|Home | About | Journals | Submit | Contact Us | Français|
Our goal is to understand the pathogenesis of amyloid-β (Aβ) deposition in the Alzheimer's disease (AD) brain. We established a cell culture system where central nervous system-derived neuronal cells (CAD cells) produce and accumulate within their processes large amounts of Aβ peptide, similar to what is believed to occur in brain neurons, in the initial phases of AD. Using this system, we show that accumulation of Aβ begins within neurites, prior to any detectable signs of neurodegeneration or abnormal vesicular transport. Neuritic accumulation of Aβ is restricted to a small population of neighboring cells that express normal levels of amyloid-β precursor protein (APP) but show redistribution of BACE1 to the processes, where it colocalizes with Aβ and markers of late endosomes. Consistently, cells that accumulate Aβ appear in isolated islets, suggesting their clonal origin from a few cells that show a propensity to accumulate Aβ. These results suggest that Aβ accumulation is initiated in a small number of neurons by intracellular determinants that alter APP metabolism and lead to Aβ deposition and neurodegeneration. CAD cells appear to recapitulate the biochemical processes leading to Aβ deposition, thus providing an experimental in vitro system for studying the molecular pathobiology of AD.
Alzheimer's disease (AD) belongs to a group of neurodegenerative disorders associated with the formation of large protein aggregates (42). Thus, the neuritic plaques found in AD brains contain at their core deposits of the amyloid-β (Aβ) peptide generated by proteolytic processing of the transmembrane protein amyloid-β precursor protein (APP) (45).
APP has a complicated life cycle. A large portion of the newly synthesized APP is degraded by the lysosomal pathway (11). However, a fraction of APP is processed intracellularly by two mutually exclusive pathways, which generate polypeptides that are secreted, degraded, or released into the cytosol (43). The two pathways use either α-secretase or β-secretase activities to cleave APP at distinct sites (close to the transmembrane domain) and release into the vesicle lumen large, soluble amino-terminal polypeptides (sAPP-α or sAPP-β) of unknown function that are ultimately secreted (see Fig. Fig.1H).1H). The membrane-bound, carboxy-terminal fragments (CTFs) that result from this cleavage (i.e., CTF-α or CTF-β) have a relatively long half-life and are either degraded or further processed by an intramembrane proteolytic activity, γ-secretase. This proteolytic processing releases the short polypeptides p3 or Aβ into the vesicle lumen and CTF-γ into the cytoplasm. Some of the Aβ and p3 peptides are secreted into the extracellular space and either enter the circulation or are cleared in some other way. Under the pathological conditions of AD, Aβ (in particular Aβ42) aggregates and becomes incorporated into neuritic plaques, processes thought to initiate the cascade that leads to neuronal loss and dementia (19). Although APP is ubiquitously expressed, accumulation, oligomerization, and aggregation of Aβ, followed by its incorporation into plaques, occur only in specific regions of the brains of AD patients, under circumstances that are not understood.
Although the Aβ-containing plaques are extracellular, increasing evidence suggests that production, oligomerization into protofibrils, and accumulation of Aβ occur intracellularly, within neuronal processes, rather than extracellularly (reviewed in reference 17), and that the incorporation of aggregates into plaques is a late-stage event. Indeed, in neurons from mouse models of AD expressing mutant forms of human APP, Aβ is generated in many intracellular compartments, including the endoplasmic reticulum, the Golgi apparatus, the secretory vesicles, endosomes, and autophagic vacuoles (43, 53, 58). In these models, accumulation of Aβ is associated with aging and is likely facilitated by the overexpression of APP and the familial AD mutation present in APP (46, 49). However, a significant fraction of AD cases are found in individuals that express normal levels of nonmutated APP, and most have no genetic predisposition known to alter Aβ metabolism (42). In addition, plaque formation appears to affect only certain neuronal populations in specific brain regions, indicating that some neurons are more prone than others to accumulate Aβ (7). This suggests the intriguing hypothesis that unknown, cell-specific, intrinsic factors sensitize certain neurons to neurodegeneration through increased production and accumulation of Aβ.
In this study, we asked whether neuronal cells maintained in culture may preferentially accumulate Aβ peptide and form large deposits. As an experimental system, we chose the central nervous system-derived, catecholaminergic cell line CAD. In our previous work, we characterized CAD cells with regard to intracellular transport and posttranslational processing of APP (37-39). We showed that, as in primary neurons, in these cells APP is processed by secretase activities to generate CTFs that are translocated into the nucleus (37). We also showed that both full-length APP and the CTFs become phosphorylated at a critical threonine residue (corresponding to Thr668 in APP695) and that this phosphorylation regulates transport of a fraction of APP into neuronal processes (38), and of CTFs into the nucleus (37). Phosphorylation of APP in CAD cells is extensive and normally occurs via a specific signaling pathway that requires the activity of c-Jun NH2-terminal kinase (JNK) (38). Since increased APP phosphorylation was recently linked to the overproduction of Aβ (29), CAD cells may represent a good system to study whether intracellular generation of Aβ could lead to formation and accumulation of Aβ aggregates similar to those characteristic of AD. Oligomerization of Aβ has been recently reported for neuronal cells overexpressing human APP with the Swedish mutation (49), but few studies investigated cells expressing normal levels of endogenous APP (46).
We found that a small fraction of the CAD cells normally exhibit Aβ deposits throughout their neurites. These deposits were largely concentrated at neurite terminals, where they colocalized with β-secretase. Surprisingly, the Aβ-depositing cells appeared in isolated islets, suggesting their clonal origin. We conclude that a small number of neuronal cells normally show biochemical and neuropathological features of degenerating neurons present in AD brains. Our data suggest that intracellular determinants, in addition to genetic and extracellular environmental factors, may contribute to the onset of Aβ deposition and AD.
Primary antibodies used in this study include mouse anti-Alzheimer precursor protein A4, recognizing an amino-terminal epitope (residues 66 to 81; MAB348, clone 22C11); rabbit anti-β-amyloid 1-40 (AB5074P) and anti-β-amyloid 1-42 (AB5078P) (Chemicon, Temecula, CA); mouse anti-human amyloid-β protein (clone 4G8, recognizing residues 17 to 24, and clone 6E10, recognizing residues 1 to 17; antibody 6E10 cross-reacts with mouse Aβ when present at high concentrations, in spite of the fact that the recognized epitope is only partially conserved between humans and rodents); rabbit anti-rodent Aβ (recognizing residues 3 to 16) (Signet, Dedham, MA); rabbit anti-APP (no. 2452; raised against a polypeptide from the cytoplasmic domain of APP; Cell Signaling Technology, Beverly, MA); rabbit anti-pAPP (no. 44-336Z) and rabbit antioligomer antibody (AHB0052, clone A11) (BioSource International, Camarillo, CA); rabbit anti-early endosomal antigen 1 (EEA1; Affinity BioReagents, Golden, CO); and rabbit anti-JIP1 and rabbit anti-Rab7 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Rabbit anti-BACE1 antibodies were obtained from Bruce Lamb (Cleveland Clinic Foundation, Cleveland, OH) (9) and Riqiang Yan (Cleveland Clinic Foundation) (56). Microtubules were stained with a rabbit antitubulin antibody (ICN Biomedicals, Aurora, OH).
The mouse central nervous system-derived, catecholaminergic cell line CAD (obtained from Dona Chikaraishi [Duke University Medical School, Durham, NC] and James Wang [Cogent Neuroscience, Inc., Durham, NC]) (40) was grown in 1:1 F-12-Dulbecco’s modified Eagle medium, supplemented with 8% fetal bovine serum and penicillin-streptomycin. Cell differentiation was induced by removal of serum from the culture medium. Mouse (embryonic day 16.5) cortical neurons were grown in neurobasal medium with B-27 supplement, l-glutamine, and penicillin-streptomycin for 5 days, with a change of medium on day 4. Neurons were maintained in growth medium until fixation for immunocytochemistry.
CAD cell populations enriched in cells that accumulate Aβ were obtained by selecting small groups of cells in microtiter wells and screening the obtained cultures for Aβ accumulating cells by immunocytochemistry with antibody 6E10 (after differentiation). Occasionally, the procedure was repeated, using cultures that appeared to be enriched in such cells after the first round of selection.
CAD cells were transfected with human APP695 by using FuGene 6 (Roche Diagnostics, Indianapolis, IN). A construct in pcDNA3 of APP695 was obtained from Li-Huei Tsai (Harvard Medical School, Howard Hughes Medical Institute, Boston, MA).
Differentiated CAD cells were rinsed twice with phosphate-buffered saline (PBS) and extracted in sodium dodecyl sulfate sample buffer for 5 min at 95°C. Extracts were separated in 16.5% Tris-Tricine gels (Bio-Rad, Hercules, CA) or 14% Tris-glycine gels, followed by transfer to polyvinylidene difluoride membranes. Aβ was detected by immunoblotting using alkaline phosphatase-coupled secondary antibodies and colorimetric visualization of the reaction product (36). Peptides with a molecular size corresponding to Aβ (monomers and oligomers) were detectable only in wet membranes, probably because the peptides had penetrated the membrane. Therefore, membranes were scanned while wet.
Transfected or nontransfected CAD cells and primary cultures of cortical neurons were fixed for 20 min in PBS containing 4% formaldehyde and 4% sucrose, then permeabilized with 0.3% Triton X-100 (20 min, 20°C), and processed for single or double antigen labeling as previously described (35). Double labeling with antibodies 6E10 and A11 or 6E10 and anti-Aβ carboxy-terminal-end antibodies were done by coincubation of the specimens with the two primary antibodies. Successive incubation usually led to preferential labeling of the neuritic deposits with the antibody applied first. This was likely due to steric hindrance between antibodies detecting the same or vicinal epitopes. Occasionally, detergent extraction was omitted, or cells were extracted with Triton X-100 prior to fixation. Secondary antibodies coupled to Alexa dyes were from Molecular Probes (Eugene, OR). Actin filaments were stained with a fluorescein-phalloidin conjugate (Molecular Probes). Nuclear DNA was stained with 4,6-diamidino-2-phenylindole (DAPI; Pierce Biotechnology, Rockford, IL). Digital images were obtained with an Olympus IX81 or a Nikon Optiphot microscope (100× oil, 20×, 40× objectives) equipped with cooled charge-coupled device cameras and collected using Image-Pro Plus (Media Cybernetics, Inc., Silver Spring, MD) or Optronics Magnafire image analysis software. Images were processed for contrast and brightness by Adobe Photoshop. Colocalization of 6E10-labeled deposits with β-secretase or oligomers at neurite terminals was quantified using Image-Pro Plus (32).
To detect cell surface APP that becomes endocytosed, CAD cells were incubated for 30 min at 37°C in the presence of anti-rodent Aβ antibody, then rinsed 3 times with cold PBS, and fixed. Endocytosed antibody, bound to the Aβ region of APP, was detected by immunocytochemistry with fluorescently labeled anti-rabbit immunoglobulin G (IgG). In control experiments, cells were incubated in the absence of anti-rodent Aβ antibody. To estimate fluid phase uptake of antibody, cells were incubated with nonimmune rabbit IgG. Under the conditions of the uptake experiment and immunostaining procedure, the amount of fluid-phase endocytosed IgG was negligible.
Necrotic CAD cells were detected by staining with propidium iodide (PI; Sigma-Aldrich, St. Louis, MO). Briefly, cells were incubated for 5 min with culture medium containing 0.5 mg/ml PI prior to rinsing and fixation.
As previously reported (37, 40), CAD cells extend long processes when cultured in the absence of serum (Fig. 1A to C); these processes contain APP that is relatively homogeneously distributed throughout their length (Fig. 1D to G). To determine if Aβ accumulations are present in CAD cells, we used two monoclonal antibodies, 4G8 and 6E10, that have been extensively employed to detect, by immunocytochemistry, Aβ deposits in AD brains. We also tested a polyclonal antibody raised to a peptide from the amino-terminal region of rodent Aβ. Although these antibodies also detect full-length APP and some of its cleavage products (4G8: APP, CTF-α, and CTF-β; 6E10 and anti-rodent Aβ: APP, CTF-β, and sAPP-α [Fig. [Fig.1H]),1H]), accumulations of Aβ, if present, should be easily detected by fluorescence microscopy as bright, intense spots (49).
In Western blots of CAD cell lysates, all anti-Aβ antibodies faintly detected a polypeptide band corresponding to Aβ (our unpublished results). This result was expected, since the Aβ peptide is notoriously difficult to detect by Western blotting of cell lysates.
Using immunocytochemistry, we investigated the overall staining pattern given by the above listed antibodies in differentiated CAD cells. Antibody 4G8 revealed a continuous distribution of APP in CAD cell processes, which made the detection of any sporadic Aβ accumulations difficult (our unpublished results). By contrast, the anti-rodent Aβ antibody identified a small number of cells that showed accumulations of immunoreactive material within neurites, and especially at neurite terminals (Fig. 2A and B). This intense staining was detected over the background labeling that most likely represented total APP, which made it difficult to identify and assess the distribution of these deposits. We therefore turned to antibody 6E10, which detects poorly endogenous levels of mouse APP but is reactive towards APP and Aβ, if these are present at high concentrations (Signet data sheet 9320 of 2001 for antibody 6E10 and our unpublished results).
Immunocytochemistry of differentiated CAD cells using the 6E10 antibody yielded a surprising result. While most cells showed no immunoreactivity, sporadic clusters of cells were intensely stained, particularly at their processes (Fig. 2C to F). Labeling along these neurites was nonhomogeneous and was often highly concentrated at the terminals. We referred to these highly fluorescent areas as accumulations of 6E10-immunoreactive material and hypothesized that they most likely contain Aβ aggregates rather than full-length APP. It is unlikely that this labeling represents nonspecific binding of the antibody to epitopes unrelated to APP. Indeed, antibody 6E10 did not label any structures in cortical neurons derived from the brains of APP knockout mice (our unpublished results).
To determine whether the 6E10-immunoreactive material within the neurite terminals represents full-length APP or cleavage products of APP, we performed double-labeling experiments with 6E10 and with several antibodies that detect other APP epitopes (Fig. (Fig.1H1H and and3).3). First, we used an antibody that recognizes the carboxy-terminal domain within the full-length APP or within the CTFs generated by the action of secretases. This antibody labeled the cell bodies more intensely than the neurites and did not particularly stain the 6E10-immunoreactive material (Fig. 3A to C). This result suggests that these deposits do not contain significant amounts of full-length APP or CTFs.
Since the labeling pattern of 6E10-immunoreactive cells strongly resembled the distribution of Thr668-phosphorylated APP (pAPP) (Fig. (Fig.2G)2G) (1, 23, 38), we tested whether the 6E10 immunoreactivity colocalized with pAPP, using an antibody that specifically recognizes pAPP. We found no significant overlap between the labeling with these antibodies (our unpublished results; see also Fig. 9G and H). On the contrary, the 6E10-immunoreactive material was often present in cells that contained little or no pAPP within their terminals (Fig. 3D and E).
We also did not detect significant colocalization of 6E10 staining with staining by an antibody to the amino-terminal domain of APP (our unpublished results). However, we detected a high degree of colocalization of the 6E10 immunoreactivity with the labeling by an antibody to rodent Aβ (Fig. 4A to D). To further verify that the neuritic deposits stained with these antibodies contain Aβ peptide, we immunolabeled CAD cells with antibodies that detect the cleaved carboxy termini of either Aβ40 or Aβ42 (Fig. (Fig.4E-L).4E-L). Like the 6E10 antibody, the Aβ carboxy-terminal-end antibodies preferentially stained groups of cells (Fig. 4G and H), with significant labeling at the neurite terminals (Fig. 4E to L), which frequently colocalized—at least in part—with the 6E10 labeling (Fig. 4M and N). We noted that these Aβ terminal-end antibodies (particularly, anti-Aβ42) also showed significant labeling in the cell body, in a region that may correspond to the endoplasmic reticulum (Fig. 4G to J). This result is in line with previous reports that detected significant amounts of Aβ42 in the endoplasmic reticulum of NT2N neurons (10, 46). As with antibody 6E10, the carboxy-terminal-end antibodies showed no labeling of APP−/− neurons, which confirmed their specificity (our unpublished results). In summary, the neuritic deposits detected by antibody 6E10 also reacted with four additional antibodies to the Aβ peptide, two of which are carboxy-terminal end specific.
To determine whether the neuritic accumulations contained oligomerized Aβ, we used a well-characterized antibody that specifically recognizes the oligomeric state of polypeptides (26). As shown in Fig. 4O and P, this antibody extensively labeled CAD cell processes, and in many cases this labeling significantly colocalized with antibody 6E10 labeling (Fig. 4Q to S). Since the antioligomer antibody preferentially detects species larger than the octamer (26), the deposits likely include large-molecular-size oligomers. Taken together, these results support the notion that the accumulations of 6E10-immunoreactive material contain bona fide Aβ peptides. Therefore, we referred to this 6E10 immunoreactivity as Aβ accumulations. Since they were not detected in nonpermeabilized cells (our unpublished results), we conclude that the Aβ accumulations are largely intracellular. In addition, because a significant fraction of these Aβ accumulations were resistant to extraction with nonionic detergent (applied prior to fixation) and cross-reacted with the antioligomer antibody, it is likely that they also contain oligomeric forms of Aβ. Taken together, these results indicate that CAD cells develop accumulations of Aβ within their neurites and that at least a fraction of this Aβ is oligomerized.
Since CAD cells that contained Aβ accumulations usually appeared in clusters, we reasoned that they may have additional common features. One of the possibilities might be that these cells undergo neurodegeneration and death. A close examination of the cells showing Aβ accumulations did not reveal any detachment of their processes from the cell bodies, thus eliminating the possibility that these cells were undergoing a wallerian-type degenerative process.
Next, we examined whether these cells show signs of plasmalemmal disintegration typical for necrosis. Cell cultures were incubated with PI prior to fixation and immunolabeling with the antibody 6E10. Only a few cells showed nuclear PI staining (indicative of plasma membrane damage and intracellular penetration), even in the areas that contained many cells with Aβ accumulations (Fig. 5A to C). Only rarely, we found groups of several necrotic, PI-positive cells, but these cells did not contain Aβ accumulations (Fig. 5D to F). Examination of CAD cell cultures also indicated that cells rich in Aβ deposits did not show signs of apoptosis: they showed neither blebbing of the cell membrane nor fragmentation of nuclei, as revealed by DAPI staining (Fig. 5G to N). Taken together, these results indicate that CAD cells containing Aβ accumulations do not show signs of degeneration or cell death.
Cells that showed Aβ accumulations exhibited a variety of phenotypes. Under normal culturing conditions, many of these cells extended short processes, even when maintained for several days under differentiating conditions (Fig. 2C to F and 6A and B). However, Aβ accumulations were also detected in cells that extended one (Fig. (Fig.6K)6K) or more (Fig. (Fig.6J)6J) long processes. In these cases, the deposits were distributed throughout the processes, filling varicosities and accumulating heavily at the terminals (Fig. 6G and J). Accumulation of Aβ appeared to start very early during differentiation and neurite extension, since antibody 6E10 stained regions of cells from which processes appeared to be just forming (Fig. 6C and D) as well as advancing growth cones (Fig. 6E and F). Usually, all neurites of an affected cell exhibited Aβ accumulations, most of which was concentrated at their terminals (Fig. 6G to I).
As noted above, Aβ accumulations were preferentially detected within the neurites. In addition, the 6E10 antibody—like the antioligomer antibody and the Aβ carboxy-terminal-end antibodies (Fig. 4E to J)—also labeled the perinuclear regions of these cells (Fig. 6B and K), in an area that contains the endoplasmic reticulum, the Golgi apparatus, and the endosomes. Since endosomes are a major site of active β-secretase residence (27), they might be the site of generation of the perinuclearly detected Aβ.
At high resolution, Aβ, as detected by the antibody 6E10, appeared to be associated with a heterogeneous population of vesicle-like particles of various dimensions throughout the neurites (Fig. 7A to C), or localized at the neurite terminals (Fig. 7D to L). A similar particulate, vesicle-like labeling pattern was seen with anti-Aβ carboxy-terminal-end (Fig. 4K and M) and antioligomer (Fig. (Fig.4R)4R) antibodies. Based on their size, we think that most of the larger vesicular structures are of endosomal or autophagosomal origin.
The Aβ-containing, vesicular structures at the terminals of processes (Fig. 7D to L) could represent either transport vesicles that reached the neurite terminals or endosomes or autophagic vesicles generated locally, at the terminals. To test whether the Aβ was localized to early endosomes, we performed double-labeling experiments with the antibody 6E10 and antibodies that specifically label early endosomes. As exemplified in Fig. 7Q to S for EEA1, neuritic Aβ accumulations did not significantly colocalize with early endosomes, which were largely concentrated in the cell body of CAD cells (Fig. (Fig.7R7R).
To further confirm that the detected Aβ is not associated with early endosomes within neurites, we tested for colocalization of the 6E10-detectable accumulations with bona fide, endocytosed APP. Since cell surface APP is known to become partially taken up by endocytosis (54), we traced the internalized APP by incubating CAD cells for 30 min with a rabbit antibody to rodent Aβ (recognizing an extracellular epitope in APP), followed by washing and fixation of the cells. We then detected the bound and internalized antibody with fluorescent anti-rabbit IgG. As shown in Fig. 7M to P, the anti-rodent Aβ antibody that was internalized during the 30-min time interval became concentrated in the cell body, in a perinuclear region, thus marking the early endosomal compartment. Little of this antibody was detected within processes and at their terminals (Fig. (Fig.7O).7O). Importantly, the Aβ deposits detected with antibody 6E10 were largely segregated from the internalized anti-rodent Aβ antibody and accumulated within neurites (Fig. 7M and N). Taken together, these results suggest that a large fraction of the Aβ detected with antibody 6E10 is localized in a post-early-endosomal compartment, outside the cell body. They also exclude the possibility that the detected intracellular Aβ might originate from the endocytosis of an extracellular, soluble Aβ pool at neurite terminals. Double-immunolabeling experiments with antibody 6E10 and Rab7, a marker for late endosomes (12) and autophagic vacuoles (18), suggested that a significant fraction of the Aβ detected at neurite terminals resides in late-endosomal or autophagic compartments (Fig. 7T to W). These results suggest that, while the process of generation of Aβ from APP might, in principle, begin in membrane-bound compartments in the cell body, concentration of Aβ occurs only within neurites, most likely in late endosomes and autophagic vacuoles.
To begin to understand what factors might favor increased generation of Aβ in a selected population of cells, we investigated whether CAD cells showing Aβ accumulations express more APP. As shown in Fig. 3A to C, Aβ-containing cells did not show increased levels of APP (compared to neighboring, normal cells), as detected with an antibody recognizing an epitope within the cytoplasmic domain of APP. Moreover, exogenous expression of APP at high levels did not lead to the accumulation of significantly increased amounts of 6E10-stainable material within neurites of transfected cells (our unpublished results). These results indicate that accumulation of Aβ is not necessarily a consequence of increased levels of APP; more likely, it is the result of increased production of Aβ through the action of secretases.
Next, we asked whether cells that accumulate Aβ show increased levels of secretases, the proteases responsible for Aβ generation. We focused on the major β-secretase, BACE1 (22, 30, 44, 51, 55), using immunocytochemical detection. In most cells, BACE1 was localized primarily within the cell bodies and to a lesser extent within processes (Fig. (Fig.8).8). While BACE1 levels did not appear to be significantly increased in the cell bodies of Aβ-producing cells, in over 80% of cells that contained large neuritic Aβ accumulations, increased levels of BACE1 were detected within the neurites, where it colocalized with Aβ (detected with 6E10 antibody; colocalization coefficient, 87 ± 8 [mean ± standard deviation]) (Fig. (Fig.8).8). This remarkable result suggests that accumulation of Aβ within neurites of CAD cells is coincident with pools of β-secretase, abnormally localized within neurites.
What are the physiological consequences of the accumulation of Aβ peptide within cell processes? We began to address this question by analyzing the transport of vesicular cargoes within the neurites that contained Aβ accumulations. This is a relevant question, considering the reports that link APP processing and Aβ deposition to abnormal axonal transport (24, 47).
First, we examined the overall integrity of the cytoskeletal networks that support vesicular transport, the microtubules, and the actin filaments. As shown in Fig. 9A to F, staining of CAD cells with antibodies to tubulin and with phalloidin detected normal distributions of microtubules and actin filaments in the processes that contained large accumulations of Aβ (labeled with the antibody 6E10). Next, we examined the localization of pAPP and JNK-interacting protein 1 (JIP-1), two vesicular cargoes of the microtubule motor kinesin-1 (34, 39). We and others previously showed that the accumulation of pAPP (39) and JIP-1 (52) at neurite endings is a direct consequence of their transport by kinesin-1, and that abnormal transport leads to decreased accumulation of these proteins at the terminals. As shown in Fig. 9G to J, the levels of pAPP and JIP-1 that accumulated at the terminals of processes in CAD cells that contained extensive amounts of Aβ were similar to those detected in nonaffected cells. In addition, we found no correlation between the amount of pAPP and Aβ deposits present at neurite terminals. This result indicates that transport of vesicular cargoes into neurites is not generally perturbed by the presence of large amounts of accumulated Aβ. It also suggests that increased generation of Aβ can occur in the absence of detectable signs of abnormal axonal transport.
As described above, cells with increased accumulations of Aβ in their processes represent a small fraction of the total cells and appear in islet-like clusters, segregated from the dominant, conventional cells (Fig. (Fig.2).2). Since CAD cells divide when cultured in the presence of serum, it seems likely that the cells within each separate cluster originate from one or a few cells that are prone to increased generation of Aβ. If this is true, then one could obtain, by dilution cloning, a population of CAD cells that is enriched in cells exhibiting the Aβ deposition phenotype. Indeed, using such a procedure, we obtained cultures that contained increased numbers of cells with 6E10-positive Aβ accumulations within neurites (Fig. (Fig.10).10). Cells showing this phenotype were detected very early after differentiation was induced, as soon as they started to extend processes (Fig. 10B and C). When maintained for longer periods of time under differentiation conditions, some of these cells extended long processes, loaded with Aβ (Fig. 10A). Western blots of concentrated lysates of CAD cells from such cultures showed the presence of monomeric and polymeric Aβ, as detected with an antibody to rodent Aβ and the 6E10 antibody (Fig. 10D). Notably, the 6E10 antibody detected oligomeric species of Aβ with higher sensitivity than the anti-rodent Aβ antibody. Future studies will be aimed at using such CAD cell cultures to investigate the determinant factors and the mechanism of Aβ generation and accumulation within neurites.
This experimental system should be a good model for the study of the generation and accumulation of Aβ, complementing studies with bona fide primary neurons. In this respect, we note that primary cultures of cortical neurons from normal mice occasionally contain cells that show accumulations of material immunoreactive towards antibodies to Aβ, particularly in the cell body but also within neurites (Fig. (Fig.11).11). Like in CAD cells, the immunofluorescent material within neurites is detected as large puncta, reminiscent of late endosomes or autophagic vacuoles (Fig. 11D). However, unlike CAD cells, the cortical neurons mostly lacked the heavy accumulation of Aβ at neurite terminals, which likely explains the significantly reduced neuritic immunoreactivity with antibody 6E10 (Fig. 11A), compared to that in CAD cells. We also note that CAD cells originate from neurons in the brain stem (see Discussion) and may thus have different properties than cortical neurons.
In this study, we report the establishment of a cell culture system in which central nervous system-derived cells produce and accumulate large amounts of Aβ peptide from endogenous APP within their neurites. This process is remarkably similar to what is now believed to occur within the neurites of affected neurons in the initial phases of AD. Using this system, we confirm that significant accumulation of insoluble Aβ begins within neurites, in particular at their terminals, long before Aβ deposits are detected extracellularly. Accumulation of Aβ within neurites is normally restricted to a small population of neighboring cells that also show redistribution of BACE1 to the processes, where it colocalizes with Aβ. These events occur exclusively in differentiated cells, prior to any detectable signs of neurodegeneration or abnormal transport of vesicular cargoes into neurites. Taken together, these results suggest that the process of conversion of Aβ to a detergent-resistant form and its neuritic accumulation is initiated in a small number of neurons by intracellular determinants that alter APP metabolism.
Animal models and cell culture systems manipulated to express human versions of mutant, disease-prone APP or components of the APP processing machinery (e.g., presenilin-1, BACE1) remain important tools for the identification of molecular mechanisms that operate in familial forms of AD. In contrast, the availability of neuronal lines that show a propensity to form intraneuritic Aβ deposits at endogenous levels of APP and secretases may provide insight into the pathogenic mechanisms that alter Aβ metabolism in sporadic AD forms that may not have a genetic basis. Such cell lines are the previously reported human NT2N neurons (46) and the mouse cell line CAD, used in this study. Initially established by targeted oncogenesis in transgenic mice, CAD cells have spontaneously lost the original oncogene, that encoding simian virus 40 T antigen. They are diploid, chromosomally stable cells that exhibit biochemical and morphological characteristics of primary neurons (40). In previous work, we demonstrated that the resolution achieved by fluorescence microscopy in the thin processes of CAD cells allows the resolution of individual transport packages even without the use of confocal microscopy (39). CAD cells are a promising model system for studying neurodegenerative diseases, including torsion dystonia (20, 25), Parkinson's disease (2), and AD (29). As we show here, CAD cells are particularly suited for the investigation of APP metabolism in conjunction with its transport into processes. This is important, in view of the recent evidence that transport and processing of APP are intimately entangled (24, 28, 41, 47). CAD cells, which are likely derived from the locus coeruleus in the brain stem (40, 48), are relevant to the neuropathology of AD. Indeed, neurons in the locus coeruleus aberrantly express cell cycle proteins (8) and are largely affected by cell death in AD (6, 14, 59), in spite of the fact that this brain region lacks high densities of neuritic plaques and tangles.
The first important result of this study is that CAD cells, expressing endogenous levels of APP, normally develop deposits of Aβ that are detergent resistant, indicative of oligomerized Aβ. Results of immunocytochemistry done with an antioligomer antibody that does not recognize monomeric or fibrillar Aβ confirm that at least a fraction of the Aβ detected in CAD cells is in an oligomeric form. We also note that the Aβ accumulations reported in this study are best detected with antibody 6E10, which binds to Aβ oligomers and fibrils, in addition to the monomers (26; see also the BioSource International product data sheet for the antioligomer antibody AHB0052). This result is in line with the known fact that antibodies that recognize amino-terminal regions in Aβ (e.g., 6E10) detect oligomerized Aβ more efficiently than antibodies that bind to internal regions (such as 4G8) (4). The fact that under our experimental conditions the 6E10 antibody indeed detects Aβ deposits was confirmed by colocalization with two antibodies to rodent Aβ (Fig. 4A to D and our unpublished results) and the absence of immunoreactivity to antibodies to APP regions outside Aβ (Fig. (Fig.3).3). Finally, experiments using Aβ carboxy-terminal-end antibodies (Fig. 4E to L) indicate that both Aβ40 and Aβ42 are present—to various extents—in CAD cells. While these deposits likely also contain CTFs, in addition to Aβ, these may either be below the detection limit of our assays or have their carboxy-terminal epitopes masked. In this respect, we note that the exact composition and the aggregation state of the Aβ polypeptides present in these accumulations cannot be revealed by immunocytochemistry and biochemistry alone. Indeed, the accessibility of antibodies to the various Aβ epitopes is certainly differentially affected by the aggregation state of the Aβ polypeptide. For example, some epitopes that are accessible in Aβ oligomers might not be accessible in Aβ fibrils. This likely explains why double-labeling experiments with antibodies to distinct epitopes in Aβ showed—in many cases—only partial colocalization. Studies to further characterize these deposits with regard to composition and aggregation state of the polypeptides are under way.
The second important result is that Aβ accumulates in the distal portion of CAD cell neurites, in particular at their terminals, a process thought to occur in early stages of AD (5, 16). Oligomerization of Aβ within processes and synapses of cultured neurons expressing human APP with the Swedish mutation (derived from Tg2576 mice) has been recently described (49). However, unlike Tg2576 neurons, where Aβ gradually accumulates within processes over time in culture (49), CAD cells show preferential neuritic deposits of Aβ starting very early during differentiation. Eventually, all processes of affected CAD cells become filled with deposits, irrespective of neurite length and neurite number per cell. Whether and how these intracellular Aβ accumulations can become extracellular (likely by cell disintegration or a form of exocytosis) remains to be established in future studies.
A third important result of our study is that cells that accumulate Aβ can still differentiate and function normally, at least for some time in culture. Importantly, these cells still appear to normally transport to neurite terminals vesicular cargoes, such as those containing pAPP and JIP-1. Moreover, cells with Aβ accumulations do not show signs of abnormal cytoskeleton, neurodegeneration, or cell death. These results suggest that while a deficient axonal transport may cause neurodegeneration (41, 47), accumulation of Aβ within neurites can certainly occur in the absence of any detectable abnormal intraneuritic transport. We are currently conducting a more detailed investigation of vesicle transport in CAD cells that accumulate Aβ.
A fourth important result is that cells that contain Aβ accumulations within neurites also contain BACE1, the major β-secretase, at the same location, coincident with Aβ. The extent of BACE1 accumulation within the neuronal processes of these cells is abnormal, since this enzyme is normally localized to Golgi compartments and early endosomes in the cell body (27, 56). This result suggests that mislocalization of BACE1 may cause the production and accumulation of Aβ within processes and synaptic regions. This result is in line with a recent report that proposes that BACE1 localization—in addition to its expression level (13, 21, 57)—determines the amount of generated Aβ and its accumulation in plaques (28).
The colocalization of Aβ with secretases within neurites and at terminals suggests that Aβ may be generated during transport through the processes. This does not necessarily mean that this Aβ is contained in Golgi-derived secretory vesicles, particularly because BACE1 is active only in the acidic environment provided by endosomes. Our results of colocalization with early endosomal markers and uptake experiments clearly show that the neuritic Aβ does not accumulate in early endosomes, which are concentrated in the cell body region. It is likely that the Aβ is contained in late endosomes, such as multivesicular bodies, that may originate by maturation from early endosomes generated in the cell body and are then transported anterogradely, down the processes. This hypothesis is supported by our data on colocalization with late endosomal markers and is consistent with recent reports that identify the intraneuritic compartments that contain Aβ in neurons as multivesicular bodies (50). Alternatively, the Aβ could be contained in autophagic vesicles that form within the neurites. Indeed, a recent study showed that macroautophagy may be an important pathway for Aβ generation in AD (58). Although not yet detected in CAD cells, macroautophagy might be selectively triggered in some of these cells, ultimately leading to the abnormal generation and accumulation of Aβ. Further studies are required to identify the exact pathway of vesicular transport that results in accumulation of Aβ at the distal end of neurites.
An important result of our study is that cells that accumulate Aβ within neurites appear to originate by division from a small number of cells, present in the culture, that possess the propensity to form Aβ accumulations. Thus, intracellular determinants conferring the Aβ phenotype would be clonally transmitted to the progenitors of a few cells that exhibit the biochemical and neuropathological features of degenerating neurons present in the brains of AD patients. These intracellular determinants are not necessarily genetically inherited. As recently shown, expression of neuronal genes can be spontaneously altered by retrotransposition, resulting in neuronal somatic mosaicism, a phenomenon seen both in cultured neurons and in the brains of adult mice, in vivo (33). Thus, events triggering Aβ oligomerization and accumulation may be initiated randomly within single cells (16). Overall, our data suggest that intracellular determinants present in a small number of neurons may contribute to the onset of Aβ deposition in AD.
It is intriguing that while other studies done with cultured cortical and hippocampal neurons from APP transgenic mice occasionally identified neuritic staining for Aβ (49), preferential accumulation at neurite terminals—as found by us in CAD cells—has not been reported. We think that this may be due to differences between neurons from different brain regions. In this respect, neurons derived from the brain stem, such as the CAD cells, may hold the key to explaining the onset of plaque formation in AD. The locus coeruleus neurons innervate many brain regions, including the cerebral cortex and the hippocampus (3, 31), where neuritic plaques are abundant in AD. Plaque formation in the cerebral cortex and hippocampus could be seeded by oligomerized Aβ that accumulates at the terminals of projections of neurons with their cell bodies in the brain stem. As our results clearly show, the locus coeruleus-derived CAD cells accumulate oligomerized Aβ primarily at the terminals of their processes, which become swollen and contain varicosities. Similarly, tyrosine hydroxylase-containing nerve terminals—extending from the locus coeruleus—are markedly enlarged in the proximity of neuritic plaques in AD brains and in mouse models of AD (15). We propose that spontaneous accumulations of Aβ at the terminals of brain stem neurons that project into the cortex and hippocampus may nucleate the formation of the neuritic deposits. Thus, the neuropathology of AD may actually begin in subcortical regions and then spread to the cortex and hippocampus. This hypothesis remains to be tested.
In conclusion, we propose a novel cell culture system for the study of AD-like early events in the generation and accumulation of Aβ in neuronal cells. We show how this system can be used to address questions of processing and transport of APP and to potentially identify molecular determinants relevant to AD pathology. Finally, we provide support for the hypothesis that redistribution of BACE1, which may spontaneously occur in a small population of neurons, may lead to abnormal generation and regional accumulation of Aβ. These accumulations may in time become extracellular and serve as seeds for development of neuritic plaques. Our results thus suggest a clonal origin of abnormal Aβ metabolism and plaque formation. Further studies, possibly using CAD cell cultures enriched in cells that accumulate Aβ, are required to test this hypothesis.
This work was supported by National Institutes of Health grant 5RO1GM068596-02 and a Mt. Sinai Health Care Foundation scholarship (V.M.).
We thank Dona Chikaraishi and James Wang for kindly providing the CAD cell line; Samantha Cicero and Karl Herrup for kindly providing the cortical neuron cultures; and Li-Huei Tsai, Ming-Sum Lee, Bruce Lamb, and Riqiang Yan for kindly providing cDNA constructs and antibodies. We also thank Karl Herrup and Bruce Lamb for many fruitful discussions on topics covered in this paper.