|Home | About | Journals | Submit | Contact Us | Français|
The amyloid-β precursor protein (APP), a transmembrane protein that undergoes proteolytic cleavage into defined fragments, has been implicated in axonal transport. The proposed role of APP as vesicle receptor for the microtubule motor, kinesin-1 has relevance for the pathogenesis of Alzheimer's disease. Nevertheless, this function, which relies on the transport to the cell periphery of full-length APP rather than its cleavage fragments, remains controversial. Other proposed functions of APP, such as regulating transcription, neurogenesis, cell movement, or neurite growth also rely on APP's presence as full-length protein at the cell surface, implying that APP cleavage occurs after its transport to the cell periphery. To test this hypothesis, we mapped the localization of various APP epitopes in neurons in culture and in the mouse brain. Surprisingly, epitopes from the amino-terminal, carboxy-terminal, and central (amyloid-β; Aβ) domain of APP each showed a distinct distribution throughout the cell, and rarely colocalized. Within neurites, these epitopes were localized to distinct transport vesicles that associated with different sets of microtubules and, occasionally, actin filaments. Carboxy-terminal APP fragments were preferentially transported into neurites as phosphorylated forms, and - unlike the amino-terminal and Aβ fragments - entered the lamellipodium and filopodia of growth cones, and concentrated in regions of growth cone turning and advancement. We conclude that, under normal conditions, the proteolytic cleavage of APP largely occurs prior to its sorting into axonal transport vesicles, and the cleaved fragments segregate into separate vesicle populations that reach different destinations, and thus have different functions.
Alzheimer's disease (AD), the prevalent neurodegenerative disorder of old-aged humans, has a complex symptomatology that includes deterioration of cognitive function and progressive memory loss. Although AD has been associated with the presence of neuritic plaques and neurofibrilary tangles in specific brain regions, this symptomatology is likely caused by synaptic dysfunction and neuronal loss (Haass and Selkoe, 2007). What triggers this neuronal pathology is still not known.
An interesting idea is that neuronal degeneration in AD may be caused by abnormal axonal transport (Kamal et al., 2001; Roy et al., 2005). It was proposed that the transmembrane protein, amyloid-β precursor protein (APP), from which the amyloid-β peptide (Aβ) is derived by proteolysis, plays an active role in transport, by anchoring the plus-end motor kinesin-1 to vesicles that contain APP, the APP processing machinery, and other cargo proteins (Kamal et al., 2000; Kamal et al., 2001). In this scenario, a delayed transport may result in premature cleavage of APP into fragments, followed by release of kinesin-1 from the vesicle and early termination of transport. This situation would favor aggregation of Aβ within the neurites with detrimental consequences on neuronal function and survival (Kamal et al., 2001).
To function as kinesin-1 receptor, APP should travel within neurites as an intact protein, capable to anchor the motor to the vesicle. Hence, in normal conditions, a significant fraction of APP should be present within the neuronal processes as full-length protein rather than cleaved polypeptides (Fig. 1A). Other proposed functions of APP, such as regulating transcription (Cao and Sudhof, 2001), neurogenesis (Ma et al., 2008), cell movement (Sabo et al., 2001), or neurite growth (Sabo et al., 2003) also rely on APP's presence as full-length protein at the cell surface, implying that APP cleavage occurs after its transport to the cell periphery. This hypothesis was never thoroughly tested for the endogenous APP, with few studies addressing APP transport from the point of view of its proteolytic processing (Buxbaum et al., 1998; Kamal et al., 2001; Lazarov et al., 2002; Goldsbury et al., 2006). This problem can be best explored with immunocytochemistry in neurons expressing normal levels of APP (Muresan and Muresan, 2005b), as exogenous expression of tagged APP (Kaether et al., 2000; Stamer et al., 2002; Goldsbury et al., 2006) at above-physiological APP levels can perturb its complex metabolism.
To study the potential transport into neurites of APP fragments, as opposed to transport of full-length APP, we mapped the localization of epitopes from different APP regions in neurons in culture and in situ, and found that they distribute largely to non-overlapping vesicle populations within the neurites. We then showed that phosphorylated carboxy-terminal fragments of APP, but not other APP-derived polypeptides, enter the lamellipodium and filopodia of growth cones (where they co-distribute with actin filaments), and become concentrated in regions of growth cone turning and advancement. These results suggest that APP is proteolytically processed prior to delivery into axons, and that the resulting cleavage products are sorted to distinct vesicle populations that are independently transported to different destinations.
The primary antibodies used in this study are: rabbit anti-APP (#2452, affinity-purified; Cell Signaling Technology, Beverly, MA) raised against a synthetic peptide corresponding to residues surrounding Thr668 of human APP695; rabbit anti-APP (AB5352, raised to a nine amino acid peptide from APPs carboxy-terminus; Chemicon, Temecula, CA); rabbit anti-APP (recognizing residues 676-695 of APP695; C9, affinity-purified) (Kimberly et al., 2005); rabbit anti-APP, raised against a 22 amino acid synthetic peptide derived from the carboxy-terminus of APP [CT695, affinity-purified; Invitrogen (Zymed), Carlsbad, CA]; mouse anti-human Aβ(4G8, purified IgG; reacts with an epitope conserved in rodents) and rabbit anti-rodent Aβ (Signet, purified IgG, Dedham, MA); rabbit anti-β-amyloid 1-40 (AB5074P, affinity-purified) and anti-β-amyloid 1-42 (AB5078P, affinity-purified) (Chemicon; these antibodies recognize the free carboxy-terminal end of Aβ); mouse anti-APP (MAB348, clone 22C11, purified IgG recognizing an amino-terminal epitope, amino acids 66-81; Chemicon); rabbit anti-APP, N-terminal (amino acids 46-60; A8967, IgG fraction; Sigma, St. Louis, MO); mouse anti-Alz 90 (MAB349; purified immunoglobulin), raised against a synthetic peptide corresponding to amino-acids 511-608 of APP pre A4695 (Chemicon). A rabbit antibody to phosphorylated APP [pAPP; #44-336Z; raised against a phospho-peptide (containing phospho-threonine) from the region surrounding Thr668 of human APP695; affinity-purified and negatively preadsorbed using a corresponding, non-phosphorylated peptide; BioSource International, Camarillo, CA] reacts with phosphorylated, but not non-phosphorylated forms of APP (Muresan and Muresan, 2005a, b). A mouse antibody to phospho-(Thr) MAPK/CDK substrate (#2321) that detects phosphothreonine only when followed by proline was from Cell Signaling Technology. A rabbit antibody to calsyntenin-1 (CLSTN1; alcadein-α) was from Protein Tech Group, Inc., Chicago, IL. Mouse (B-7) and rabbit (M-300) anti-JIP-1 antibodies were from Santa Cruz Biotechnology, CA. Microtubules were stained with mouse antibodies to α-tubulin (B-5-1-1) or to acetylated tubulin (6-11B-1) (Sigma). Actin filaments were stained with a fluorescein-phalloidin conjugate (Molecular Probes, Eugene, OR).
Mouse CNS-derived, CAD cells (Qi et al., 1997), were grown in 1:1 F12:DME medium, containing 8% fetal bovine serum and penicillin/streptomycin. Differentiation was induced by culturing without serum (Qi et al., 1997). Embryonic day 16.5 mouse cortical neurons were grown in Neurobasal Medium with B-27 supplement, L-glutamine, and penicillin/streptomycin for 5 days.
APP-YFP (Kaether et al., 2000), APP-Thr668Glu (Muresan and Muresan, 2005b), or an EGFP-tagged carboxy-terminal fragment of APP (C58-GFP) (Kinoshita et al., 2002) were transfected into CAD cells using FuGene 6 (Roche Diagnostics).
Brain tissue was collected from male Sprague-Dawley rats (killed by asphyxiation with CO2), placed in ice-cold phosphate-buffered saline, pH 7.4, and homogenized in HEPES-acetate buffer containing 1% NP-40 (Muresan and Muresan, 2005b). CAD cells were washed with phosphate-buffered saline, and extracted either in HEPES-acetate buffer containing 2% TritonX-100, or in Tris-HCl buffer containing 1% NP-40 (Muresan and Muresan, 2005b). The extracts were analyzed for the presence of APP by Western blotting (using 10% SDS-PAGE gels). Antibody incubations were done in the presence or absence of competitor polypeptides at 100:1 molar excess over IgG.
CAD cells and primary cortical neurons were fixed (20 min in phosphate-buffered saline containing 4% formaldehyde and 4% sucrose) (Goslin and Banker, 1991), then permeabilized (0.3% Triton X-100, 20 min at 20°C), and processed for single or double antigen labeling as described (Muresan and Muresan, 2005b). In some experiments, cells were fixed in 4% formaldehyde alone (omitting sucrose); in others, fixation (4% formaldehyde) and permeabilization (0.1% Triton X-100) was done simultaneously. Occasionally, cells were fixed and permeabilized for 3-5 min with cold methanol. Secondary antibodies coupled to Alexa dyes (488 and 594) were from Molecular Probes. When required, double labeling experiments with two primary antibodies raised in rabbit were done by using fluorescently labeled, monovalent Fab fragments of anti-rabbit IgG (Jackson Immunoresearch Laboratories, West Grove, PA) (Li et al., 2002; Muresan and Muresan, 2005b). Transfected APPYFP was detected via the YFP fluorescence. For in situ localization of APP, cryosections were cut from fixed (4% paraformaldehyde, 4°C, overnight), cryoprotected (30% sucrose, 4°C, overnight), and OCT compound-embedded brains from 18-month old R1.40 mice (Yang et al., 2006).
Digital images were obtained with an Olympus IX81 microscope (20X, 40X, and 100X objectives; Tokyo, Japan) equipped with Semrock, Inc. (Rochester, NY) filters and cooled CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan), and collected using Image-Pro Plus software (Media Cybernetics, Silver Springs, MD). Images were processed for contrast and brightness with Adobe Photoshop. The distribution of fluorescent particles along neurites was analyzed in thresholded, inverted, grayscale images, using the “plot profile” function of the NIH ImageJ 1.39u software (Muresan and Muresan, 2007).
For colocalization studies of two epitopes viewed in different channels, alignment of the two colors was confirmed through staining of test samples probed with a mouse primary antibody (e.g., anti-JIP-1) followed by a mixture of anti-mouse secondary antibodies tagged with Alexa 488 and Alexa 594. As shown in supplemental Fig. 1U-W, there was perfect colocalization of the particle distributions detected in the two channels. The extent of colocalization of epitopes from different APP regions was estimated by quantifying the percent coincidence of fluorescent particles between two channels. Briefly, pairs of inverted, grayscale images were thresholded at the mean intensity (determined in Adobe Photoshop CS2), to eliminate most of the diffuse, non-particulate labeling. Total number of fluorescent particles in each image of the pair, and the number of particles common to both images, were counted. This procedure provided a better quantitation of particle colocalization than the percentage of pixel overlap between the two channels.
The percent distribution of individual APP epitopes among different neuronal compartments (i.e., soma, neurites, and neurite terminals) was determined by quantifying the distribution of fluorescence intensity within each cell using Image J software. Surface plot analysis and conversion to mask of fluorescent particle distribution was done using Image J software. Statistical analysis was done using a two sample t test for the two-tailed hypothesis (Zar, 1999). For each experimental condition, data were derived from at least three separate experiments.
Three types of experiments were done to control for non-specific binding. 1) Non-specific binding of secondary antibodies was assessed in experiments that omitted the primary antibodies, where images have been acquired and processed as in the presence of the primary antibodies. 2) Specificity of antibodies for their cognate epitopes was verified in control experiments that used incubation of the specimens in the presence of competing polypeptide used at 100:1 molar excess over IgG (Muresan and Muresan, 2005a, b). We used the following competing polypeptides: Aβ40 and Aβ42 peptides, HPLC purity (AnaSpec, Inc., San Jose, CA; peptides were solubilized according to manufacturer's instructions); a polypeptide that encompassed a 12-amino acid region centered on Thr668 in the APP cytoplasmic domain (termed here “short polypeptide”); and a biotinylated polypeptide encompassing the entire APP cytoplasmic domain (termed here “long polypeptide”). 3) Antibody 22C11 was pre-adsorbed on a membrane that contained the APP region of an overloaded transfer of rat brain lyzate (similar to the blot in Fig. 2D) to remove the anti-APP immunoreactive species from the IgG fraction, or preadsorbed on a membrane that contained transferred BSA. As seen in Fig. 2D, antibody 22C11 shows no non-specific binding even in Western blots that were overloaded with rat brain extract.
The specificity of the anti-pAPP antibody for the phosphorylated, but not non-phosphorylated APP epitope was verified in Western blots of in vitro phosphorylated versus non-phosphorylated GST fusions with the APP cytoplasmic domain. In vitro phosphorylation of the fusion proteins was done with bacterially expressed GST-cyclin-dependent kinase 5 (Cdk5) and GST-p25 (an activator of Cdk5) (Iijima et al., 2000). Additional specificity control experiments showing elimination of labeling with the anti-pAPP antibody upon incubation with excess, phosphorylated, but not non-phosphorylated polypeptide corresponding to the APP cytoplasmic domain were described in our previous publications (Muresan and Muresan, 2004, 2005a, b). As specified in the data sheet #01 3004 for the anti-pAPP antibody (44-336Z), treatment of transfers of CAD cell extracts with lambda phosphatase completely eliminated antibody binding, showing strict specificity for pAPP.
Although intensively investigated, transport of APP within neurons is still incompletely understood, especially because of its complex posttranslational modifications, which include phosphorylation and extensive proteolytic processing (Fig. 1A). While the possibility that a fraction of APP might be transported into neurites as fragments generated by secretase cleavage was occasionally considered (Yamazaki et al., 1995; Buxbaum et al., 1998; Lazarov et al., 2002; Goldsbury et al., 2006), many studies of APP transport in real time used carboxy-terminally tagged APP, and assumed that the transport of the tag is representative of that of full-length APP (Kaether et al., 2000; Stamer et al., 2002). However, this approach does not discern between transport of the full-length APP and of the cleaved fragments containing the carboxy-terminal region (CTFs). To begin addressing the transport of endogenous APP in relation to its processing, we used immunocytochemistry to compare the overall patterns of localization of various APP epitopes in neuronal cells (i.e., CAD cells, primary neurons, and neurons in situ), focusing on their general distribution within the cell. We note that immunocytochemistry is the only procedure currently available to investigate localization of endogenous proteins. The employed antibodies recognize distinct domains of APP, are well characterized in terms of specificity (Grundke-Iqbal et al., 1989; Weidemann et al., 1989; Hilbich et al., 1993; DeGiorgio et al., 2002; Muresan and Muresan, 2004, 2005b, a) (see also results reported here), and detect both full-length APP and truncated APP forms containing their cognate site (Fig. 1A and supplemental Fig. 1T). Importantly, we included in our study antibodies that detect the cleaved carboxy-termini of either Aβ40 or Aβ42, which selectively bind to these polypeptides. Thus, the combined use of these antibodies should allow determining whether full-length APP or shorter polypeptides derived from APP are being detected.
Immunocytochemistry of cultured CAD cells [a CNS-derived, neuronal cell line (Qi et al., 1997) with relevance to AD (Muresan and Muresan, 2008)] showed that the epitopes corresponding to different domains of APP mostly localize to distinct regions of the cell: the antibodies against the cytoplasmic domain of APP consistently labeled mostly the cell soma, while those recognizing epitopes within the amino-terminal domain labeled to a large extent the neurite terminals (Fig. 1B, D, J and supplemental Fig. 1A-O). Antibodies to the Aβ region labeled primarily the neurites (with little or no accumulation at terminals), and the cell bodies (Fig. 1C). A comparable result was obtained with primary cortical neurons (Fig. 1E-I). Control experiments done 1) in the absence of primary antibodies, 2) in the presence of competing polypeptides, or 3) using negatively preadsorbed antibodies showed that the immunolabeling was specific (Fig. 1K and L, and supplemental Fig. 2). The segregated distribution of APP epitopes was seen under various fixation and permeabilization conditions, including the omission of sucrose from the fixative, simultaneous addition of the detergent and fixative (Fig. 1M-O), or cold methanol fixation (data not shown).
While the pattern of segregation of APP epitopes was similar in most cells, we also noticed subcellular distributions of APP epitopes that differed from those described above (unpublished results), suggesting that the level of expression and the pattern of processing of APP may occasionally differ among individual cells of similar type [see also (Muresan and Muresan, 2006)]. However, these were exceptions. We also noticed that antibodies recognizing carboxy-terminal epitopes (e.g., C9, AB5352, CT695) occasionally showed accumulations at neurite terminals (supplemental Fig. 1P, Q), a distribution that is typical for the localization of pAPP (Muresan and Muresan, 2005b) (see also supplemental Fig. 7). This terminal localization is likely due to the binding of these antibodies to phosphorylated APP species, as suggested by double labeling experiments done with these antibodies and an antibody to pAPP (see below and supplemental Fig. 1R and S). Finally, we note that the above-described segregation of epitopes from different regions of APP is typical for differentiated CAD cells, and less evident in non-differentiated cells.
The extensive absence of colocalization of epitopes from different regions of APP indicated that the antibodies largely detect APP fragments, not full-length APP, suggesting that a significant fraction of intracellular APP is cleaved into fragments. However, these results could also be obtained if, in some parts of the cell, certain epitopes in APP were “masked” by interactions with APP binding proteins, or otherwise inaccessible to the antibodies. The likelihood of such interactions is increased for the cytoplasmic domain of APP, known to bind proteins that contain phosphotyrosine-binding (PTB) domains. To address this concern, we chose for this study antibodies that recognize epitopes in APP outside the region involved in interactions with such proteins (Muresan and Muresan, 2004). As we recently showed, these antibodies can simultaneously detect different APP epitopes on the same vesicle (Muresan and Muresan, 2005b) (see also Fig. 7C-E). In addition, Western blots of differentiated CAD cell cultures showed that CAD cells contain significant amounts of APP cleavage products (CTFs and sAPPs; Fig. 2A and C). Also, we have previously shown that CAD cells contain significant amounts of the AICD (Muresan and Muresan, 2004), as well as Aβ (in monomeric and oligomeric forms) (Muresan and Muresan, 2006), which is consistent with the immunocytochemical observations reported here. However, the fact that the CAD cell cultures maintained in differentiating culturing conditions contain cells that are at various stages during differentiation makes it impossible to directly compare the results of immunocytochemistry with those of Western blots. We conclude that APP fragments are largely produced intracellularly - not at the cell surface, and are retained within the cell, where they become segregated in different neuronal regions.
The segregation of APP-derived polypeptides could occur by independent transport of the already cleaved polypeptides to the specific intracellular sites. To address whether APP is transported into neurites as full-length protein or as fragments, we studied the relative distribution of epitopes from different APP domains in transport vesicles along the neurites of cultured cells, where individual cargo vesicles can be accurately resolved and analyzed for epitope co-localization. As exemplified in Fig. 3A, epitopes from the Aβ and carboxy-terminal region of APP mostly localize to different transport carriers throughout the processes, indicating that they are transported separately. Control experiments done in the absence of primary antibodies (Fig. 3B), or in the presence of excess polypeptides corresponding to various APP regions (Fig. 3C, D) confirmed the low background labeling and the specificity of binding. Quantification of the vesicle population that contained both APP epitopes indicated that only ~(21±2)% (mean±SD) of the vesicles that contain APP carboxy-terminal epitopes also contain the Aβ epitope, and thus may carry the full-length APP or the CTFs. Conversely, only ~(15±2)% (mean±SD) of the vesicles that contain the Aβ epitope also contain APP carboxy-terminal epitopes, which indicates that many of these vesicles likely contain the γ-secretase cleavage product [i.e., the APP intracellular domain (AICD)]. Similarly, vesicles that contain APP amino-terminal epitopes were mostly segregated from those containing carboxy-terminal epitopes and the Aβ peptides, a result that was confirmed with a plethora of antibodies to APP epitopes, including terminal-end-specific anti-Aβ antibodies (supplemental Fig. 3A-P). Again, control experiments with excess competitor peptide confirmed specificity of binding (supplemental Fig. 3Q-T). Thus, the transport of APP into neurites occurs to a large extent as fragments, and not as full-length protein. The processes that determine the relative abundance of the different APP cleavage products within neurites remain to be investigated.
We previously showed that the fraction of APP that is phosphorylated at Thr668(pAPP) is transported into neurites independently of the non-phosphorylated APP (Muresan and Muresan, 2005b), and that, unlike the more abundant, non-phosphorylated APP, pAPP preferentially accumulates at the neurite terminals. The findings that APP fragments, rather than full-length APP, are transported along the neurites also raise the possibility that some of the neuritically detected pAPP might represent phosphorylated CTFs (pCTFs) or phosphorylated AICD (pAICD), not full-length pAPP. We investigated this possibility in very thin processes of CAD cells with antibodies against the ectodomain of APP and with antibodies that specifically recognize the Thr668-phosphorylated forms of APP (Fig. 3E-I). The strict specificity of the anti-pAPP antibody for phosphorylated APP species was validated in our previous studies (Muresan and Muresan, 2005a, b), and is re-confirmed here (Fig. 2B). While part of the labeling with the two antibodies was coincident (consistent with the presence of the full-length pAPP), we clearly detected a large population of vesicles that stained for the phospho-epitope, but not for the amino-terminal epitope (Fig. 3E-H). Quantification of the vesicle population that contained both APP epitopes indicated that only ~(19±4)% (mean±SD) of the vesicles that contain APP phospho-epitopes also carry amino-terminal epitopes (and thus may carry the full-length APP), with the majority likely containing pCTFs and pAICD. Moreover, the distribution pattern of the vesicle populations carrying the two APP epitopes differed significantly, with those containing phospho-epitopes appearing as clusters of vesicles separated by vesicle-free regions (Fig. 3G).
The segregation of APP amino-terminal epitopes from phospho-epitopes was particularly striking at the growth cone, where the latter localized to the peripheral (P) domain and the transition (T) zone, while the former to the central (C) domain of the growth cone (Fig. 4A-C). Importantly, both epitopes were present in vesicle-like structures that distributed along filamentous tracks, suggesting that these are transport vesicles (supplemental Fig. 4). Amino-terminal APP epitopes co-localized with the microtubule shaft that entered the growth cone, and extended along arc-shaped tracks (likely to be microtubules accompanying F-actin arc structures), which are typical for the T-zone (supplemental Fig. 4B, D, E) (Schaefer et al., 2002). APP phospho-epitopes were also present in the C-domain and T-zone, but clearly on different tracks than the APP amino-terminal epitopes (supplemental Fig. 4A, F-H). In addition, the phosphoepitopes were detected throughout the P-domain of the growth cone, where they aligned along filamentous tracks that extended into filopodia (supplemental Fig. 4C).
The above-described distribution of APP phospho-epitopes was typical for growth cones with wide, flattened lamellae, not yet committed to turning the direction of advancement (we call these growth cones “exploratory” growth cones). By contrast, in turning growth cones, committed to a new direction of extension, the APP phosphoepitopes showed a polarized distribution, with high concentration within the advancing protrusions (Fig. 4D-G and supplemental Fig. 5A-D). In such protrusions, the APP phospho-epitopes co-distributed with microtubules (Fig. 4H, I, and supplemental Fig. 5A-D). A different situation was found in filopodia that surrounded the growth cone, where APP phospho-epitopes extended to the tip of the filopodia, beyond the stable microtubules (Fig. 4J-O). A similar distribution was seen in the thin, filopodia-like processes emanating laterally from the main neuronal process (Fig. 4P-R, and supplemental Fig. 5E, F), which are typical for CAD cells (Li et al., 2005; Muresan and Muresan, 2006). Filamentous actin, detected with phalloidin, was found throughout the filopodia and filopodia-like lateral processes (Fig. 5A-G), suggesting that APP phosphoepitopes may penetrate into filopodia by moving on actin filaments rather than microtubules, a hypothesis that will be tested in future studies. Interestingly, APP amino-terminal and - to some extent - Aβ epitopes were less abundant in filopodia or filopodia-like processes (Fig. 5H-R), in line with our hypothesis that APP-derived polypeptides are independently transported to different destinations.
The results described above indicate that phosphorylated APP carboxy-terminal fragments, including the pAICD, are being transported into neurites, clearly in association with vesicle-like particles. Moreover, we found that an exogenously expressed, GFP-tagged AICD becomes phosphorylated, and some of the pAICD-GFP is transported to the neurite terminals in vesicle-like structures (supplemental Fig. 6). How the pAICD, which is presumably soluble and localizes in part to the nucleus (Muresan and Muresan, 2004), could become associated with transport vesicles remains to be established in future studies.
A close examination of the distributions of the different APP epitopes showed that these are likely transported along different microtubule tracks within neurites. For example, while amino-terminal APP fragments are transported along a limited set of tracks in the central portion of the neurites, fragments containing carboxy-terminal and Aβ epitopes are distributed throughout the entire width of the process, apparently being transported along a larger set of microtubule tracks (Fig. 3I and supplemental Fig. 3E-P). In vivo, microtubules are subjected to post-translational modification in several ways, including acetylation and detyrosination of α-tubulin (Luduena et al., 1992). Since the motility properties of kinesin-1 along microtubules with different post-translational modifications differ (Reed et al., 2006; Dunn et al., 2008), it was possible that transport of the different APP fragments occurs along sets of microtubules that are differentially altered by such modifications. In support of this hypothesis, we found that APP amino-terminal fragments, but not CTFs and Aβ peptides co-localized strictly with acetylated microtubules (supplemental Fig. 7A-T). Phosphorylated APP species and the c-Jun NH2-terminal kinase (JNK)-interacting protein-1 (JIP-1), a protein that is co-transported with pAPP (Muresan and Muresan, 2005b), showed a distribution pattern distinct from that of acetylated microtubules (supplemental Fig. 7G-R and supplemental Fig. 8E-H). By contrast, calsyntenin-1, a protein reported to localize to the central region of growth cones (Konecna et al., 2006), showed a distribution typical for the acetylated microtubules, and localized to regions that also contained amino-terminal APP epitopes (supplemental Fig. 8A-D), suggesting that calsyntenin-1 and APP amino-terminal fragments use similar transport pathways. Importantly, we found that the segregation of APP amino- from carboxy-terminal epitopes, with the confinement of the former to distinct filamentous tracks, occurs already in the cell body, prior to entrance into neurites. This was particularly evident in cells with flattened neurites, such as the one depicted in supplemental Fig. 7U, V.
Taken together, these results indicate that, in cultured neurons, APP is proteolytically processed prior to sorting into transport vesicles, and the polypeptides derived from it are transported within neurites independently, packaged in separate transport vesicles.
We asked whether the segregation of APP epitopes found in CAD cells and primary neurons in culture also occurs in brain neurons, in situ. To answer this question, we employed the transgenic mouse, R1.40-YAC that expresses an AD-specific mutant of the human APP at low levels (2-3-fold higher than endogenous APP levels) (Lamb et al., 1997; Lehman et al., 2003). We focused on the Purkinje cells in the cerebellum, which are ideally laid out in the tissue to allow concomitant examination of APP distribution in the cell body and throughout the processes. As shown in Figure 6, the antibodies to carboxy- and amino-terminal epitopes of APP labeled the cell bodies and the distal dendritic segments of Purkinje cells, respectively. An antibody to Aβ labeled the cell bodies and proximal dendritic arborizations, but not the distal processes. These labeling patterns reiterate the segregated distribution of these APP epitopes detected in cultured neurons, and support the notion that APP polypeptides are transported independently to different intraneuronal destinations.
The segregation of epitopes from different regions of APP seen in non-transfected cells was less evident in neurons that overexpressed APP. Indeed, unless cells expressed low levels of exogenous APP (Fig. 7G-K), the epitopes from different regions of overexpressed APP co-localized throughout the neurites and at their terminals (Fig. 7A-F). A typical example pertinent to the detection of APP phospho-epitopes is shown in Figure 7C-F, where CAD cells expressing the phospho-mimetic APP mutant, APPThr668Glu, were stained with antibodies to Aβ and to the carboxy-terminal phosphoepitope. Quantification of the vesicle population that contained both APP epitopes indicated that ~(93±13)% (mean±SD) of the vesicles that contain APP phosphoepitopes may carry the full-length APP, a result that is in sharp contrast to the results obtained for endogenous APP epitopes (see Figs. 1 and and3,3, and supplemental Figs. 1 and 3). We conclude that processing and transport of APP into neurites is largely dependent on the level of expression of APP.
In this study, we have addressed the transport of endogenous APP in relation to its proteolytic processing. To our surprise, we found that APP is transported into neurites as already cleaved polypeptides rather than full-length protein (Fig. 8). Within neurites, these APP cleavage products are detected within distinct transport carriers that associate with different sets of microtubules; some of the APP fragments occasionally co-distribute with actin filaments, not microtubules. These results indicate that the proteolytic cleavage of APP largely occurs prior to its final sorting into cargo vesicles, and that the cleaved fragments segregate into separate vesicle populations that reach different destinations within the cell soma, neurites, and their growth cones. We then showed that phosphorylated carboxy-terminal fragments of APP (including the pAICD), but not other APP-derived polypeptides, enter the lamellipodium and filopodia of growth cones, and become concentrated in regions of growth cone turning and advancement. Finally, we showed that these fragments might switch from a microtubule-based transport powered by kinesin-1 (operating within the neurite) to actin-based motility (operating in filopodia). Overall, our results overturn the long-hold view according to which transport of APP within neurites occurs, under normal conditions, mostly as full-length protein.
The findings that much of the intracellular APP is present as cleaved fragments, and that the cleaved fragments are transported into neurites and accumulate at the terminals should not be unexpected. Indeed, Buxbaum et al. (Buxbaum et al., 1998) found a surprisingly high content of CTFs at presynaptic terminals of entorhinal neurons; Amaratunga and Fine (Amaratunga and Fine, 1995) found that APP is transported along the axons of retinal ganglion cells as proteolytically processed protein; and Lee et al. (Lee et al., 2003) reported the presence of significant amounts of phosphorylated CTFs in mouse brain neurons. Most importantly, Sambamurti et al. (Sambamurti et al., 1992) presented evidence - more than 16 years ago - that in PC12 cells APP is cleaved intracellularly in the trans-Golgi network (TGN) or a post-Golgi compartment, and that the generated sAPP is then transported in vesicles and exocytosed. Our results are fully consistent with all these previous reports.
More recent data showed that in the mouse brain the levels of sAPP largely exceed those of full-length APP (Bai et al., 2007). This situation is also confirmed by results of immunoprecipitation experiments done to identify APP interacting proteins in vivo, which showed that antibodies to the extracellular (amino-terminal) and intracellular (carboxy-terminal) domains of APP co-precipitated distinct, mostly non-overlapping sets of proteins (Bai et al., 2007). Such a result can only be explained if the intracellular and extracellular APP epitopes are part of distinct polypeptides, which supports the conclusion that a significant fraction of APP is present as cleaved fragments, rather than full-length protein, in the mouse brain, in vivo.
APP processing by the secretase pathways is temporally and spatially regulated (De Strooper and Annaert, 2000), and the precise identification of the intracellular sites where the APP processing events occur is difficult. One reason for this situation is that localization studies of the components of α-, β- or γ-secretase do not provide information on their activation status. The γ-secretase complex - though probably inactive - assembles in the endoplasmic reticulum (ER) (Annaert et al., 1999; Kim et al., 2007), while BACE (β-secretase) is detected in the TGN and endosomes (Walter et al., 2001; He et al., 2007). Moreover, activation of γ-secretase requires proteolytic cleavage of the presenilins themselves (Capell et al., 1998). APP cleavage may thus occur almost anywhere along the secretory and recycling pathway: the TGN, the transport vesicle, the plasma membrane, the endosome, and autophagosome (Koo and Squazzo, 1994; Cook et al., 1997; Hartmann et al., 1997; De Strooper and Annaert, 2000; Kamal et al., 2001; Wilson et al., 2002; Yu et al., 2005). Our results are best explained by a scenario in which APP fragments are either generated in the TGN, or delivered (from another compartment, such as the endosome) to the TGN, and then selectively packaged into distinct transport vesicles. It is likely that full-length APP is initially targeted to the plasma membrane in the cell soma, then internalized by endocytosis, and processed in cell body endosomes (He et al., 2007). The fragments could then be retrotransported to the TGN for packaging into axonal transport vesicles. This pathway is consistent with the currently prevailing view that APP processing largely occurs in endosomes. Alternatively, it is possible that APP undergoes extensive proteolytic processing early in the secretory pathway - probably as soon as it exits the ER, rather than during recycling via endocytosis. Such a scenario is consistent with recent data showing that, in the in vivo brain, a significant fraction of all proteins found to interact with full-length APP are ER resident proteins (Bai et al., 2007), which suggests that intracellular holo-APP largely resides in the ER, and less in other intracellular compartments. Future studies will elucidate how the different APP fragments are selectively sorted to distinct transport vesicle populations.
Kinesin-1 is the main motor that transports both APP and the APP-derived polypeptides (including phosphorylated species) into neurites. This hypothesis is based on data that show that blocking transport by kinesin-1 abolishes accumulation within neurites of APP, as revealed with antibodies that detect epitopes throughout the APP molecule (Yamazaki et al., 1995; Kaether et al., 2000; Peretti et al., 2000; Muresan and Muresan, 2005b). Although KIF17 (a kinesin-2 motor) may interact with the APP binding protein, Mint1/X11α (Setou et al., 2000), data indicating that this kinesin motor is involved in APP transport along neurites are still lacking. In spite of the use of the same motor, the regulation of transport within neurites of the various APP fragments, and the characteristics of transport differ significantly. Our results suggest that amino-terminal APP fragments are preferentially transported along acetylated microtubules. The selection for a specific class of post-translationally modified microtubules appears to rely largely on factors other than the kinesin-1 motor itself. Indeed, all APP-derived polypeptides are carried into neurites by kinesin-1; yet, their transport occurs along distinct sets of microtubules.
How might kinesin-1 be recruited to the cargo vesicles that contain the different APP fragments and insure selective transport? It is possible that different kinesin-1 motors, differing in the heavy chain or light chain isoforms (Deboer et al., 2008), are recruited to specific cargo vesicles, a hypothesis that remains to be tested. Most likely, different APP-derived polypeptides (or sets of polypeptides) recruit kinesin-1 motors via different scaffolding proteins (Muresan, 2000). Like pAPP, the pCTFs and the pAICD may recruit kinesin-1 via JIP-1 (Muresan and Muresan, 2005a, b), a protein that was recently implicated in axonal development (Dajas-Bailador et al., 2008). Non-phosphorylated CTFs and AICD may recruit kinesin-1 via Fe65, a kinesin-1- (Lazarov et al., 2005) and APP-binding (Borg et al., 1996), scaffolding protein that binds with higher affinity to non-phosphorylated compared to phosphorylated APP species (Ando et al., 2001). How the presumably soluble (p)AICD attaches to the vesicle membrane is not known, and this may involve yet to be identified anchoring proteins. Alternatively, the (p)AICD may directly bind to acidic phospholipids via the amino-terminal cluster of positively charged lysine residues present in AICD.
Deciphering how kinesin-1 is recruited to vesicles that contain p3/Aβ or sAPPαβ, which are inside the vesicle, will be more difficult, since these APP fragments may not directly participate themselves in recruiting the motor machinery.
Interestingly, pCTFs (especially the pAICD) were systematically detected at locations likely to be reached by actin- rather than microtubule-based transport, such as the distal filopodia and the lateral filopodia-like processes. Although the characterization of the mechanism of pCTF transport within filopodia (including identifying the motor that powers this transport) was not the subject of the current study, myosin-X, an unconventional myosin that undergoes intrafilopodial motility (Berg and Cheney, 2002), could in principle bind to, and power the transport of pCTF-containing cargo. Indeed, myosin-X contains a PTB domain within its Band 4.1/Ezrin/Radixin/Moesin (FERM) domain (Sousa and Cheney, 2005) predicted to bind to the cytoplasmic domain of APP [and (p)CTFs].
Previous studies have proposed that APP may function in regulating some sort of actin-based motility in neurons (Sabo et al., 2001). Extending these studies in an unexpected way, our results suggest that the pCTFs (including the pAICD), not the full length APP, may play a role in growth cone turning and advancement, processes that involve both actin filaments and microtubules, as well as the action of molecular motors. It is significant that a recent study has implicated JIP-1 - the scaffolding protein that is cotransported with phosphorylated APP species (Muresan and Muresan, 2005b) - in axonal development (Dajas-Bailador et al., 2008). Combined with this study, our results raise the possibility that pCTF/JIP-1 complexes could be involved in regulating growth cone motility, and thus, neurite extension, a hypothesis that will be tested in the future.
From a technical perspective, our study emphasizes the benefit of employing multiple antibodies to study the intracellular localization of proteins (such as APP) that undergo complex post-translational modifications. We also draw attention to the fact that the segregation of APP fragments is not detected upon over-expression of APP, which suggests that the normal, physiological processing and transport of APP is largely perturbed at the high APP levels usually attained in transfected cells and some mouse models of AD.
We have provided evidence that the proteolytic cleavage of APP largely occurs prior to its sorting into transport vesicles destined for neuritic delivery, and that full-length APP represents only a fraction of the total (full-length plus cleaved) APP present within neuronal processes at any given time. We infer that a significant portion of the full-length APP is only an intermediary product with no function of its own, from which active polypeptides [sAPPs, (p)CTFs, (p)AICD, and Aβ] are generated prior to sorting into transport vesicles. Once generated, these polypeptides - some modified by phosphorylation - are packaged into distinct vesicles that travel on distinct cytoskeletal tracks to different destinations, where they exert their specific functions. For example, the pCTFs may play a role in growth cone turning and advancement, traveling on both microtubules and actin filaments. From this perspective, cleavage of APP to produce different functional molecules appears similar to the cleavage of peptide hormone precursors, such as pro-opiomelanocortin, which generates several hormones with distinct activities, but has no function as full-length protein (Alberts et al., 2002). The signaling that regulates the early APP processing and the many roles of the APP fragments remain to be determined.
We thank Dr. Dona Chikaraishi and Dr. James Wang for providing the CAD cell line; Dr. Samantha Cicero and Dr. Karl Herrup for providing the cortical neuron cultures; Dr. Carlos Dotti and Dr. Christoph Kaether for providing the APP-YFP cDNA; Dr. Li-Huei Tsai and Dr. Ming-Sum Lee for providing the APP-Thr668Glu cDNA; Dr. Brad Hyman and Dr. Ayae Kinoshita for providing the C58-GFP cDNA; Dr. Dennis Selkoe for providing the C9 anti-APP antibody; and Dr. Karl Herrup for many fruitful discussions and advice. Supported by NIH Grants GM068596 (Muresan), AG023012 (Lamb), and funds from UMDNJ (Muresan). To the DOTT, in memoriam.