed and then incubated at 37°C for 30 min. In some experiments, a lysosomal marker dye lysotracker green (1 M, Molecular Probes Inc., Eugene, OR) was included in the medium at the 37°C incubation. At the end of the 4°C or 37°C incubation, fluorescent images were obtained with a MRC1024 laser-scanning confocal microscope (Bio-Rad, Osaka, Japan).

125
I-Transferrin Internalization Assay-Cells in 24-well plates were washed and kept in KHB at 37°C. The reaction was initiated by the addition of 125 I-transferrin (10 nM) and was terminated by rapid cooling of the cells on ice. After washing with ice-cold KHB, the cells were incubated twice in 1 ml of 50 mM glycine (pH 2.5) at 4°C for 2 min, and the residual radioactivities were recovered in 1 ml of 0.1 N NaOH. Radioactivities recovered in glycine (acid-sensitive) and NaOH (acidresistant) were defined as cell surface and internalized radioactivities, respectively. Nonspecific binding in the presence of 100-fold excess of unlabeled transferrin was always less than 5% of the total radioactivities recovered.
Cholesterol Oxidase Treatment and Thin Layer Chromatography (TLC)-Cells in 100-mm dishes were washed and incubated in KHB with or without 2 units/ml cholesterol oxidase at 37°C for 2 h. Where indicated, 25 g/ml cholesterol was added simultaneously with the enzyme. Cholesterol was dissolved in ethanol at 10 mg/ml. The vehicle (0.25% ethanol) did not affect the cellular contents of cholesterol or cholestenone (data not shown). To examine the recovery from cholesterol oxidation, a set of dishes were washed with KHB and incubated in F-12, 10% FCS in a CO 2 incubator up to 1 h. Extraction of total lipids and TLC analysis were done as described (26). In brief, lipids were extracted from cell pellets in chloroform/methanol (2:1 v/v). After saponification with NaOH, samples were dried under N 2 , dissolved in chloroform/methanol, and separated on HPTLC Silica gel 60 (Merck, Darmstadt, Germany). The solvent was 100% dichloromethane, which gave the best separation of cholestenone from other lipids in pilot experiments. Pure cholesterol and cholestenone were dissolved in chloroform/methanol and used as standards. Lipids were visualized by charring the plate with CuSO 4 and heating at 180°C. Densitometric analysis was done using NIH Image software.
Anti-caveolin-1 Western Blotting-Triton-soluble and -insoluble fractions of cell lysate were prepared as described (27). In brief, cells in 100-mm dishes were harvested by incubation for 10 min in MESbuffered saline (MBS, 25 mM MES, pH 6.5, 150 mM NaCl) plus 1 mM EDTA. Cells were collected by centrifugation, resuspended in 0.5 ml of MBS plus 1% Triton X-100, and incubated on ice for 20 min. The samples were Dounce homogenized 20 times and spun for 5 min at 16,000 ϫ g to give a Triton-soluble fraction (supernatant). The pellets were suspended in MBS plus 1% Triton X-100 by sonication to give a Triton-insoluble fraction. Protein concentrations were determined using a BCA protein assay kit (Bio-Rad). The samples were dissolved by SDS-12.5% polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane. The blot was probed with anticaveolin-1 according to the manufacturer's instructions, and bound antibody was detected using an ECL system (Amersham Pharmacia Biotech).
Immunofluorescence-Cells on poly-L-lysine-coated glass coverslips were stained with anti-ETA or anti-caveolin-1 using an indirect fluorescent technique. In brief, cells were fixed for 10 min in PBS, 4% paraformaldehyde, permeabilized in 50% ethanol, and incubated in Block Ace (Dainippon Pharmaceutical Co., Suita, Japan). They were then incubated with anti-ETA (1 g/ml) or anti-caveolin-1 (1 g/ml) at 4°C for 1 h. Bound antibody was detected with a biotinylated secondary antibody and Cy2-avidin, and fluorescent images were obtained with a confocal microscope.
Statistical Analysis-Student's t test was used for the statistical analysis of the results. p values of Ͻ 0.05 were considered to be significant.

Quantification of Cell Surface myc-ETA by Flow Cytometry-
Since it has been shown that ET-1/ETA binding is unusually stable and even resistant to acidic conditions (28), conventional internalization assays using 125 I-labeled ligand appear inapplicable. To quantify cell surface ETA, we generated a CHO cell line stably expressing an ETA variant that is tagged with a myc epitope on its N-terminal extracellular domain. 125 I-ET-1 binding to membrane preparations form CHO/myc-ETA revealed an irreversible binding interaction between 125 I-ET-1 and myc-ETA. The binding was saturable with a binding capacity of 3.2 Ϯ 0.3 pmol/mg protein (mean Ϯ S.E., n ϭ 3), and a halfmaximum binding was obtained at the ligand concentration of 63 Ϯ 4 pM. Signaling activities of both ET-1 and ET-3 on [Ca 2ϩ ] i and cAMP pathways were indistinguishable between wild type ETA and myc-ETA (data not shown). Anti-myc immunostaining of the cells followed by flow cytometry successfully detected cell surface myc-ETA (Fig. 1a). As expected, exposure to ET-1 caused a leftward shift of the fluorescent peak, which indicates agonist-induced reduction of cell surface receptors (Fig. 1a). This reduction did not occur at 4°C (data not shown) and was not elicited by an ETA-selective antagonist BQ123 (Fig. 1b), excluding inhibition of anti-myc binding to the epitope by ET-1. Time-dependent determinations showed that, upon ET-1 binding, cell surface myc-ETA decreased within 10 min and reached a plateau at 30 min (Fig. 1b).
In Situ Binding and Internalization of Cy5-ET-1-As an alternative assay for myc-ETA internalization, we took advantage of the irreversible nature of ET-1/ETA binding and visualized the localization of myc-ETA using fluorescent dye-labeled ET-1 (Fig. 2). When CHO/myc-ETA cells were incubated with Cy5-ET-1 at 4°C for 1 h, Cy5-ET-1 showed a homogeneous distribution on the plasma membrane. Subsequent incubation at 37°C elicited internalization of the surface-bound li-gand, and, after 30 min, the ligand showed a patchy distribution around the nucleus. Anti-ETA immunostaining showed co-localization of Cy5-ET-1 with myc-ETA on the cell surface at 4°C and in the intracellular compartment at 37°C, indicating that internalization of cell surface ETA can be followed by Cy5-ET-1. In the cells incubated at 4°C, anti-ETA also gave clear intracellular staining (Fig. 2, upper panels). Since this antibody recognizes the C terminus of the receptor, this staining is likely to represent ETA being glycosylated in the Golgi apparatus. These "reserve" receptors may be responsible for the anti-ETA staining that does not colocalize with Cy5-ET-1 at the 37°C incubation (Fig. 2, lower panels).
Effects of Nystatin/Filipin and Chlorpromazine/Hypertonic Sucrose on myc-ETA Internalization-To determine the internalization pathway of myc-ETA, we examined the effects of nystatin and filipin, both of which disrupt internalization via caveolae (29 -31), and chlorpromazine and hypertonic sucrose, both of which disrupt internalization via clathrin-coated pits (32)(33)(34). None of these drug treatments altered ET-1 binding capacities of the cells as examined by 125 I-ET-1 binding assays (data not shown). Quantification of cell surface myc-ETA by flow cytometry showed that ET-1-induced internalization was significantly inhibited both by nystatin and filipin, whereas it was barely affected by chlorpromazine and hypertonic sucrose (Fig. 3a), suggesting that myc-ETA is internalized via caveolae but not clathrin-coated pits. In both the case of nystatin and filipin, the inhibition was not complete. Alternative usage of clathrin-coated pits, however, seemed unlikely because simultaneous application of chlorpromazine with nystatin caused no further inhibition (Fig. 3a). The results from Cy5-ET-1 internalization assays were rather more clear-cut; the internalization was almost completely blocked by nystatin but not at all by chlorpromazine (Fig. 3b). As control experiments, we tested the effects of the same treatments on internalization of 125 I-transferrin, a ligand internalized via clathrin-coated pits (35). As expected, 125 I-transferrin internalization was effectively inhibited by chlorpromazine and hypertonic sucrose but not by nystatin nor filipin (Fig. 3c).
ET-1-induced Intracellular Translocation of Caveolin-1-CHO cells express a relatively high level of caveolin-1 (36, 37), which is a well characterized structural component of caveolae (3). To obtain further evidence for myc-ETA internalization via caveolae, we tested whether it was accompanied by caveolin-1 translocation. Anti-caveolin-1 immunostaining showed that, after the binding of Cy5-ET-1 at 4°C, caveolin-1 was localized both in the plasma membrane and intracellular compartments (Fig. 4a). The intracellular caveolin-1 is likely to reside in the Golgi apparatus (27). Subsequent incubation for 30 min at 37°C caused disappearance of caveolin-1 from the plasma membrane, indicating ET-1-induced intracellular translocation of caveolin-1 (Fig. 4a). As in the case of myc-ETA, this translocation was blocked by nystatin but not by chlorpromazine (Fig. 4b). Final destinations of caveolin-1 and ET-1/myc-ETA complexes, however, appeared to be different from each other because, although there was a partial overlap, the intracellular distribution of anti-caveolin-1 immunoreactivity was obviously different from that of Cy5-ET-1 at the end of the 37°C incubation (Fig. 4a).
myc-ETA Internalization in Cholesterol Oxidase-treated Cells-The integrity of caveolae appears to be dependent on cell surface cholesterol (3), and oxidation of caveolar cholesterol translocates caveolin-1 to the Golgi region (27). To further delineate the role of caveolae in ETA internalization, we tested the effect of cholesterol oxidase treatment of the cells on ET-1induced myc-ETA internalization.
TLC analyses of total cellular lipids showed that cholesterol oxidase (applied at 2 units/ml for 2 h) caused an obvious decrease in the level of cholesterol and generation of cholestenone, an oxidation product of cholesterol (Fig. 5, a and b). Neither effects were observed when the enzyme was inactivated with Triton X-100 (38) prior to application (data not shown). Simultaneous addition of free cholesterol (25 g/ml) with the enzyme partially blocked the reduction of cellular cholesterol but did not affect the generation of cholestenone. The effects of the enzyme were reversible; the cholesterol content returned to a normal level and cholestenone disappeared within 1 h after removal of the enzyme and incubation in F12/10% FCS (serum chase, Fig. 5, a and b). Anti-caveolin-1 Western blotting showed that, as reported for primary-cultured fibroblasts (27), cholesterol oxidase treatment caused translocation of caveolin-1 from a Triton-insoluble fraction to a soluble fraction (Fig. 5c). In accordance, immunostaining showed intracellular accumulation of caveolin-1 in cholesterol oxidasetreated cells (Fig. 5d). Both Western blotting and cell staining showed that this translocation was blocked by simultaneous addition of free cholesterol. As in the case of the effects on cellular lipids, the translocation was reversible; caveolin-1 dis-

FIG. 3. Effects of nystatin/filipin and chlorpromazine/hypertonic sucrose on myc-ETA internalization.
In a-c, the indicated agents were applied 30 min before the ET-1 exposure and were present throughout the subsequent incubations. Concentrations of the drugs in b and c are the same as indicated in a. a, flow cytometry. CHO/myc-ETA cells were incubated with 100 nM ET-1 for 30 min. Cell surface myc-ETA was quantified by anti-myc immunofluorescence and flow cytometry. b, Cy5-ET-1 localization. Cells were incubated with Cy5-ET-1 (10 nM) at 4°C for 1 h, washed, and then further incubated at 37°C for 30 min. Shown are the fluorescent images of Cy5 at the end of the 4°C or 37°C incubation. c, internalization of 125 I-transferrin. 125 I-Transferrin incorporation was quantified as described under "Experimental Procedures." In a and c, the internalization was expressed as relative to that in the absence of drugs (100%). Each bar represents means Ϯ S.E. of three independent experiments. *, p Ͻ 0.01, significantly different from the values of non-treated cells. chlor, chlorpromazine. appeared from the Triton-soluble fraction within 1 h of serum chase (Fig. 5c), and the intracellular accumulation disappeared within the same time range (Fig. 5d).
Cholesterol oxidase treatment did not affect the 125 I-ET-1 binding capacity of the cells (data not shown). Anti-myc flow cytometry showed that, in the enzyme-treated cells, ETA was still internalized upon binding of ET-1. In contrast to that in untreated cells (Fig. 3b); however, myc-ETA internalization was significantly inhibited both by chlorpromazine and hypertonic sucrose (Fig. 6a). The blocking effects of chlorpromazine/ hypertonic sucrose were not observed in cells treated with the enzyme in the presence of free cholesterol (Fig. 6a) and were lost after serum chase for 1 h (data not shown). The same enzyme treatment did not affect the cell surface localization of Cy5-ET-1 at 4°C and subsequent internalization at 37°C. In accordance with the results of flow cytometry, Cy5-ET-1 internalization under this condition was obviously blocked by chlorpromazine (Fig. 6b). This blocking effect of chlorpromazine was not observed in cells treated with free cholesterol and was lost after serum chase (data not shown). In the anti-myc flow cytometry, nystatin still caused a significant inhibition of the internalization (Fig. 6a) but the reduction (ϳ25%) was not as prominent as it was in untreated cells (ϳ60%). In the Cy5-ET-1 internalization assay, nystatin caused a partial blockage of internalization and the internalized Cy5-ET-1 was located not in the ordinary perinuclear space but in some compartments closer to the plasma membrane (Fig. 6b). As control experiments, we examined internalization of 125 I-transferrin in cholesterol oxidase-treated cells and found negative effects; as in untreated cells (Fig. 3c), internalization was effectively inhibited by chlorpromazine but not at all by nystatin (Fig. 6c).
Effects of Expression of a Dominant Negative Form of ␤-Arrestin on Cy5-ET-1 Internalization-The pharmacological evidence described above suggests that the internalization pathway of myc-ETA was switched from caveolae to clathrincoated pits by oxidation of cell surface cholesterol. To confirm this switching, we transfected the cells with pME/arrestin-(319 -418), which encodes a dominant negative form of human  (2), or cholesterol oxidase plus free cholesterol (3). After the enzyme treatment, cell were washed and chased in F12/10% serum for 0.5 h (4) or 1 h (5). a, TLC analysis of cellular lipids. b, densitometric quantification of the spots on TLC. Each bar represents means Ϯ S.E. of three independent experiments. c, anti-caveolin-1 Western blotting. Ten g of protein was loaded in each lane. Molecular masses are given on the left (kDa). d, anti-caveolin-1 immunostaining. Cells were fixed and processed for immunofluorescence as described in the legend to Fig. 4.   FIG. 6. myc-ETA internalization in cholesterol oxidase-treated cells. Cells were incubated at 37°C for 2 h without the enzyme or with cholesterol oxidase (2 units/ml) plus/minus free cholesterol (25 g/ml) and processed for anti-myc flow cytometry (a). Cells incubated with cholesterol oxidase (minus free cholesterol) were also processed for Cy5-ET-1 internalization assay (b) and 125 I-transferrin incorporation assay (c). The assay procedures and the concentrations of the applied drugs are described in the legend to Fig. 3. In a and c, the internalization was expressed as relative to that in untreated cells (100%). Each bar represents means Ϯ S.E. of three independent experiments. *, p Ͻ 0.01, significantly different from the values of untreated cells. (24). In these experiments, we used co-transfection of pEGFP to label the transfected cells. In untreated cells, expression of ␤-arrestin-(319 -418) caused no effect on Cy5-ET-1 internalization as shown by perinuclear clustering of Cy5 fluorescence in cells expressing GFP. In contrast, expression of ␤-arrestin-(319 -418) almost completely blocked Cy5-ET-1 internalization in cholesterol oxidase-treated cells (Fig. 7).

␤-arrestin
Double Labeling with Cy5-ET-1 and Lysotracker Green-Both caveolae-and clathrin-coated pit-mediated internalization pathways lead to endosomes from which ligand/receptors are transported to final destinations (1, 10 -12). In some systems, cellular destinations of receptors are altered depending on the internalization pathways to endosomes (39). To examine whether or not the destination of ET-1/myc-ETA complexes depends on the internalization pathway, we double-labeled the cells with Cy5-ET-1 and a lysosomal marker lysotracker green. After 30 min of incubation at 37°C, Cy5-ET-1 showed partial colocalization with lysotracker green; after 2 h of incubation, it showed an almost complete overlap. Cholesterol oxidase treatment did not affect the colocalization of Cy5-ET-1 with lysotracker green (Fig. 8). DISCUSSION Flow cytometric quantification of cell surface myc-ETA indicated that a major portion of cell surface myc-ETA was inter-nalized upon exposure to ET-1. However, a significant portion of myc-ETA (ϳ40%, Fig. 1) continued to reside on the cell surface after the exposure, despite the fact that the concentration of ET-1 applied (100 nM) was more than 5,00 times higher than that of 125 I-ET-1 required to saturate the binding sites in membrane preparations. Because in situ Cy5-ET-1 binding assays showed that, at least at the detectable level, all of the occupied receptors are rapidly internalized upon agonist binding (Fig. 2), it is reasonable to suppose that the residual antimyc binding sites represent unoccupied receptors that could be recognized by the antibody but not by ET-1. It is unlikely that the modified receptors function abnormally due to the incorporated myc epitope, because binding affinities to agonists as well as signaling activities on [Ca 2ϩ ] i and cAMP pathways were indistinguishable between the wild type ETA and myc-ETA. In addition, a mutagenesis study showed that the N-terminal half of the extracellular domain has little to do with the binding capacities of ETA (40). Although it is unknown why these receptors are left unoccupied in the presence of excess ligand, this is not a specific phenomenon for myc-ETA; similar observations have been reported both for epitope-tagged versions of platelet-activating factor receptors (20) and opioid receptors (41).
Here we presented both pharmacological and molecular biological evidence for caveolae as a default internalization pathway for myc-ETA in CHO cells. Both in flow cytometry and in situ binding assays, agonist-induced internalization of myc-ETA was inhibited by nystatin and filipin but not by chlorpromazine and hypertonic sucrose (Fig. 3). Accordingly, the dominant negative form of ␤-arrestin failed to inhibit internalization of Cy5-ET-1 (Fig. 7). These findings are in accordance with the observation by Chun et al. (19) that ETA expressed in COS cells colocalized with caveolin-1 on the cell surface. Unlike their finding of patchy distribution of the two molecules on the cell surface, our immunofluorescence imaging revealed rather homogenous distribution of both ETA (Fig. 2) and caveolin-1 (Fig. 4). The two molecules, however, were recovered in the same fractions in cell fractionation experiments, and, again in accordance with the observation by Chun et al. (19), the recovery did not depend on the presence of an agonist. 2 Therefore, it is plausible to conclude that, both in COS and CHO cells, expressed ETA is anchored to caveolae in the absence of a ligand and, upon binding of an agonist, is internalized through this pathway.
The most distinct finding of the present study was that cholesterol oxidase treatment switches the internalization pathway of ET-1/ETA complexes from caveolae to clathrincoated pits. Both in flow cytometry and in situ binding assays, chlorpromazine effectively blocked agonist-induced myc-ETA internalization in cholesterol oxidase-treated cells (Fig. 6). Accordingly, the dominant negative form of ␤-arrestin completely inhibited the internalization of Cy5-ET-1 under the same condition (Fig. 7). These results clearly indicate that agonist-induced myc-ETA internalization in cholesterol oxidase-treated cells involves clathrin-coated pits. An observation that apparently contradicts this conclusion was the effect of nystatin. Although less in degree compared with that in untreated cells, it still caused a significant inhibition of ET-1/myc-ETA internalization, whereas it had no effect on that of 125 I-transferrin (Fig. 6). Assuming that nystatin binds to oxysterols, there appear to be two possible explanations for the effect of nystatin in cholesterol oxidase-treated cells. One is that a part of ET-1/ myc-ETA complexes still uses caveolae as an internalization pathway; this is, however, unlikely because of the almost complete inhibition of internalization both by chlorpromazine and the dominant negative ␤-arrestin. An alternative explanation is that the internalization pathways used by 125 I-transferrin and ET-1/myc-ETA complexes are different from each other (although both involve clathrin-coated pits) and nystatin inhibits some steps specific for that used by ET-1/myc-ETA complexes. One possible step is the transfer of ET-1/myc-ETA complexes from their original sites (i.e. caveolae) to clathrin-coated pits. Thus, although the precise mechanism of nystatin action is unknown, the observed effects of nystatin do not exclude the involvement of clathrin-coated pits in the internalization of ET-1/myc-ETA complexes in cholesterol oxidase-treated cells.
Cholesterol is the most abundant sterol of the plasma membrane, and cholesterol oxidase has long been used to discriminate between the cell surface and the intracellular cholesterol pools (42). Treatment of the cells with this enzyme causes depletion of caveolins from the cell surface due either to depletion of free cholesterol or generation of oxysterol ( Fig. 5 and Ref. 27). It has also been long known that GPCR functions including both ligand binding and G protein activation are regulated by the level of cell surface cholesterol (43), and we have shown for the first time that the internalization pathway of a GPCR can also be regulated by cell surface cholesterol.
Given the association of ETA and caveolin-1 on the cell surface (although it is not known whether it is direct or indirect, Ref. 19), the most likely explanation for the pathway switching is that it is secondary to the presence or absence of caveolin-1 on the cell surface, i.e. in control cells (and also in nystatin-treated cells), ET-1/ETA complexes are anchored to caveolae because of the presence of caveolin-1, while in cholesterol oxidase-treated cells, they could diffuse away from caveolae to clathrin-coated pits because of the absence of caveolin-1. Two lines of evidence support this notion. First, simultaneous addition of free cholesterol with the enzyme partially blocked the reduction of cellular cholesterol (Fig. 5, a and b) and prevented both translocation of caveolin-1 (Fig. 5, c and d) and pathway switching of myc-ETA (Fig. 6). Because free cholesterol did not affect the generation of oxysterol, oxysterol is not primarily responsible for caveolin-1 translocation or pathway switching. Second, serum chase following the enzyme treatment restored the cholesterol level (Fig. 5, a and b), cell surface caveolin-1 (Fig. 5, c and d), and caveolae-mediated myc-ETA internalization. Thus, it appears that the translocation of caveolin-1 and the pathway switching of myc-ETA closely parallel each other and both events depend on the cellular cholesterol level. The data presented in the present study, however, do not exclude an alternative possibility that the switching is a direct consequence of free cholesterol depletion. It has been shown that, in fibroblasts, cholesterol oxidase treatment depletes cell surface caveolins but does not disrupt the morphological integrity of caveolae (27). Some types of cells including lymphocytes and neuronal cells do not express caveolins but do possess caveolae-like structures on the cell surface (44,45). Therefore, future studies using such cell lines without caveolin expression may provide an answer whether the pathway selection depends on the cell surface cholesterol per se or on the presence or absence of caveolins.
Internalized GPCRs are either recycled back to the plasma membrane or transported to lysosomes for degradation. Both in cholesterol oxidase-treated and -untreated cells, internalized Cy5-ET-1 showed a good colocalization with a lysosomal marker lysotracker green after 2 h of incubation at 37°C (Fig.  8). Therefore, unlike cholecystokinin receptors expressed in CHO cells whose destinations depend on the internalization pathway (39), agonist-occupied ETA appears to end up in lysosomes regardless of whether it is internalized through caveolae or clathrin-coated pits.
Besides intracellular destinations, the usage of differential internalization pathways may result in differential signaling activities of agonist-occupied ETA. Both caveolin and clathrin work as a scaffold for assembly of signaling molecules including G proteins (4) and cytoplasmic protein kinases (46,47). Although not definitely proven, it is well anticipated that the sets of molecules assembled are different between caveolae and clathrin-coated pits. We did not, however, address this issue in the present study because of the potential effects of cholesterol oxidation on the protein assembly.
Although pathophysiological significance of our findings remain uncertain, it will be of interest to examine whether the switching of the internalization pathway as observed in CHO cells occurs in vascular smooth muscle cells (VSMC) associated with atherosclerotic lesions. VSMC are a rich source of ETA (17), and circumstantial evidence for the idea includes cell surface oxidation of VSMC that is linked to the atherosclerotic change of the vascular wall (48) and the intracellular translocation of caveolin-1 that accompanied the phenotypic change of VSMC from a contractile to synthetic phenotype (49). Possible