Caveolae, Plasma Membrane Microdomains for α-Secretase-mediated Processing of the Amyloid Precursor Protein*

Caveolae are plasma membrane invaginations where key signaling elements are concentrated. In this report, both biochemical and histochemical analyses demonstrate that the amyloid precursor protein (APP), a source of Aβ amyloid peptide, is enriched within caveolae. Caveolin-1, a principal component of caveolae, is physically associated with APP, and the cytoplasmic domain of APP directly participates in this binding. The characteristic C-terminal fragment that results from APP processing by α-secretase, an as yet unidentified enzyme that cleaves APP within the Aβ amyloid sequence, was also localized within these caveolae-enriched fractions. Further analysis by cell surface biotinylation revealed that this cleavage event occurs at the cell surface. Importantly, α-secretase processing was significantly promoted by recombinant overexpression of caveolin in intact cells, resulting in increased secretion of the soluble extracellular domain of APP. Conversely, caveolin depletion using antisense oligonucletotides prevented this cleavage event. Our current results indicate that caveolae and caveolins may play a pivotal role in the α-secretase-mediated proteolysis of APP in vivo.

isoforms; APP 695 is the brain-specific isoform. The A␤ amyloid peptide is generated by the processing of APP with ␤and ␥-secretases (3). Alternatively, APP is processed by ␣-secretase, which cleaves APP within the A␤ sequence, thereby precluding the formation of A␤ (4). The identities of these secretases remain unknown.
In some inherited forms of Alzheimer's disease, point mutations have been identified within the coding sequence of the APP gene (5). These mutations co-segregate with the disease phenotype and cause Alzheimer's disease. Understanding the molecular function and processing of APP is therefore critical to unraveling the molecular basis of Alzheimer's disease.
One approach to elucidate the function of APP is to identify APP-interacting proteins. At least five distinct classes of molecules have been identified as APP binding partners as follows: G o (6), Fe 65 (7), X11 protein (8), Fe 65 -like protein (9), and APP-BP1 (10). The APP domain that interacts with G o has been localized to residues His 657 -Lys 676 within the cytoplasmic domain of APP 695 . G o is a brain-specific member of heterotrimeric GTP-binding protein (G-protein) family. The in vivo interaction between APP 695 and G o results in apoptotic cell death (11) and inhibition of cAMP response element trans-activation (12). In contrast, functional consequences of interactions between APP and other binding partners have not yet been described. However, it is likely that these APP-interacting protein molecules directly participate in APP-mediated cell signaling events.
A recent report described APP enrichment within caveolaelike domains in neurons (13). Caveolae are plasmalemmal microdomains where multiple signaling molecules are concentrated. Growth factor-mediated and G-protein-mediated signaling cascades have been shown to be initiated within these microdomains (14 -16).
A major protein component of caveolae is caveolin. Molecular cloning has recently identified a caveolin gene family. Three distinct caveolin genes have been identified, cav-1, -2, and -3 (17)(18)(19). In addition, two isoforms of caveolin-1 (Cav-1␣ and Cav-1␤) are derived from a single transcript from alternate translation initiation sites. Caveolins-1 and -2 are most abundantly expressed in adipocytes, endothelial cells, and fibroblastic cell types, whereas the expression of caveolin-3 is musclespecific. Caveolin proteins form homo-oligomers that bind cholesterol and glycosphingolipids. These protein-protein and protein-lipid interactions are thought to be essential for caveolae formation (20).
The molecular mechanism that underlies the recognition of signaling molecules by caveolin has recently been elucidated (27). Like other scaffolding proteins involved in signal transduction, caveolin contains a modular protein domain, termed the caveolin scaffolding domain. The caveolin scaffolding domain, in turn, recognizes a specific consensus motif within its interacting partners (28). As APP can function as a G o -coupled "receptor" (29) and contains a caveolin binding motif within its cytoplasmic domain, caveolin could provide a means for sequestering APP with caveolae membrane domains.
As APP is expressed in many non-neuronal cell types where caveolin-1 is abundantly expressed, we initiated this study to determine whether APP localizes within caveolae and if caveolin plays a role in this sequestration event. Here, we show that (i) endogenous APP co-fractionates with caveolin-1 in three distinct cell lines (COS-7, HEK293, and MDCK); (ii) recombinantly expressed APP co-immunoprecipitates with caveolin-1 when expressed in COS-7 cells; (iii) a GST fusion protein encoding the cytoplasmic domain of APP interacts directly with purified recombinant caveolin-1 protein. These three independent lines of evidence support the hypothesis that APP physically associates with caveolin.
Functionally, we show that the ␣-secretase cleavage product of APP is enriched within caveolae and that this cleavage event occurs at the cell surface. Specifically, we find that recombinant overexpression of caveolin-1 promoted ␣-secretase-mediated proteolysis of APP. The production of the ␣-secretase cleavage product of APP was abrogated by caveolin-1-based antisense oligonucleotides that effectively block caveolin-1 expression. These in vivo functional data clearly indicate that caveolae and caveolin proteins play a pivotal role in ␣-secretase-mediated processing of APP on the plasma membrane.

EXPERIMENTAL PROCEDURES
Materials-The cDNAs for canine caveolin-1 were as we described previously (30). Ni-NTA-agarose for purification of polyhistidine-tagged proteins was from Qiagen. Antibodies and their sources were as follows: anti-caveolin IgG polyclonal and monoclonal (2234) (gifts of Dr. John R. Glenney, Transduction Labs), anti-myc epitope IgG (monoclonal, 9E10; Santa Cruz Biotech), anti-HA epitope IgG (rat and mouse (12CA5) monoclonal, Boehringer Mannheim). A variety of other reagents were purchased commercially as follows: fetal bovine serum (JRH Biosciences), purified myelin basic protein (Sigma), and pre-stained protein markers (Bio-Rad and NOVEX). The anti-GST antibody was from Santa Cruz Biotechnology, and anti-myelin basic protein antibody was from Dako.
Cell Culture and cDNA Transfection-COS-7, HEK293, and MDCK cells (all from ATCC) were grown in DMEM plus 10% fetal bovine serum supplemented with penicillin and streptomycin (complete DMEM). cDNA transfections were conducted as described previously (36). Briefly, COS-7 cells were seeded at 1 ϫ 10 6 cells per 100-mm dish. Twenty-four hours later, DNA (6 g) of interest was transfected with 30 l of LipofectAMINE (Life Technologies, Inc.) in 5 ml of DMEM. 24 h post-transfection, the media were changed to complete DMEM. 48 h post-transfection, media or cells were harvested and subjected to further analyses.
cDNA Constructions-For construction of the cDNA encoding HAinserted APP, the APP 695 cDNA in pCDNA-1 (37) was digested with XmaI and dephosphorylated with calf intestine alkaline phosphatase (New England Biolabs). The oligonucleotides that encode HA sequence possessing XmaI restriction sites at both ends were phosphorylated, annealed, and ligated with the dephosphorylated APP cDNA. The orientation and sequence of the inserted fragment was confirmed by DNA sequencing. The HA peptide sequence was inserted between residues Pro 307 and Gly 308 of APP 695 . For construction of the GST fusion protein cDNA encoding the cytoplasmic domain of APP (Lys 649 -Asn 695 ), this domain was amplified as a fragment with additional restriction cleavage sites at both ends (EcoRI and BamHI sites). After PCR, the amplified fragment was digested with EcoRI and BamHI and subcloned in frame into the vector pGEX4T-1 (Amersham Pharmacia Biotech).
Detergent-free Purification of Caveolin-rich Membrane Domains-Cultured cells were grown to confluence in 150-mm dishes and used to prepare caveolin-enriched membrane fractions, as described previously (38). After two washes with ice-cold phosphate-buffered saline, cultured cells (two confluent 150-mm dishes) were scraped into 2 ml of 500 mM sodium carbonate, pH 11.0, and homogenized sequentially with a loosefitting Dounce homogenizer (10 strokes), a Polytron tissue grinder (three 10-s bursts; Kinematica GmbH, Brinkmann Instruments), and a sonicator (three 20-s bursts; Branson Sonifier 250, Branson Ultrasonic Corp.). The homogenate was then adjusted to 45% sucrose by the addition of 2 ml of 90% sucrose in MBS (25 mM Mes, pH 6.5, 0.15 M NaCl) and placed at the bottom of an ultracentrifuge tube. A 5-35% discontinuous sucrose gradient was formed above (4 ml of 5% sucrose, 4 ml of 35% sucrose; both in MBS containing 250 mM sodium carbonate) and centrifuged at 39,000 rpm for 16 -20 h in an SW41 rotor (Beckman Instruments). A light-scattering band confined to the 5-35% sucrose interface contained caveolin but excluded most other cellular proteins. The caveolae-rich fractions were diluted 3-fold with MBS and centrifuged at 15,000 rpm for 20 min at 4°C. The pellets were used as "purified caveolae-enriched membranes." This protocol separated caveolin from the bulk of cellular membranes and cytosolic proteins. This fractionation scheme is based on the specific buoyant density of caveolin-rich membrane domains and their resistance to solubilization by sodium carbonate. By using this scheme, endogenous MDCK (38) and COS-7 caveolin (39) are purified 2,000-fold relative to total cell lysates; approximately 90 -95% of caveolin is recovered in fractions 4 and 5 of these sucrose density gradients while excluding greater than 99.95% of total cellular proteins. By using this same fractionation scheme, cav-myc and cav-myc-H 7 co-fractionated with endogenous caveolin, an indication that the myc and polyhistidine tags do not interfere with the correct targeting of caveolin (38). Sodium carbonate extraction is routinely used to determine if proteins are firmly attached to membranes, and caveolin is not solubilized by sodium carbonate. By using this scheme, endogenous caveolin and cav-myc-H 7 were not only recovered almost quantitatively in fractions 4 and 5, while excluding most cellular proteins, but also were separated from the glycosylphosphatidylinositol-linked plasma membrane marker, carbonic anhydrase IV (38). This is consistent with recent observations that glycosylphosphatidylinositol-linked proteins are not concentrated directly within caveolae but may reside in close proximity to the "neck regions" of caveolae within intact cells (40).
As an alternative approach to purify caveolae, a protocol developed by Smart et al. (41) was employed. A plasma membrane fraction was prepared from 10 100-mm dishes of confluent tissue culture cells. Each dish was washed twice with 5 ml of buffer A (0.25 M sucrose, 1 mM EDTA, 20 mM Tricine, pH 7.8). Cells were collected by centrifugation at 1,400 ϫ g for 5 min (Sorval RT6000: 3000 rpm) and resuspended in 1 ml of buffer A and homogenized 15 times with Teflon glass homogenizer. Homogenized cells were centrifuged at 1,000 ϫ g for 10 min (Sorval RT6000: 2500 rpm), and the supernatant was subjected to Percoll gradient centrifugation. The sample was overlaid on top of 23 ml of 30% Percoll solution in buffer A and centrifuged at 83,000 ϫ g (30,000 rpm) for 30 min in a Beckman Ti-60 rotor. A plasma membrane fraction was collected, and the volume was adjusted to 2 ml in buffer A. 1.84 ml of 50% Optiprep in buffer B (0.25 M sucrose, 6 mM EDTA, 120 mM Tricine, pH 7.8) was added to 0.16 ml of buffer A (23% Optiprep solution), which was mixed with the sonicated plasma membrane fraction. The entire solution was placed at the bottom of a Beckman SW41 rotor tube and overlaid onto a linear 10 -20% Optiprep gradient (prepared by diluting 50% Optiprep in buffer A and B) and centrifuged at 52,000 ϫ g (18000 rpm) for 90 min using SW41 rotor (Beckman). The bottom fraction was collected (non-caveolae membrane). The top 5 ml of the gradient (fractions 1-6) was collected and mixed with 50% Optiprep in buffer B, which was then placed on the bottom of SW41 rotor tube and overlaid by 2 ml of 5% Optiprep in buffer A. The membrane fractions were centrifuged at 52,000 ϫ g for 90 min. An opaque band located just above the 5% interface was designated the caveolae fraction (41).
Immunoblotting of Gradient Fractions-From the top of each gradi-ent, 1-ml gradient fractions were collected to yield a total of 12 fractions. Caveolin migrates mainly in fractions 4 and 5 of these sucrose density gradients (38). Gradient fractions were separated by SDS-PAGE and transferred to Immobilon-P SQ sheets (Millipore). After transfer, sheets were stained with Ponceau S to visualize protein bands and subjected to immunoblotting. For immunoblotting using ECL, incubation conditions were as described by the manufacturer (Amersham Pharmacia Biotech), except we supplemented our blocking solution with both 1% bovine serum albumin, 3% nonfat dry milk, and 0.02% sodium azide. Chemical Cross-linking Studies-After transfection of cDNAs, cells were incubated with DMEM plus 10% Me 2 SO containing 48 g/ml bis-[␤-(4-azidosalicylamindo)ethyl] disulfide (Pierce) for 10 min at RT. Media were discarded, and cells were exposed to ultraviolet light (wavelength 365 nm) (model LM20E, VWR Scientific) for 15 min at 4°C. Cells were solubilized in a lysis buffer A (10 mM Tris/HCl, pH 8.0, 1 mM EDTA, 1 mM dithiothreitol, 60 mM octylglycoside, 1% Triton X-100, 1 mM aprotinin, 1 mM phenylmethylsulfonyl fluoride) with sonication (three 20-s bursts, Branson Sonifier 250, Branson Ultrasonic Corp). After centrifugation at 15,000 rpm for 20 min at 4°C, supernatants were collected and subjected to immunoprecipitation/immunoblot analysis. Cross-linking reagents were cleaved by boiling samples at 95°C for 5 min in the presence of 5% ␤-mercaptoethanol, and the samples were subjected to SDS-PAGE.
Immunoprecipitation of Caveolae-enriched Fractions-Caveolae-enriched fractions were diluted with Mes-buffered saline (MBS) and solubilized in buffer A without Triton X-100. After dialysis against buffer A without detergent, immunoprecipitation was conducted with anti-HA (Boehringer Mannheim) or anti-myc antibodies (Santa Cruz Biotechnology) overnight at 4°C. Fifty l of a ϳ50% slurry protein G-Sepharose (Amersham Pharmacia Biotech) was added and further incubated for 2 h. The beads were washed 3 times with Tris-buffered saline (TBS, 10 mM Tris/HCl, pH 8.0, 0.15 M NaCl), once with TE buffer (0.1 M Tris/HCl, pH 6.8, 1 mM EDTA), and subjected to SDS-PAGE/immunoblotting.

Affinity Purification of Caveolin-myc-H 7 Expressed in Insect Cells
Using Ni-NTA-Agarose-The Sf21 cells overproducing His-tagged caveolin-1 (76) were collected and solubilized in buffer A. Ni-NTAagarose (200 l) was then pre-equilibrated with TBS. Solubilized caveolin-containing cell extracts were then added to the resin and incubated for 6 h at 4°C on a rotating platform. After binding, the beads were allowed to gently settle by gravity (5 min on ice) and washed extensively (four times, 5 min each; twice with TBS and twice with TBS plus 30 mM imidazole). After washing, bound proteins were eluted with TBS containing 200 mM imidazole, as per the manufacturer's instructions (Qiagen).
Purification of GST-APP Cytoplasmic Region-An overnight culture of Escherichia coli harboring plasmids encoding either GST-APP or GST alone was purified as described previously (42).
Interaction of Caveolin-myc-H 7 with GST-APP Cytoplasmic Region Proteins-The interaction of GST-APP fusion proteins with caveolin-1myc-H 7 was evaluated as described for the interaction of caveolin with baculovirus-expressed heterotrimeric G-protein subunits (21). Briefly, GST or purified GST-APP cytoplasmic fusion proteins bound to glutathione-agarose beads were extensively prewashed with phosphate-buffered saline and lysis buffer containing protease inhibitors. These beads contained ϳ100 pmol of a given fusion protein per 100 l of packed volume. Approximately 100 l of this material was incubated with 1 g of caveolin-1-myc-H 7 on a rotating platform overnight at 4°C. After binding, the beads were washed extensively (6 -8 times) with buffer B (50 mM HEPES/NaOH, pH 7.5, 120 mM NaCl, 1 mM EDTA, 0.5% CHAPS, and protease inhibitors). Finally, bound proteins were eluted with 100 l of elution buffer containing 50 mM Tris, pH 8.0, 1 mM EDTA, 1% Triton X-100, 10 mM reduced glutathione, and protease inhibitors. The eluate was mixed 1:1 with 2ϫ sample buffer and subjected to SDS-PAGE (10% acrylamide) and Western blot analysis with anticaveolin polyclonal antibody (1:1000 dilution, Transduction Laboratories). Horseradish peroxidase-conjugated secondary antibodies (1:10000 dilution, Bio-Rad) were used to visualize bound primary antibodies by ECL (Amersham Pharmacia Biotech). In place of caveolin-myc-H7, purified myelin basic protein (a negative control) and FE 65 (a positive control) were also used for binding experiments. For detection of myelin basic protein, anti-myelin basic protein antibody was used. For the binding study of FE 65 and APP, the gene of human FE 65 PTB2 domain (484 -612 amino acids) was first cloned (43) by PCR with a set of 5Ј and 3Ј primers (5Ј primer, ACGCGTCGACTCACAGGAAGGAGAG-GAAACGCAC; 3Ј primer, GCGGATCCTGTGAGGCACCTGCCAA-GAAC) using human hippocampus cDNA library (CLONTECH) as a template. PCR product was digested with BamHI-SalI and ligated into pCAL-n-EK vector (Stratagene) for generating a fusion protein cDNA which contains calmodulin binding peptide and FE 65 -PTB2 fusion domains. Expression of calmodulin binding peptide-FE 65 -PTB2 protein was induced by isopropyl-1-thio-␤-D-galactopyranoside and was purified using calmodulin affinity resin (Stratagene) as described previously (44). The purified fusion protein (1 g) was used for binding analysis as described above. After transfer of eluted samples separated in SDS-PAGE, the membrane was blocked with 2% skim milk and 2% bovine serum albumin in PBS, 0.02% sodium azide overnight at 4°C. The membrane was incubated with 5 g/ml biotinylated calmodulin (Calbiochem) in washing buffer (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1 mM CaCl 2 ) for 30 min at RT, washed 3 times with the same washing buffer (10 min each), incubated with 1/20,000 diluted streptavidinhorseradish peroxidase-conjugated (Pierce) for 30 min at RT, and washed 3 times with the same washing buffer. Finally bands were visualized by diaminobenzidine staining (Vector).
Immunocytochemistry-Transfected cells (50,000 cells) were split onto 4-well chamber slides (Becton Dickinson). Twenty-four hours later, the cells were incubated for 15 min at 8°C with anti-HA rat monoclonal antibody (5 g/ml, Boehringer Mannheim) and cholera toxin B subunit conjugated with peroxidase (1 g/ml, Sigma) in complete DMEM (45). Cells were washed with PBS (4 ϫ 5 min) and fixed in PBS containing 4% paraformaldehyde. After washing with PBS (5 ϫ 5 min), nonspecific binding of antibodies was blocked with 3% normal goat serum (Life Technologies, Inc.) in PBS for 30 min at RT. Cells were then incubated with anti-cholera toxin rabbit antibody (Calbiochem) in PBS containing 3% normal goat serum for 1 h at RT. The chamber slides were washed with PBS (4 ϫ 5 min) and incubated for 60 min at RT with secondary antibodies (Texas Red-conjugated goat anti-rabbit IgG (preabsorbed with rat and mouse IgG), 1:300 dilution, Jackson ImmunoResearch, and biotin-conjugated goat anti-rat IgG, 1:300 dilution, Boehringer Mannheim) and washed in PBS and in 0.1 M NaHCO 3 , pH 8.2 (4 ϫ for 5 min). Cells were incubated with avidin-fluorescein isothiocyanate conjugate (1/300 dilution, Vector) in NaHCO 3 buffer for 1 h at RT. Slides were mounted with Vectashield (Vector) and analyzed with a Leica TCS NT confocal microscope.
Cell Surface Biotinylation-Cells were washed twice with ice-cold PBS and incubated with PBS containing 0.5 mg/ml sulfo-NHS-biotin (Pierce) for 30 min at 4°C (46), washed three times with ice-cold PBS, and placed in complete DMEM. Four hours after incubation at 37°C in a CO 2 incubator, media were harvested and mixed with the same volume of 2ϫ RIPA buffer, followed by incubation with Neutravidin beads (Pierce). After overnight incubation at 4°C, the Neutravidin beads were washed three times with RIPA buffer and once with TE buffer. Final samples were mixed with sampling buffer containing ␤-mercaptoethanol, boiled for 3 min, and subjected to immunoblot analysis with 22C11.
Antisense Oligonucleotide Introduction-COS-7 cells were seeded at a density of 1 ϫ 10 6 cells per 100-mm dish. Twenty-four hours later, 3 g each of antisense caveolin-1 oligonucleotide (TTTGCCCCCAGACAT; complementary to the 15-base initiation sequence of human caveolin-1) and the HA-APP 695 cDNA were co-transfected into COS-7 cells using 30 l of LipofectAMINE. Twelve hours later, the media were discarded and replaced with 8 ml of complete DMEM. Forty-eight hours post-transfection, media were harvested and mixed with 2ϫ RIPA buffer. As a control, a sense-caveolin-1 oligonucleotide (ATGTCTGGGGGCAAA) which corresponded to the 15 bases of the initiation sequence was used. The media and cells were subjected to further analyses.

RESULTS
Interaction of APP with Caveolin-1: A Requirement for the Caveolin Scaffolding Domain-To investigate whether endogenous APP is associated with caveolae membranes, caveolae-enriched membrane domains from COS-7 cells were purified by a detergent-free procedure and were ultracentrifuged in a sucrose density gradient. Immunoblot analysis of the resulting gradient fractions with APP antibody (22C11) detected the vast majority of APP in fractions 4 and 5 (Fig. 1A, upper panel). Similarly, caveolin-1 was confined to the same region of the gradient (fractions 4 and 5), indicating that caveolae are contained within these fractions (Fig. 1A, lower panel). Densitometric analysis revealed that greater than 90% of the total APP-reactive material was recovered in fractions 4 and 5. These results demonstrate that molecules containing the 22C11 epitope are primarily enriched with caveolae microdomains in COS-7 cells.
To extend these initial observations, the same fractionation procedure was employed on HEK293 and MDCK cells. In HEK293 cells, endogenous 22C11 immunoreactive bands were again highly enriched in fractions 4 and 5 and precisely coincided with the distribution of caveolin-1 (Fig. 1B). In MDCK cells, three major bands were detected by immunoblot analysis with 22C11 (Fig. 1C). Approximately 80% of the total 22C11 immunoreactive material was confined to the caveolin-rich fractions. In addition, when the independent fractionation procedure involved homogenization in Triton X-100 containing buffers, essentially identical results were obtained (not shown). Therefore, it is likely that the 22C11 immunoreactive proteins physically associated with specialized caveolin-enriched membrane microdomains in these three distinct cell lines.
The antibody 22C11 recognizes both APP and other molecules closely related to APP, termed APLPs (47). To determine if APP itself is confined to caveolae-enriched domains, we next employed a recombinant approach. COS-7 cells were co-transfected with epitope-tagged forms of APP and caveolin-1 (the HA epitope was inserted into the extracellular domain of APP 695 and the myc epitope was placed at the extreme C terminus of caveolin-1). These transiently transfected cells were then subjected to the same detergent-free fractionation procedure described above. HA-tagged APP 695 and myc-tagged caveolin-1 were confined mainly to fraction 4 and were undetectable in the remaining 11 fractions (Fig. 2A). These data indicate that the majority of recombinant APP 695 and caveolin-1 co-fractionate and are localized within caveolae-enriched membrane domains.  HEK293 (B), and MDCK cells (C) grown to confluence in 150-mm dishes were used to prepare caveolin-enriched membrane domains, as described previously (38). After two washes with ice-cold phosphate-buffered saline, cells (two confluent 150-mm dishes) were scraped into 2 ml of 500 mM sodium carbonate, pH 11.0. Homogenization was carried out using a loose-fitting Dounce homogenizer, a Polytron tissue grinder, and a sonicator. The homogenate was adjusted to 45% sucrose and placed at the bottom of an ultracentrifuge tube. A 5-35% discontinuous sucrose gradient was formed above (4 ml of 5% sucrose, 4 ml of 35% sucrose; both in MBS containing 250 mM sodium carbonate) and centrifuged at 39,000 rpm for 16 -20 h. From the top of each gradient, 1-ml gradient fractions were collected to yield a total of 12 fractions. Gradient fractions were separated by SDS-PAGE (4 -20% linear gradient gel) and transferred to Immobilon-P SQ . After transfer, the sheets were subjected to immunoblotting with 22C11 (upper panel) and anti-caveolin-1 antibodies (lower panel). Arrows indicate APP/ APLP (upper panel) and caveolin-1 (lower panel).
G-protein-coupled receptors (the endothelin receptor) and growth factor receptors can form physical complexes with caveolin-1 (48,49). In the case of epidermal growth factor receptor, caveolin-1 interacts with the epidermal growth factor receptor kinase domain via the caveolin scaffolding domain. As APP is physically and functionally coupled to the heterotrimeric G-protein, G o (29), APP may also interact directly with caveolin-1. Direct interaction of caveolin-1 with receptors and downstream signaling molecules is thought to facilitate sequestration of multiple signaling molecules within caveolae membranes. To test this hypothesis, we employed a chemical crosslinking approach coupled with immunoprecipitation. COS-7 cells were transiently co-transfected with epitopetagged APP and caveolin-1 as described above and subjected to chemical cross-linking with the membrane-permeant photoreactive agent, 9-bis-[␤-(4-azidosalicylamindo)ethyl] disulfide. After photolysis the cross-linked cells were solubilized with a buffer containing 1% Triton X-100 and 60 mM octylglycoside. Samples were then immunoprecipitated with anti-myc IgG to recover epitope-tagged caveolin. After immunoprecipitation, samples were subjected to immunoblot analysis to visualize HA-tagged APP.
As shown in Fig. 2B, two major bands corresponding to caveolin-1␣ and caveolin-1␤ were detected in these immunoprecipitates using the myc antibody for Western blot analysis (Fig. 2B, lane 2). Anti-HA immunoblot analysis revealed that HA-tagged APP was also detected in these immunoprecipitates. In the mock-transfected cells, neither caveolin-1 nor APP immunoreactivity were detected by Western blotting with anti-HA or anti-myc (Fig. 2B, lane 1). The results establish that caveolin-1 forms a physical complex with APP in intact COS-7 cells. Under the same experimental conditions expect using no chemical cross-linking reagent, the physical association of APP and caveolin was not observed, suggesting that treatment with chemical cross-linking reagent is necessary to observe sufficient binding of APP and caveolin in this transient expression system.
To investigate further the specificity of this interaction, the Cell lysates without chemical cross-linking were also subjected to immunoblotting with anti-HA antibodies (lower panel). C, recombinant APP and caveolin-1 co-immunoprecipitate (IP). After transfection of HA-APP with either caveolin-1␤ (cav-1) or caveolin-1 lacking residues 61-100 (⌬61-100), cells were subjected to detergent-free caveolae purification as in Fig. 1. Caveolae fractions were diluted with Mes-buffered saline, collected by centrifugation, and solubilized in buffer A lacking Triton X-100. After centrifugation at 15,000 rpm for 30 min, supernatants were collected and subjected to dialysis against buffer A without detergent. Immunoprecipitation was conducted with (i) anti-HA (HA) or normal rat IgG (NRG) (left panel), (ii) with anti-myc (myc) or normal mouse IgG antibody (NMG) (middle panel) and (iii) with anti-HA or anti-myc (right panel). The samples were subjected to immunoblotting with anti-myc, HA, or AC-1 antibodies. D, purified recombinant APP cytoplasmic domain interacts with purified recombinant caveolin-1. Recombinant caveolin-1 was affinity purified from the Sf21 cells overproducing His-tagged caveolin using the Ni-NTA-agarose. Recombinant GST-APP or GST was purified from E. coli harboring plasmids encoding either GST-APP or GST alone. GST-APP or GST alone were pre-bound to glutathione-agarose beads and incubated with caveolin-1-myc-H 7 . Bound proteins were eluted with buffer containing 10 mM reduced glutathione and were subjected to SDS-PAGE (12% acrylamide) and Western blot analysis with anti-caveolin-1 or anti-GST antibodies. Under the same binding conditions, the binding capacity of myelin basic protein and FE 65 PTB2 domain (484 -612 amino acids) was also determined. Myelin basic protein was detected by anti-myelin basic protein antibody, whereas FE 65 was detected by biotin conjugated-calmodulin binding, followed by diaminobenzidine staining. same experiment was performed using other forms of caveolin-1. We used myc-tagged caveolin-1␤, the shorter isoform of caveolin-1 that lacks caveolin residues 1-31, and myc-tagged caveolin-1-(⌬ 61-100), an internal deletion mutant that lacks the caveolin-1 scaffolding domain. Immunoprecipitation and Western blotting with anti-myc IgG detected both caveolin-1␤ and caveolin-1-(⌬ 61-100) (Fig. 2B, lanes 3 and 4), as expected. However, immunoblot analysis with anti-HA antibody revealed that APP only co-immunoprecipitated with caveolin-1␤ but not with caveolin-1-(⌬ 61-100). These results indicate that caveolin residues 1-31 are not required for interaction of APP with caveolin-1; in addition, they implicate caveolin residues 61-100 in this process, demonstrating a crucial role for the caveolin-1 scaffolding domain in the recognition of APP by caveolin-1 in vivo.
In another independent approach, we employed detergentfree immunoprecipitation. COS-7 cells were transiently cotransfected with epitope-tagged forms of APP and caveolin-1 (caveolin-1␤ and caveolin-1-(⌬ 61-100)) as described above. After detergent-free cell fractionation, caveolae containing fractions were collected and solubilized with 60 mM octylglycoside. After dialysis to remove detergent, these caveolin-rich membranes were subjected to immunoprecipitation with either anti-myc or anti-HA antibodies. Immunoblot analysis revealed that APP is associated with caveolin-1␤ but not caveolin-1-(⌬ 61-100) (Fig. 2C). Also, normal mouse IgG did not immunoprecipitate either HA-tagged APP or the myc-tagged forms of caveolin-1, confirming the specific nature of these interactions. The results indicate that APP and caveolin form physical complexes and they are consistent with chemical cross-linking studies of intact cells (Fig. 2B).
Although the experiments described above indicate that APP and caveolin-1 form a physical complex in intact cells, it remains unknown whether this interaction is direct or occurs via another protein that may act as a bridge to link APP with caveolin-1. To address this issue, the interaction between a purified GST fusion protein encoding the short cytoplasmic tail of APP and purified recombinant caveolin-1 containing both myc and His tags (caveolin-1-myc-His 7 ) was investigated. GST alone or GST-APP cytoplasmic tail immobilized on glutathioneagarose beads were incubated with purified recombinant caveolin-1. After binding, the beads were extensively washed with buffers containing high salt and detergents, and the bound proteins were then specifically eluted with buffer containing excess reduced glutathione. These eluates were then subjected to immunoblot analysis with anti-caveolin-1 IgG. Fig.  2D shows that caveolin-1 specifically associated with GST-APP cytoplasmic tail but not with GST alone.
We also used FE 65 APP binding domain as a positive control and myelin basic protein as a negative control for these experiments. Under our conditions, FE 65 PTB2 domain (484 -612 amino acids) but not myelin basic protein bound GST-APP cytoplasmic tail (Fig. 2D), indicating that the employed experimental conditions were appropriate to evaluate the direct interaction between the APP cytoplasmic region and its associated molecules. Taken together, these results clearly indicate that APP interacts directly with caveolin-1 through a portion of the cytoplasmic tail of APP.
Cell surface caveolae can be detected by cholera toxin B subunit binding (45,50). To investigate the possibility that cell surface caveolae contain APP, we immunostained living COS-7 cells with anti-HA and cholera toxin antibodies following incubation of the cells with cholera toxin B subunit after transfection of recombinant HA-APP and myc-caveolin cDNAs. As shown in Fig. 3, HA (green) and cholera toxin B subunit staining (red) are co-localized on the cell surface. These results support the hypothesis that APP is localized within cell surface caveolae in living cells.
In summary, we have shown here that (i) both endogenous and recombinant APP co-fractionate with caveolin-1, (ii) APP forms a physical complex with caveolin, (iii) the scaffolding domain of caveolin and the cytoplasmic domain of APP participate in this reciprocal interaction, and (iv) cell surface caveolae contain APP.

A C-terminal Degradation Product of APP Is Localized within Caveolae-rich Fractions, and the Generation of This APP Product Is Specifically Dependent upon the Expression of
Caveolin-1-The ␣-secretase-generated degradation product of APP can be detected in COS-7 cells by immunoblotting with C-terminal APP antibodies (51). To determine if ␣-secretasemediated APP degradation occurred in our experimental system, COS-7 cells were transiently transfected with HA-tagged APP and subjected to detergent-free subcellular fractionation. These fractions were then analyzed by immunoblotting with C-terminal APP antibodies (AC-1). As shown in Fig. 4, two bands confined to caveolin-rich fraction 4 were detected; the high molecular weight band represents intact APP 695 , whereas the low molecular weight band is a short fragment of APP that contains a C-terminal epitope recognized by the AC-1 antibody. As this low molecular weight band was not detected by immunoblotting with an antibody directed against the extracellular domain of APP (22C11) (Fig. 2), these results indicate that this short fragment represents a C-terminally derived APP degradation product. To obtain a more precise estimate of the molecular mass of this C-terminal APP fragment, the same sample was subjected to Tris/Tricine SDS-PAGE and immunoblotting with AC-1. A 10-kDa band was detected, supporting the hypothesis that it is derived from APP ␣-secretase processing (51). To substantiate further these observations, we subjected the samples to immunoprecipitation with three independent antibodies directed against the C-terminal region of APP (termed 369, CT15, and AC-1). These immunoprecipitates were then immunoblotted with antibodies directed against the A␤ amyloid region of APP (termed 4G8 and 6E10). 4G8 recognizes residues 18 -24 of the amyloid peptide, whereas 6E10 recognizes residues 1-17 of the amyloid peptide. As the ␣-secretase cleavage site is located at residue 17 or 18 of the amyloid peptide, these antibodies should allow us to the precisely determine the origin of this C-terminal low molecular weight product of APP. As shown in Fig. 4B, the short APP fragment that was immunoprecipitated with antibodies directed against the C terminus of APP is immunoreactive with the 4G8 antibody but not with 6E10. These data indicate that the short C-terminal APP fragment contains both the C terminus of APP and the 4G8 epitope, but it does not contain the 6E10 epitope. These studies provide additional evidence that this 10-kDa fragment represents an ␣-secretase processed product of APP (␣APPct). These results also establish that both intact APP and ␣APPct co-fractionate with caveolin-1 and are concentrated within caveolae-enriched fractions.
As ␣-secretase promotes secretion of the APP extracellular domain (APPs), we investigated the possibility that APPs are present in the media containing our transfected cells. The media from cells transfected with HA-APP were collected and subjected to immunoprecipitation with anti-A␤ peptide antibodies followed by immunoblotting with 22C11. As shown in Fig. 4C, a band corresponding to APPs was detected by immunoprecipitation with 6E10 but not with 4G8. Detection of a soluble fragment of the extracellular domain of APP that contains residues 1-17 of the A␤ amyloid peptide (the epitope of 6E10) provides additional support for ␣-secretase cleavage of APP in COS-7 cells. Taken together, these findings demonstrate that APP is correctly processed at its ␣-secretase cleavage site in COS-7 cells.  Fig. 1. Gradient fractions were collected, separated by SDS-PAGE (4 -20% linear gradient gel), and subjected to immunoblotting with AC-1 antibody which is directed against the C terminus of APP. Inset, lysates from the transfected cells were also subjected to Tris/Tricine SDS-PAGE (10 -20% linear gradient gel) and to immunoblotting with AC-1. B, the C-terminal fragment is an ␣-secretase-cleaved product of APP. COS-7 cells transfected with HA-APP were subject to immunoprecipitation with various C-terminal APP antibodies (369, CT-15, and AC-1) followed by immunoblotting with anti-A␤ antibodies (4G8 which is directed against residues 18 -40 of A␤ peptide and 6E10 which is directed against residues 1-17). C, a soluble form of the extracellular domain of APP (APPs) is shed into the culture media. The culture media of cells transfected with HA-APP were collected and subjected to immunoprecipitation with anti-A␤ antibodies (4G8 and 6E10) followed by immunoblotting with 22C11. D, APP and its C-terminal fragment are concentrated in caveolae membrane fractions. The Optiprep method was employed to confirm further the observation that both APP and its C-terminal fragment are in caveolae. Plasma membrane and caveolae membrane fractions were obtained from COS-7 cells transfected with HA-APP by the method developed by Smart et al. (41), and 4 g each of protein was loaded, separated by SDS-PAGE (4 -20% linear gradient gel), and subjected to immunoblotting with AC-1 antibody (upper panel) and caveolin antibody (middle panel). The protein concentration profile of fractionated samples is also presented (lower panel).
As an alternative approach, we employed a method developed by Smart et al. (41) to purify caveolae (Optiprep approach) to substantiate further our observation that intact full-length APP and its ␣-secretase-cleaved C-terminal fragment are localized in caveolae. We first transiently transfected HA-APP and myc-caveolin-1 into COS-7 cells and then purified caveolae by the Optiprep method. In this approach, the bulk of protein which was recovered in fractions 7-13, termed "plasma membrane fractions," was successfully separated from the "caveolae membrane fraction" (fractions 1-6) (Fig. 4D, lower panel), consistent with the method originally described by Smart et al. (41). By using this approach, full-length APP and its processed product were dramatically enriched within the caveolae membrane fraction (Fig. 4D, upper panel). As reported previously, the plasma membrane fraction did not contain caveolin, which was specifically recovered in the caveolae membrane fraction (Fig. 4D, middle panel). Importantly, the full-length APP and its processed product were not detected in these plasma membrane fractions (Fig. 4D, upper panel). These data clearly support the hypothesis that not only full-length APP but also its short product processed by ␣-secretase are present within caveolae microdomains.
These observations prompted us to investigate if caveolae play any role in the ␣-secretase processing of APP. One possibility is that the relative amount of caveolin expression could modulate whether APP undergoes ␣-secretase processing. To test this hypothesis, we (i) co-transfected COS-7 cells with HA-tagged APP and myc-tagged caveolin-1 and (ii) assessed ␣-secretase activity by measuring the amounts of ␣APPct in cell lysates and of APPs in the culture media. In addition, we performed detergent-free subcellular fractionation on these samples. As shown in Fig. 5A, two major bands corresponding to intact APP and ␣APPct that were confined to fractions 4 and 5 were detected by immunoblot analysis with C-terminal APP antibodies. The ratio ␣APPct to intact APP was increased significantly by co-expression with caveolin-1 (as compared with HA-APP expressed alone). Densitometric analysis revealed that the ratio increased from 1:1 to 5:1. In contrast, overexpression of the caveolin mutant which lacks the APP binding domain (caveolin-1-(⌬ 61-100)) did not increase ␣APPct (Fig. 5A), indicating that caveolin binding to APP promotes APP degradation.
The culture medium was then collected from these cells, immunoprecipitated with an anti-A␤-(1-28) antibody, and analyzed by Western blotting with 22C11. As shown in Fig. 5B, APPs were detected in the media, and this immunoreactivity was abolished by preincubating the anti-A␤ antibody with a synthetic peptide containing A␤ residues 1-16. In contrast, non-immune rabbit IgG did not immunoprecipitate a band corresponding to APPs. These results clearly indicate the APPs in the media are an end product of ␣-secretase-processed APP.
The amount of APPs secreted into the media from cells transiently transfected with HA-APP and myc-tagged forms of caveolin-1 was compared (Fig. 5C). Co-transfection of caveolin-1 with APP significantly increased the amount of APPs secreted into the media detected by 22C11 immunoblotting. In contrast, co-transfection with an internal deletion mutant of caveolin lacking the APP binding domain (caveolin-1-(⌬ 61-100)) failed to increase the amount of APP secreted into the media. Immunoblots with 22C11 and myc antibodies showed similar expression levels of HA-APP and caveolin among the different transfectants. These results establish that caveolin-1 overexpression promotes ␣-secretase processing of APP and that the caveolin scaffolding domain is required for this effect.
As caveolae represent a subcompartment of the plasma membrane, ␣-secretase processing should be observed on the cell surface. To investigate this possibility, we measured the amount of ␣-secretase cleavage product of APP that was shed from the cell surface by exploiting the membrane-impermeant probe, sulfo-NHS-biotin. COS-7 cells were co-transfected with HA-APP and myc-tagged caveolin-1, subjected to cell surface biotinylation, and placed in fresh media for 4 h. The media were then collected, and cell surface biotinylated molecules were specifically recovered by incubating the media with Neutravidin beads. After extensively washing the beads with buffer containing high salt and detergents, these samples were immunoblotted with 22C11. As shown in Fig. 5D (right panel), biotinylated APPs were only detected in the medium of cells co-transfected with APP and caveolin-1; in contrast, no biotinylated APPs were detected in the medium of cells transfected with APP alone (middle panel). Immunoblotting of these cell lysates confirmed that APP expression was similar in the presence or absence of caveolin transfection (left panel). These data indicate that APPs are created on the cell surface, and its production is dependent on caveolin expression, further supporting the hypothesis that caveolae organelles and the caveolin-1 protein play a pivotal role in ␣-secretase-mediated APP degradation.
If caveolin-1 overexpression greatly promoted ␣-secretasemediated degradation of APP, loss of endogenous caveolin-1 expression should block the basal production of APPs. To examine if the expression of endogenous caveolin-1 is strictly required for this processing event, we introduced caveolin-1 antisense oligonucleotides to deplete the cells of endogenous caveolin-1. The caveolin-1 antisense oligonucleotide included the first 15 bases of the 5Ј-coding sequence of the caveolin-1 gene. This antisense oligonucleotide and the corresponding sense nucleotide were introduced into COS-7 cells by transient transfection in combination with HA-tagged APP. The media were collected, and the amount of APPs produced was detected by immunoprecipitation with the anti-A␤-(1-28) antibody followed by Western blotting with 22C11. Expression of endogenous caveolin-1 was virtually undetectable after transfection with the antisense oligonucleotide indicating that the synthesis of endogenous caveolin-1 was efficiently abrogated; in contrast, the sense oligonucleotide had no effect on the expression of endogenous caveolin-1 (Fig. 6B). Co-transfection with the antisense oligonucleotide decreased the amount of APPs produced and secreted to the media by more than 80% as compared with cells transfected with HA-APP alone (Fig. 6A). Accordingly, the amount of ␣APPct produced was dramatically inhibited by co-transfection of HA-APP with the caveolin-1 antisense oligonucleotide, without affecting the production of intact APP (Fig.  6B). Importantly, co-transfection with the sense oligonucleotide had little or no effect on the production of APPs or ␣APPct (Fig. 6, A and B). DISCUSSION We have presented several independent lines of evidence that APP localizes within caveolae membranes and interacts directly with caveolin-1, a principal structural component of caveolae. More specifically, we have shown that (i) APP and caveolin-1 co-fractionate when cells are subjected to two independent detergent-free subcellular fractionation protocols previously used to purify caveolae from cultured cells; (ii) APP and caveolin-1 can form a physical complex in intact cells as shown by chemical cross-linking and co-immunoprecipitation experiments; (iii) association of APP with caveolin-1 occurs via a direct interaction between the C-terminal cytoplasmic domain of APP and the caveolin-1 scaffolding domain; (iv) cell surface caveolae contain APP; (v) the characteristic C-terminal fragment that results from APP processing by ␣-secretase was also localized within caveolae; (vi) ␣-secretase processing was sig-nificantly promoted by recombinant overexpression of caveolin-1 in intact cells, resulting in increased secretion of the soluble extracellular domain of APP; and (vii) depletion of endogenous caveolin-1 using antisense oligonucleotides prevented this cleavage event. It is therefore likely that a physical interaction between APP and caveolin-1 functionally sequesters APP within caveolae membranes and thereby compartmentalizes the ␣-secretase-mediated proteolysis of APP.
The caveolin-1 region that interacts with APP includes the caveolin scaffolding domain, a short stretch of 20 amino acids that directly participate in the recognition of multiple classes of signaling molecules. Recent evidence suggests that the caveolin scaffolding domain functions as a modular protein domain that interacts with peptide and protein ligands that contain a specific caveolin binding motif (28). APP contains a predicted caveolin binding motif within its C-terminal cytoplasmic tail. It is likely that this sequence is recognized by the caveolin scaffolding domain and functions to sequester APP within caveolae microdomains.
Cholesterol and sphingolipids are highly concentrated with caveolae domains. Recent findings indicate that transport of newly synthesized cholesterol from the endoplasmic reticulum to the plasma membrane is mediated via caveolin proteins (52,53). In addition, caveolin-1 directly binds cholesterol, and cho-FIG. 5. Overexpression of caveolin-1 promotes ␣-secretase processing of APP. A, production of the C-terminal fragment of APP is greatly enhanced by overexpression of caveolin (cav-1) but not by overexpression of caveolin-1 lacking residues 61-100 (cav-1⌬61-100). COS-7 cells were transfected with HA-APP and myc-caveolin-1 or myc-caveolin lacking residues 61-100. 48 h post-transfection, cells were harvested and subjected to detergent-free caveolae fractionation. Gradient fractions were collected, separated by SDS-PAGE (4 -20% linear gradient gel), and subjected to immunoblotting with AC-1 and anti-myc antibodies. The equivalent expression of recombinant caveolin-1 and mutant caveolin-1 was confirmed by immunoblotting with the anti-myc antibody (data not shown). B, media from cells co-transfected with HA-APP and myc-caveolin-1 were subjected to immunoprecipitation with anti-A␤-(1-28) antibody (anti-A␤) (lanes 1 and 2) or normal rabbit immunoglobulin (NRG) (lane 3) and analyzed by immunoblotting with 22C11. Prior to immunoprecipitation, the anti-A␤-(1-28) antibody was pre-absorbed with an A␤-(1-16) synthetic peptide and was then used for immunoprecipitation (lane 2). C, media from cells transfected with or without HA-APP (ϩAPP or ϪAPP) together with either mock (Ϫcav-1), caveolin (cav-1), or caveolin which lacks residues 61-100 (⌬61-100) were subjected to immunoprecipitation (IP) with an anti-A␤-(1-28) antibody or 1153 and immunoblotted with 22C11 (lower panel). Cell lysates (100 g) were also subjected to immunoblotting with 22C11 (upper panel) or anti-myc antibodies (middle panel). D, APP degradation occurs on the cell surface. Left panels, cells transfected with HA-APP alone or in combination with myc-tagged caveolin-1 were surface-labeled with the cell-impermeant probe sulfo-NHS biotin. After incubation for 0 (0h) or 4 h (4h), the media were collected and subjected to precipitation with Neutravidin beads. Precipitated proteins were to immunoblotted with 22C11. Right panel, cells transfected with HA-APP alone or in combination with myc-tagged caveolin-1 were subjected to immunoprecipitation with CT-15, followed by immunoblotting with 22C11. lesterol is required for the insertion of recombinant caveolin-1 into model lipid membranes (54,55). There are several links between cholesterol metabolism and the pathogenesis of Alzheimer's disease. Cholesterol affects APP processing by interfering with the activity of ␣-secretase (56). An isoform of a major cholesterol carrier protein, apoE, is a genetic risk factor for Alzheimer's disease (57). Thus, these links to cholesterol metabolism may simply reflect the caveolar localization of APP.
The processing of APP has been intensively studied since the APP molecule was first identified. Normal processing of APP occurs at the so-called ␣-site by ␣-secretase, which generates a non-amyloidogenic extracellular domain of APP (APPs) (3,58). This processing event makes the extracellular domain of APP containing a Kunitz-type protease inhibitor available in the extracellular space. This protease inhibitor activity is thought to prevent thrombin from injuring neurons (59,60). In addition, ␣-secretase is a target for the development of potential agents for the treatment of Alzheimer's disease, as ␣-secretase may cleave and solubilize A␤ amyloid peptide, a major insoluble component of the senile plaque. Recently, it has been reported that APPs activate microglial cells and thereby enhance the local release of neurotoxic agents (61). Therefore, APPs possess a broad range of biological effects. As our current evidence suggests that APP is processed at its ␣-site within caveolae membranes, this may help in the identification and cloning of the protein or proteins that possess ␣-secretase activity.
Several previous reports support the view that ␣-secretasemediated cleavage of APP occurs on the cell surface (62)(63)(64)(65). These observations fit well with our current findings that APP is processed at its ␣-site within caveolae, as caveolae are specialized microdomains of plasma membrane. APP requires membrane anchoring for its cleavage by ␣-secretase suggesting that the protease itself is a membrane-anchored protein (64). Interestingly, caveolae have been reported to contain a protease activity (66). The cleavage mechanism for APP may be similar to that of other proteolytic activities known to cleave a number of single membrane-spanning precursors (tumor necrosis factor and notch) at the cell surface to release their ectodomains to the media (67).
␣-Secretase cleavage can also occur intracellularly (68 -70). Therefore, a second mechanism should exist that involves an intracellular compartment that may be independent of plasma membrane caveolae. Alternatively, this phenomenon may be explained by the presence of intracellular forms of caveolae, i.e. plasmalemmal vesicles. In our experiments, we could not detect a cell surface cleavage product of APP in cells that were singly transfected with APP (Fig. 5D); in contrast, in cells co-transfected with APP and caveolin-1, APPs were produced at the cell surface. Overall cleavage of APP by ␣-secretase may be modulated by the cycle of caveolae internalization and recycling (APP could cycle between cell surface and intracellular populations of caveolae). The ␣-secretase cleavage occurs not only at Leu 17 but also at Val 18 in the amyloid peptide (3), suggesting the existence of multiple ␣-secretases. Thus, caveolae and an undefined intracellular compartment may possess different ␣-secretases that cleave APP at distinct ␣-sites.
Protein kinase C activation promotes the ␣-secretase-mediated processing of APP (71), by an unknown molecular mechanism. Since this PKC␣ is a well established component of caveolae (72) and PKC activators inhibit caveolae internalization (potocytosis) (73), it is possible that PKC activation prolongs the amount of time APP spends in cell surface caveolae and therefore in close contact with ␣-secretase activity.
APP is known to transit through clathrin-coated pits and vesicles on its way to endosomes and lysosomes (74). Recent findings suggest a number of signaling receptors that are cleared from the cell surface via clathrin-coated pits are first present within caveolae microdomains (15,21,75). In fact, internalized caveolae microdomains and clathrin-coated vesicles may be targeted to a common endosomal pool (77). This scenario would provide two distinct mechanisms for clearing signaling receptors from the cell surface. In fact, ␣APPct was reported to be enriched in lysosomes (62). It therefore is likely that APP is first present within caveolae where ␣-secretase processing takes place, whereas the remaining intact APP may be cleared from the cell surface via clathrin-coated pits and targeted to endosomes and lysosomes for proteolytic processing.
Caveolae-like membrane domains exist in neuronal cells and have been characterized by a number of independent laboratories (13,75,78). In addition, Allinquant and co-workers (13) provided preliminary evidence that axonal APP is concentrated within caveolae-like structures in neurons. Thus, it is likely that a neuron-specific form of caveolin may exist that promotes caveolae formation in neuronal cells and interacts with neuronally expressed APP. In Alzheimer's disease, caveolae dysfunction may cause reduced activity of ␣-secretase and accumulation of toxic A␤ amyloid peptide.