Alzheimer amyloid protein precursor is localized in nerve terminal preparations to Rab5-containing vesicular organelles distinct from those implicated in the synaptic vesicle pathway.

In order to localize amyloid protein precursor (APP) in nerve terminals, we have immunoisolated vesicular organelles from nerve terminal preparations using antibodies to Rab5 and synaptophysin. These immunoisolates were then analyzed by electron microscopy and by immunoblotting. The synaptophysin immunoisolates represented a nearly homogeneous population of small synaptic vesicles, with less than 10% contamination by other organelles, and very little APP. In contrast, Rab5 immunoisolates contained, in addition to small synaptic vesicles, substantial numbers of large uni- and bilamellar vesicles and high levels of APP. Thus, it appears that nerve terminal APP is contained predominantly in large vesicular organelles, distinct from synaptic vesicles and from the synaptic vesicle recycling pathway.

In order to localize amyloid protein precursor (APP) in nerve terminals, we have immunoisolated vesicular organelles from nerve terminal preparations using antibodies to Rab5 and synaptophysin. These immunoisolates were then analyzed by electron microscopy and by immunoblotting. The synaptophysin immunoisolates represented a nearly homogeneous population of small synaptic vesicles, with less than 10% contamination by other organelles, and very little APP. In contrast, Rab5 immunoisolates contained, in addition to small synaptic vesicles, substantial numbers of large uni-and bilamellar vesicles and high levels of APP. Thus, it appears that nerve terminal APP is contained predominantly in large vesicular organelles, distinct from synaptic vesicles and from the synaptic vesicle recycling pathway.
Amyloid protein precursor (APP) 1 is a type I membranespanning glycoprotein which is ubiquitously expressed in mammalian cells (1). Proteolytic processing of APP results in the generation of an ϳ40-residue amyloid (A␤) fragment which accumulates in the brains of individuals with Alzheimer's disease (AD). A central role for A␤ in the pathogenesis of AD is indicated by the discovery that various mutations within or flanking the A␤ region of APP cosegregate with affected status in individuals from several families with autosomal dominant AD. For this reason, the localization, trafficking, and processing of APP has generated great interest.
Immunocytochemical studies of APP in cultured cells and in brain tissue reveal that a predominant fraction of APP is localized to the endoplasmic reticulum and the Golgi apparatus (2,3). The localization of APP to biosynthetic organelles can be explained in part by the very high rate of synthesis and turnover of this protein. In axonal and synaptic compartments of brain tissue, APP is localized to large vesicular structures (2,3). APP is also found on the surface of cultured cells, from where it can be internalized and converted to A␤ (4 -7). APP is transported by fast axonal transport in central and peripheral neurons (8,9). Soluble APP (APP s ) can be released at synapses, 2 although it is not known whether the soluble fragments are generated at the synapse. Pools of unprocessed APP are also transported retrogradely from axonal or synaptic compartments to neuronal cell bodies and dendrites (10,11). It is not known which membrane trafficking pathways are involved in anterograde and retrograde transport of APP in neurons or whether APP-containing vesicles are related to the recycling pathway of synaptic vesicles within the nerve terminal.
In an attempt to study the nature of the APP-containing vesicles in nerve terminals, we have isolated APP-containing membranes from synaptosomes of rat forebrain and from PC12 cells. To differentiate vesicles participating in the synaptic vesicle pathway from other endocytic vesicles, organelles were immunoisolated using immobilized monoclonal antibodies directed against either synaptophysin, synaptobrevin, or the endosomal GTPase, Rab5. Synaptophysin and synaptobrevin are found in synaptic vesicle membranes (12)(13)(14)(15)32). Rab5 is a resident of early endosomes that is ubiquitously expressed in all endosomes, irrespective of the nature of the endocytic pathway (16 -20). Our data suggest that APP resides on endocytic trafficking organelles which are clearly distinct from organelles involved in synaptic vesicle recycling, suggesting a hitherto unknown trafficking pathway in nerve terminals of fully differentiated neurons.
Preparation of Immunobeads-Immunobeads were prepared as described (19). For coupling to beads, IgG was purified from ascites using Protein G-Sepharose (Pharmacia Biotech Inc.). The purified IgG was dialyzed for 3 days against 150 mM NaCl with 7 changes. Following dialysis, antibodies were centrifuged at 10,000 ϫ g for 15 min, and the supernatant was used for coupling to Eupergit C1Z beads (Rohm Pharma, Darmstadt, Germany) as described (19). Beads were tested for antibody coupling by SDS-PAGE.
Immunoisolation of Vesicular Organelles from Nerve Terminal Preparations-Nerve terminal preparations (synaptosomes) were purified as described by Nicholls (24). Purified synaptosomes were pelleted, resuspended in a minimal volume of 140 mM NaCl, 10 mM glucose, 5 mM KCl, 1 mM MgCl 2 , 1.2 mM Na 2 HPO 4 , 20 mM HEPES, pH 7.4, and hypotonically lysed by diluting 10-fold in ice-cold H 2 O and homogenizing at 2,000 rpm, using 8 strokes in a glass/Teflon homogenizer. HEPES and NaCl were added to final concentrations of 10 mM and 100 mM, respectively, and the lysed synaptosomes were spun for 15 min at * This work was supported by United States Public Health Service Grant AG09464 (to P. G.) and by a D. Collen and a Phillips Fellowship of the Belgian American Educational Foundation (to W. G. A.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
18,700 ϫ g to separate the synaptic vesicular organelles from synaptosomal membrane debris. The supernatant from this fraction was used for the immunoisolation of vesicular organelles. Immunoisolation was conducted as described previously (19). Briefly, 2 ml of the supernatant from the nerve terminal preparation were incubated with 10-l bed volume of immunobeads for 45 min at 4°C under slow rotation. Beads with bound material were pelleted for 10 min at 360 ϫ g and washed 3 times with 0.32 M sucrose, 10 mM HEPES, pH 7.4. Bead-bound organelles were then analyzed by immunoblotting (see below) and/or electron microscopy (25).
Immunoisolation of Vesicular Organelles from PC12 Cells-Immunoisolation of vesicles from PC12 cells was carried out as described (22). Briefly, cells were homogenized by passing 20 times through a ball bearing cell cracker with a 0.0006-inch clearance. Homogenates were centrifuged for 20 min at 10,000 ϫ g, and the resultant supernatants were incubated with anti-synaptophysin or anti-Rab5 immunobeads for 1 h at 4°C. The beads were sedimented and washed as described above, and bound material was analyzed by SDS-PAGE followed by immunoblotting.
Semiquantitative and Quantitative Immunoblotting-For immunoblot analysis, washed beads were resuspended in Laemmli loading buffer, boiled for 5 min, and spun down. The supernatant was subjected to SDS-PAGE followed by immunoblotting. Membranes were probed for synaptophysin with antibody C1.7.2 and for APP with antibodies 369 or 3129. Detection was carried out using peroxidase-labeled secondary antibodies and enhanced chemiluminescence (ECL, Amersham) or with radioiodinated Protein A followed by PhosphorImager (Molecular Dynamics, Mountain View, CA) autoradiography.

RESULTS
Previous research has shown that APP is transported to the nerve terminal, and that full-length APP can be transported retrogradely from the nerve terminal (8,10,11,26). The relationship of these pathways for the transport of APP to the trafficking pathways for small synaptic vesicles was investigated in the present study. Immunoblotting of a conventionally purified (27) small synaptic vesicle fraction showed that there was a small amount, but no enrichment, of APP immunoreactivity (Fig. 1A, Ves). In order to localize APP and synaptophysin in this vesicle preparation, immunolabeling of frozen ultrathin sections was carried out using either anti-synaptophysin (c1.7.2) (Fig. 1B) or anti-APP (369) (Fig. 1C) antibodies. Immunolabeling with anti-synaptophysin antibodies demonstrated that the vast majority of the small synaptic vesicles contained synaptophysin, as expected. In contrast, only very few structures, which were typically larger than small synaptic vesicles, contained APP (Fig. 1C, arrowheads). The paucity of profiles which were immunoreactive for APP is consistent with the results of immunoblotting and presumably represents trace amounts of contaminating vesicles of unknown origin. Thus, these results indicate that APP is virtually absent from small synaptic vesicles.
To further characterize APP-containing organelles in the synapse, we chose a two-step procedure to isolate synaptic organelles. First, synaptosomes were prepared using a combination of differential centrifugation and Ficoll density gradient centrifugation. This procedure yields a fraction that is highly enriched in nerve terminals, with only low levels of contamination by soma-derived organelles, myelin, or mitochondria (28,29). Second, these purified nerve terminals were lysed by osmotic shock to release internal organelles, followed by immunoisolation of organelles using methacrylate beads coated with antibodies directed against synaptophysin, synaptobrevin II, or Rab5.
Comparison of vesicular organelles immunoisolated with Rab5 and synaptophysin antibodies revealed significant differences. The synaptophysin immunoisolates contained predominantly (Ͼ90%) small synaptic vesicles (typically Ͻ60 nm in diameter, Fig. 2A). Rab5 immunoisolates contained, in addition to small synaptic vesicles (Fig. 2B), a substantial number of other distinct vesicles, including large unilamellar vesicles ( Fig. 2, C and D), large bilamellar vesicles (Fig. 2, E and F), and multivesicular bodies (Fig. 2, G and H). It should be noted that the distinction between unilamellar and bilamellar vesicles was not always clear-cut (e.g. Fig. 2, C and D); in these cases, the profiles were counted as unilamellar. There was a 4-fold increase in the proportion of large unilamellar vesicles and a 2-fold increase in the proportion of large bilamellar vesicles, in Rab5 immunoisolates relative to synaptophysin immunoisolates (Fig. 2, lower panel). Whereas the small synaptic vesicles in both synaptophysin and Rab5 immunoisolates were heavily immunoreactive with anti-synaptophysin as shown by immunogold labeling of immunoisolates (Fig. 2, A and B), the large unilamellar vesicles, bilamellar vesicles, and multivesicular bodies showed negligible immunoreactivity (data not shown). We were unable to obtain satisfactory immunolabeling of the Rab5 immunoisolates with 369 due to nonspecific adsorption of antibody to the beads.
The levels of APP were determined in vesicular organelles immunoisolated from nerve terminal preparations. Since the immunoisolates contained significant amounts of immunoglobulins derived from the isolation procedure, total protein could not be used as a basis for comparing the various preparations. Instead, we used the levels of synaptophysin for comparison. Synaptophysin is found in small synaptic vesicles in synaptophysin, synaptobrevin, and Rab5 immunoisolates (e.g. Fig. 2, A and B). Since the frequency of small synaptic vesicles in the various types of immunoisolates was so high (78 -91%; Fig. 2, lower panel), it enabled us to use synaptophysin levels as the basis for comparison. Samples of each immunoisolate, containing equivalent amounts of synaptophysin, were subjected to immunoblotting with an antibody (369) raised against a peptide corresponding to the cytoplasmic domain of APP or an FIG. 1. APP is not enriched in conventionally purified synaptic vesicles. Synaptic vesicles were purified using conventional techniques (27). A, 50 g of each fraction were resolved by 6% SDS-PAGE; immunoblotting with antibody 369 was carried out using ECL for detection. B and C, the purified vesicles (Ves) were frozen, sectioned, and reacted with monoclonal anti-synaptophysin antibodies (B) or rabbit anti-APP antibodies (369) (C), followed by gold-labeled goat-antimouse and goat-anti-rabbit antibody, respectively. Subsequently, grids were lightly stained with uranyl acetate to visualize the tissue (gray regions). H, homogenate; S 1 , 800 ϫ g supernatant; P 2 , 9,200 ϫ g pellet (crude synaptosomes); S 2 , 9,200 ϫ g supernatant; P 2 Ј, repelleted P 2 ; S 2 Ј, supernatant from repelleted P 2 ; L, lysed P 2 Ј; S 3 , 100,000 ϫ g supernatant of S 2 ; P 3 , 100,000 ϫ g pellet of S 2 (microsomal pellet); LS 1 , 25,000 ϫ g supernatant of lysed P 2 Ј (synaptosomal cytoplasm and small membranes); LP 1 , 25,000 ϫ g pellet of lysed P 2 Ј (synaptic densities, mitochondria, etc.); LS 2 , 165,000 ϫ g supernatant of LS 1 (soluble proteins); LP 2 , 165,000 ϫ g pellet of LS 1 (small membranes, vesicles); SG 1 , first (lightest) fraction from sucrose gradient; SG 2 , second fraction from sucrose gradient; SG 3 , third fraction from sucrose gradient; SG 4 , pellet from sucrose gradient; Mem, first peak from controlled pore glass column (membrane fragments); Ves, second peak from controlled pore glass column (small synaptic vesicles). antibody (3129) raised against a peptide corresponding to the A␤ domain of APP. With either antibody, high levels of APP were observed in the Rab5 immunoisolates, but not in the synaptobrevin or synaptophysin immunoisolates (Fig. 3A). Quantitative immunoblotting for APP and synaptophysin indicated that the APP/synaptophysin ratio was more than 10-fold higher in Rab5 immunoisolates than in synaptophysin immunoisolates (Fig. 3B). Control immunoisolates, prepared using irrelevant antibodies or antibody-free beads, contained levels of APP and synaptophysin that were below the levels of detection of the assay system (data not shown).
Lysed nerve terminal preparations were used to estimate the proportion of total APP which could be depleted by the Rab5 immunobeads (Fig. 3C). For this purpose, samples of lysates were incubated with varying amounts of immobilized anti-Rab5 antibodies, followed by centrifugation and immunoblotting. With the highest amount of anti-Rab5 immunobeads used, approximately 70% of the total APP could be immunodepleted. The ability of Rab5 to dissociate from organelles (19) might decrease the proportion of APP-containing organelles recovered in Rab5 immunoisolates, either because of organelles lacking Rab5 and/or because of excessive free Rab5 competing for the available antibody. It is therefore likely that the 70% figure underestimates the amount of APP associated with Rab5-containing vesicular organelles in intact tissue.
Rab5 immunoisolates from the rat pheochromocytoma (PC12) cell line were dramatically enriched in APP, compared with synaptophysin immunoisolates (Fig. 3D). These results support the conclusion that the vesicular organelles present in Rab5 immunoisolates from nerve terminal preparations were of neuronal origin. DISCUSSION Previous studies in brain tissue using immunoelectron microscopy have shown that APP resides in large vacuolar structures in axonal and synaptic compartments (2,3). However, the nature of these APP-containing vesicles was unknown. In the present study, we have characterized APP-containing vesicular organelles obtained by immunoisolation from purified synaptosomes of rat brain. The use of synaptosomal preparations made it possible to study the distribution of APP in organelles derived from nerve terminals without significant contamination by trafficking organelles from other sources. Immunoisolations were carried out using immobilized monoclonal antibodies directed against either the synaptic vesicle proteins synaptophysin or synaptobrevin or against the endosomal GTPase, Rab5. Synaptophysin and synaptobrevin are membrane proteins specific for synaptic vesicles and are thus expected to reside in all organelles participating in the pathways of synaptic vesicle recycling and regulated exocytosis. Rab5 is a resident of early endosomes of all endocytic pathways studied to date. We found that APP was highly enriched in Rab5- containing vesicles, but virtually absent from synaptophysinor synaptobrevin-containing vesicles. Based on these results, we hypothesize that Rab5 immunoisolates contain at least two classes of organelles. One consists of small synaptic vesicles, which contain synaptophysin but not APP. The second consists of a previously unidentified class of endocytic trafficking organelles, which contain APP but little or no synaptophysin. This would account for the high levels of APP in Rab5 immunoisolates as compared with that in synaptophysin immunoisolates (Fig. 3B), as well as for the intense synaptophysin staining (Fig. 1B) and virtual absence of APP staining (Fig. 1C) of conventionally purified small synaptic vesicles.
Several studies suggest that synaptic vesicles arise after endocytosis from the presynaptic membrane (19,30,31). The apparent exclusion of APP from synaptic vesicles implies that APP resides on an organelle which is part of a separate endocytic pathway. The fact that most of the APP in the synapse is associated with a Rab5-positive compartment indicates that APP spends most of its time in the endocytic rather than the exocytic limb of its trafficking pathway. This means that APP, which reaches the nerve terminal by fast anterograde transport (8,9), is transiently exposed to the synaptic plasma membrane surface (7), reinternalized, and targeted to organelles distinct from membranes involved in synaptic vesicle recycling.
A similar example of specialized sorting in nerve terminals is provided by the polymeric immunoglobulin receptor (pIgR), which mediates transport of polymeric IgA and IgM across epithelial surfaces (33,34). When expressed in differentiated PC12 cells (31) and primary neuronal cultures (35), pIgR is primarily sorted to neuritic or axonal processes, where it is found in specialized endosomes, but is excluded from purified synaptic vesicle-like structures (31). In transfected neurons, pIgR undergoes transcytosis from the somatodendritic compartment to the axonal compartment (35). Full-length APP which, unlike the pIgR, is an endogenous neuronal protein, also undergoes transcytosis, albeit from the axonal to the somatodendritic compartment (7,11). Thus, APP might have a similar role as a transcytotic carrier of an as yet unidentified ligand which would bind to APP on the presynaptic terminal and/or axonal membrane. The APP-containing organelles which we have isolated in the present study would represent the earliest stages in this transport pathway, since they still contain Rab5, which is only transiently associated with membranes (16 -20) and is, therefore, a stage-specific marker.
A more complete biochemical characterization of these Rab5/ APP vesicles would be greatly facilitated by the development of high affinity monoclonal antibodies directed against the cytoplasmic domain of APP, which would allow the direct immunoisolation of APP-containing vesicular organelles. This would also enable the study of the fate of the APP carrier vesicles during their retrograde or transcytotic transit.
In summary, we have biochemically and ultrastucturally characterized APP-containing vesicles in rat brain synaptosomal preparations. These vesicles contain the endosomal GTPase Rab5 and only very low levels of synaptophysin. We therefore conclude that these APP carriers are derived from an endocytic pathway distinct from that involved in synaptic ves-icle recycling. The presence of APP in synaptic terminals, in proximity to synaptic vesicle-rich areas, may indicate a relation between synaptic activity and APP function. Since most of the A␤ in cultured cells is produced via an endocytic pathway (5), the Rab5/APP-containing endocytic organelles described here may play a crucial role in A␤ formation and deposition in the brain.