Axonal Amyloid Precursor Protein Expressed by Neurons in Vitro Is Present in a Membrane Fraction with Caveolae-like Properties*

,

The ␤A4 peptide, present in the extracellular amyloid plaques characteristic of Alzheimer's disease, originates from a transmembrane precursor, the amyloid precursor protein (APP). 1 Brain APP is expressed during embryogenesis and in the adult by several cell types, including neurons (for review, see Selkoe et al. (1994)).
Although the exact physiological functions of APP are poorly understood, a large body of observations supports a role in cell interactions both during development and in the adult. The suggestion that it performs an important function in development is supported by the following facts. First, the expression of reporter plasmids under the control of the APP promoter can be regulated by developmental genes of the Hox family (Violette et al., 1992). Second, APP expression increases with neuronal differentiation . Third, APP is rapidly transported down the axons to their terminals where it might be secreted (Koo et al., 1990;Morin et al., 1993;Moya et al., 1994), suggesting a synaptic function, in particular during synaptogenesis. Finally, several observations support a role for APP in neurite outgrowth, and the decreased rate of neurite elongation following APP down-regulation lends weight to the latter hypothesis (Small et al., 1994;Jin et al., 1994;Allinquant et al., 1995).
In neurons in vitro the predominant APP isoform has 695 amino acids and is found both in the somatodendritic and axonal compartments . In vivo, it is enriched at the synaptic level where its extracellular domain is released. The release of APP ectodomain requires that it be cleaved within the ␤A4 sequence and is therefore incompatible with the production of the amyloid peptide. The formation and the secretion of the amyloid peptide ␤A4 has been shown to involve reendocytosis (Koo and Squazzo, 1994).
We recently reported that, in embryonic neurons differentiating in vitro, two intracellular pools of APP coexist. A first pool is present in the entire cell (soma, dendrites, and axon) and is seen only upon strong permeabilization with detergents. A second pool, localized primarily in the axon and cell body and almost absent in the dendrites, can be observed upon mild fixation and without detergent treatment, suggesting a localization in the vicinity of the axonal membrane . In our culture conditions, the latter pool referred to as Ax-APP (for axon-specific) is not detected on nonfixed live cells, suggesting that it is not directly exposed to the extracellular medium or that it is present in a vesicular compartment that exchanges very rapidly with the plasma membrane .
Ax-APP represents only a small percentage of total APP. Interestingly, double immunostaining experiments indicate that a significant amount of Ax-APP does not colocalize with clathrin , therefore suggesting that some Ax-APP is present in a plasmalemmal compartment that differs both from the plasma membrane and from clathrincoated vesicles, and it is found in the soma and axon but not in the dendrites. It has been speculated that axons and dendrites could be the neuronal equivalents of the epithelial cell apical and basolateral domains, respectively (Dotti and Simons, 1990). The imperfect colocalization of Ax-APP and clathrin, as well as its highly polarized distribution in the axon, prompted us to examine whether Ax-APP was associated with GPI-linked molecules present in sphingolipid-rich microdomains and preferentially addressed to the apical compartment of polarized epithelia (Anderson, 1993). To examine this possibility we have compared the distribution and fractionation characteristics of Ax-APP and F3, a GPI-linked neuronal glycoprotein (Gennarini et al., 1989(Gennarini et al., , 1991. Our results confirm that Ax-APP and F3 glycoprotein are present in membranous microdomains with similar biophysical characteristics.
Madin-Darby canine kidney cells (MDCK) cells were grown to confluence on 100-mm dishes in Dulbecco's modified Eagle's medium, supplemented with 10% fetal calf serum.
Immunocytochemistry-Neurons were fixed at 5 days in vitro with paraformaldehyde (4% in PBS, pH 7.4) for 30 min at room temperature without further permeabilization . Anti-APP mAb 22C11 and anti-clathrin (Boehringer-Mannheim) were used at 5 and 0.5 g/ml, respectively. The polyclonal anti-F3 antibody (Gennarini et al., 1989) was diluted 50-fold. The appropriate fluorescein isothiocyanate-or Texas Red-conjugated secondary antibodies were from Amersham Corp. (Paris, France) or from Southern Biotechnology Associates.
Gel Electrophoresis and Immunoblotting-SDS-polyacrylamide gel electrophoresis (PAGE) and immunoblotting on cellulose or Immobilon (Millipore) were performed as described in Allinquant et al., (1994). The presence of the epitopes was revealed with alkaline phosphatase-or peroxidase-conjugated secondary antibodies (Southern Biotechnology Associates). In the latter case peroxidase was revealed using the chemiluminescence protocol according to the instructions of the manufacturer (Amersham Corp.).
Cell Extraction-Neurons (150 ϫ 10 6 ) after 5 to 7 days in vitro were rinsed twice with PBS, pH 7.4, scraped, and washed once in the same buffer. After centrifugation at 800 rpm for 5 min, cells were resuspended in 2 ml of Mes-buffered saline (MBS; 25 mM Mes, pH 6.5, 0.15 M NaCl) containing 1% Triton X-100, 1 mM dithiothreitol, and protease inhibitors (1 mM Pefablock, 1 M leupeptin, 1 M pepstatin, 0.3 M aprotinin, 1.3 M ␣ 2 -microglobulin). The extract was homogenized through a combitip (10 passages) and a 22-gauge needle (15 passages) and processed as described by Sargiacomo et al., (1993). The homogenate was adjusted to 40% sucrose and placed at the bottom of an ultracentrifuge tube. A 5-30% linear sucrose gradient in MBS without Triton X-100, was formed above the homogenate. After centrifugation at 100,000 ϫ g for 18 h at 4°C, in a TST 14-41 rotor (Kontron Instruments), 1-ml fractions were harvested from top to bottom, diluted with MBS plus 1% Triton X-100, and centrifuged at 100,000 ϫ g for 1 h at 4°C. Pellets were then resuspended in different buffers (depending on the subsequent experiment).
Phosphatidylinositol Phospholipase C Treatment-Pellets (see above) were resuspended in 100 l of MBS, pH 7.4, containing 0.5 unit/ml phosphatidylinositol phospholipase C (PI-PLC; Immunotech) and incubated for 1 h at 37°C. After centrifugation at 15,000 rpm for 1 h at 4°C in a Sigma 2K15 centrifuge, pellets and supernatants were separated by centrifugation and solubilized, and their content was analyzed by PAGE.
Confocal Laser Microscopy-Confocal laser microscopy was performed on a Molecular Dynamics 1000 system using dedicated image software. Optical sections of 0.5-m increment were scanned for double fluorescent-labeled samples with appropriate filters.
Electron Microscopy-Gradient fractions were loaded onto carbocoated grids and fixed for 10 min with 3% paraformaldehyde in PBS, pH 7.4. The grids were washed in PBS and incubated with 100 mM NH 4 Cl for 30 min, and the nonspecific sites were blocked in 0.2% PBS-gelatin for 30 min. The primary antibodies were incubated at room temperature for 2 h. After rinsing in PBS, the samples were incubated with gold-conjugated goat anti-mouse or anti-rabbit secondary antibodies diluted to 1/50 in PBS-gelatin. After rinsing five times the grids were stained with 2% uranyl acetate and examined on a Philips 400 electron microscope.

RESULTS
Colocalization of APP and F3-After 5 days in vitro, the cells were fixed with paraformaldehyde and immunostained with the anti-APP antibody without permeabilization, a procedure that gives access only to Ax-APP and does not allow revelation of the second pool of APP present in all compartments . Approximately half of the cortical neurons express Ax-APP and the staining is not uniform since, within the axons, Ax-APP-positive and -negative regions alternate. To determine the fraction of axons that coexpress Ax-APP and F3 or clathrin, cells fixed as above were doubled stained with anti-APP and anti-clathrin or anti-F3 antibodies. In three separate experiments in which approximately 1000 neurons were analyzed, we observed that 50% of the axons expressing Ax-APP also expressed F3 and/or clathrin. As illustrated in the confocal sections of Fig. 1 AB, expression of Ax-APP and F3 was often superimposed. In contrast, the colocalization of clathrin and Ax-APP was limited (Fig. 1, C and D) and not very different from that of F3 and clathrin (Fig. 1, E and F). These patterns of colocalization were specific for the axon and not observed in the cell body in which "Ax-APP" colocalized equally well with the two markers (not shown). Virtually no staining was seen in the dendrites which do not express Ax-APP . These data suggested that, at least in axons, a subfraction of APP and F3 are present in the same membranous structures.
A Small Fraction of Transmembrane APP Is Associated with Triton X-100-insoluble Light Density Complexes-GPI-anchored molecules "en route" for the cell surface are normally associated with a class of vesicles enriched with sphingolipids resulting in two distinct properties. They are resistant to nonionic detergents and their density in a sucrose gradient is lighter than that of regular plasma membranes (Sargiacomo et al., 1993). We thus investigated if a fraction of transmembrane APP could be recovered in Triton X-100-insoluble F3-containing fractions, and compared the distribution of APP and F3 on a sucrose density gradient.
In extracts from neuronal cultures at 5 days in vitro, the particulate fraction retains almost all transmembrane APP and solubilization of particulate APP with Triton X-100, releases more than 90% of the molecule, meaning that approximately 10% of total APP remains insoluble . APP present in this Triton X-100-resistant fraction can be visualized by SDS-PAGE electrophoresis, and Western blot with mAb 22C11 (Fig. 2A). As expected, significant amounts of F3 are also observed in this detergent-insoluble fraction ( Fig.  2A). Triton-insoluble complexes from neuronal cultures at 5-7 days in vitro were prepared in 40% sucrose and loaded under a linear sucrose gradient (5-30%). After centrifugation, the material was separated in three main fractions, an insoluble pellet at the bottom of the tube and two opalescent bands in the 40 and 10 -20% sucrose regions. In some cases the lighter band was split into two bands with very similar densities. Similar results were obtained with the 22C11 antibody and with antibody CT15 that specifically recognizes true APP and not its APP-like variants (Sisodia et al., 1993).
Twelve different 1-ml gradient fractions were collected, from top to bottom, diluted in Triton X-100-containing buffer and centrifuged at high speed. The resulting pellets were analyzed by PAGE and immunoblotting. As seen in the gradient of Fig.  2B, the GPI-linked F3 glycoprotein was highly enriched in low density fraction 7 (18% sucrose) with some immunoreactivity also present in fractions 3 and 4 (12% sucrose). In comparison, APP distributed equally between the clathrin-enriched fraction 12 (40% sucrose, Fig. 2, B and C) and the low density fractions 3, 4, 7, and 8 that contain F3. It is noteworthy that some clathrin was also present in fraction 7. Although the gradient of Fig. 2B is representative of most experiment, it was also found that the distributions of light APP and F3 could be more widely spread (between fractions 5 and 7, e.g. Fig. 3). Fig. 3 illustrates that treating low density fractions 5-7 with PI-PLC, an enzyme that hydrolyzes the GPI-link and thus solubilizes glypiated proteins from the membranes releases a fraction of the F3 glycoprotein. This incomplete hydrolysis probably reflects that a fraction of the molecules is not accessible to the enzyme. As expected for a trans-membrane protein, APP was not solubilized at all upon incubation with PI-PLC. F3-glycoprotein and APP were quantitatively solubilized by the addition of n-octylglucoside (not shown).
Caveolae-like Properties of the Low Density Fractions-To ensure that our protocol allowed for the isolation of vesicles of a density expected for noncoated microdomains enriched in GPI-anchored proteins (caveolae), we applied it to MDCK cells. As demonstrated in Fig. 4A, fractions 14 -7 are highly enriched in caveolin/VIP 21, a specific marker of caveolae in this cell type. Electron microscopy of these low density fractions 14 -7 from MDCK cells show a heterogenous population of membrane fragments and vesicular structures with curved shapes and stria. All these structures contain caveolin as demonstrated by immunocytochemistry (Fig. 4, B and C), therefore confirming that fractions 14 -7 prepared from MDCK cells are highly enriched in caveolae. Whole mount electron microscopy of low density fractions 5-7 from neuronal cultures shows a heterogenous population of curved shape vesicular structures (100 -500 nm), either isolated or grouped with two to three identical elements and exhibiting a striated coat (Fig. 5). The morphological aspect of these structures is very similar to that observed from MDCK or smooth muscle cell caveolae (Chang et al., 1994). Immunogold double immunocytochemistry demonstrates the presence of APP and F3 associated with these structures. Some of these "caveolae-like" vesicles can be double-labeled with the anti-F3 and anti-APP antibodies (Fig. 5, B and D), demonstrating that the two antigenes can colocalize on the same structures.

DISCUSSION
In this report we demonstrate, using immunocytochemical and biochemical approaches, that in rat forebrain neurons differentiating in vitro, a fraction of APP, enriched in the axon, is colocalized with the glypiated F3 adhesion molecule. Our results also suggest that the neurons contain a population of vesicles that share some of the properties assigned to caveolae in other cell types and that F3 and APP are present in these vesicles herein called microdomains.

FIG. 2. Presence of APP in low density membrane fractions that resist to Triton solubilization.
A, PAGE analysis and Western blotting of APP and F3 in a Triton X-100-resistant membrane fraction (15 ϫ 10 6 cells). B, high speed pellets of Triton-resistant membranes were loaded on a density sucrose gradient and each fraction from lowest (1) to highest (12) density analyzed by PAGE and Western blotting. Note that APP is retrieved from high (10 -12) and low (3, 4, and 6 -8) density fractions. The results presented here are not in contradiction with other reports demonstrating that most APP molecules are associated with clathrin-coated vesicles (Ferreira et al., 1993). Indeed, Ax-APP as defined by the fraction that is observed upon mild paraformaldehyde fixation is present only in the cell body and in the axon, and not in dendrites. Furthermore, Triton X-100 permeabilization allows the revelation of another very abundant pool of APP present in all cell compartments . From our own immunocytochemical and fractionation experiments we estimate that Ax-APP associated with the microdomains corresponds to 1-5% of total neuronal APP.
The properties of the microdomains containing Ax-APP are reminiscent of those reported for a class of vesicles called caveolae present in several cell types, but not described in neurons nor in the nervous system, due to the absence of caveolin, in neural tissue (Rothberg et al., 1992;Dupree et al., 1993;Kurzchalia et al., 1994). In spite of the absence of caveolin, the microdomains that contain Ax-APP share the following traits with bona fide caveolae.
First, Ax-APP is enriched in the axon and virtually absent in dendrites, two compartments which, according to Dotti and Simons (1990), are the equivalent, in neurons, of the MDCK apical and basolateral domains, respectively. Significant amounts of Ax-APP colocalize in light immunocytochemistry with F3, a GPI-linked glycoprotein (Gennarini et al., 1989) but not with clathrin a marker of the coated vesicle endocytic pathway. This distribution of Ax-APP is interesting in the context of several reports suggesting that glypiated proteins are enriched in caveolae and, in general, addressed to the apical compartment of polarized cells (Lisanti et al., 1988(Lisanti et al., , 1989Brown et al., 1989).
Second, APP is present in a Triton-resistant and light density membrane fraction also enriched in F3. The comigration of APP and F3 in fractions 3, 4, 7, and 8 (Fig. 2), added to the immunocytochemical data discussed in the preceding paragraph, and the APP-F3 double labeling in electron microscopy, demonstrates that these fractions are enriched with Ax-APP. Indeed, APP was also retrieved in heavier fractions that contain the bulk of clathrin, suggesting that the protocol of extraction before sucrose separation spares a significant quantity of coated vesicles or of membranes derived from coated vesicles. The presence of small quantities of clathrin in fractions 7 and 8 could reflect the actual presence of clathrin-coated vesicles in the light fractions, and possibly the presence of F3 in such vesicles. Other glypiated molecules have already been observed in coated vesicles (Mayor et al., 1994;Shyng et al., 1994). At last, fusion between the two endocytic pathways leading to multivesicular bodies could explain the presence of some clathrin in the enriched-F3 fractions. (Turek et al., 1993).
Third, when the same protocol of fractionation was applied to MDCK cells, the membranous structures resistant to Triton and present in the lighter fractions are labeled with anticaveolin. In the case of MDCK-derived fractions all structures examined by electron microscopy present traits characteristic of caveolae, in particular striated figures and caveolin. In the case of neuron-derived fractions, observations in electron microscopy demonstrate a colocalization of APP and F3 and a striated coat identical to that observed in MDCK or smooth muscle (Chang et al., 1994) caveolae.
The possibility that Ax-APP could be localized in structures resembling caveolae raises the possibility that, in analogy with epithelial cells (Smart et al. 1994a), protein kinase C activity might affect the subcellular localization of APP. We did not find that protein kinase C activators or inhibitors modify the percentage of axons with strong Ax-APP staining or the amount of Ax-APP in the Triton-resistant light density fractions (not shown). In this context it should be noted that an absence of effect of protein kinase C activity on caveolea endocytosis has been observed by Parton et al. (1994).
Taken together these data infer that Ax-APP is present in vesicular microdomains that share several properties with caveolae, with the notable exception that they lack caveolin. Whether the vesicular structure can be maintained in the absence of caveolin, a possibility suggested by Smart et al. (1994b) and/or is permitted by the presence of other nonidentified polypeptides has not been investigated. In fact, similar microdomains devoid of caveolin have recently been described in lymphocyte cell lines and in neuroblastoma cells (Fra et al., 1994;Gorodinsky et Harris, 1995) and in membrane fractions from mouse cerebellum (Olive et al., 1995).
The identification of a transmembrane molecule, such as APP, in a compartment enriched in GPI-linked molecules raises the possiblity that Ax-APP could serve as a signaling intermediate following the interaction of the glypiated proteins with their ligands to deliver an intracytoplasmic signal. In the context of neurite elongation, it is interesting that F3, NCAM-120, and Thy-1, three GPI-anchored glycoproteins present at the axonal surface, can increase (F3, NCAM) or decrease (Thy-1) the rate of axonal elongation (Durbec et al., 1992;Doherty et al., 1989, Tiveron et al., 1992. It is also noteworthy that APP can be found associated with G o , a major trimeric GTP-binding protein (Nishimoto et al., 1993;Okamoto et al., 1995) involved in neurite elongation (Brabet et al., 1990).
In a recent report, we have demonstrated that APP downregulation decreases the rate of elongation of both axons and dendrites (Allinquant et al., 1995). Since Ax-APP is almost absent in dendrites, its implication in axonal elongation would be another indication that dendritic and axonal elongation are not regulated by the same sequence of events (reviewed in Lafont and Prochiantz (1994)). A specific role of Ax-APP in the formation and the maturation of the axon or of its terminal, during synaptogenesis, is in agreement with the fact that APP is transported rapidly in the axon (Koo et al., 1990;Morin et al., 1993;Moya et al., 1994) and that Ax-APP is significantly increased upon calcium influx .
Finally, the presence of APP in microdomains is also interesting in terms of our understanding of its processing. APP contains a reendocytosis signal compatible with reinternalization (Chen et al., 1990;Haass et al., 1992), an event associated with the production and release of the ␤A4 peptide (Koo and Squazzo, 1994). In cells where caveolae have been observed the molecules associated with these structures can be delivered to multivesicular bodies and to lysozomes (Turek et al., 1993;Parton et al., 1994;Schnitzer et al., 1994). Therefore, and as suggested by a recent report from Amaratunga and Fine (1995), the presence in the axon of microdomains that contain Ax-APP and share distinct properties with caveolae might indicate the existence of an axon-specific pathway for the production and release of amyloidogenic ␤A4 peptide.