|Home | About | Journals | Submit | Contact Us | Français|
Phosphorylation of amyloid-β precursor protein (APP) at Thr668 is a normal process linked to neurite extension and anterograde transport of vesicular cargo. By contrast, increased phosphorylation of APP is a pathological trait of Alzheimer's disease. APP is overexpressed in Down's syndrome, a condition that occasionally leads to increased APP phosphorylation, in cultured cells. Whether phosphorylation of APP in normal versus high APP conditions occurs by similar or distinct signaling pathways is not known. Here, we addressed this problem using brainstem-derived neurons (CAD cells). CAD cells that ectopically overexpress APP frequently show features of degenerating neurons. We found that, in degenerating cells, APP is hyperphosphorylated and colocalizes with early endosomes. By contrast, in normal CAD cells, phosphorylated APP (pAPP) is excluded from endosomes, and localizes to the Golgi apparatus and to transport vesicles within the neurites. Whereas the neuritic APP is phosphorylated by c-Jun NH2-terminal kinase through a pathway that is modulated by glycogen synthase kinase 3β, the endosomal pAPP in degenerated CAD cells results from activation of cyclin-dependent kinase 5. Additional signaling pathways, leading to APP phosphorylation, become active during stress and mitosis. We conclude that distinct pathways of APP phosphorylation operate in proliferating, differentiating, stressed, and degenerating neurons.
A stated goal of systems biology is to accurately describe and predict how signal transduction pathways function in both normal and diseased cells (Wiley, 2006 ) and to explain diseases as alterations of the normal signaling pathways or activation of novel pathways that are normally inactive. In this study, we have identified and briefly characterized signaling pathways that all lead to the phosphorylation of a key threonine residue of amyloid-β (Aβ) precursor protein (APP)—a protein relevant to Alzheimer's disease and Down's syndrome—in the context of normal neuronal function and during degeneration.
Alzheimer's disease, a complex neurodegenerative disorder of old age humans, is characterized by two major brain lesions: the neuritic plaques and the neurofibrillary tangles (Selkoe, 2001 ). Neuritic plaques contain extracellular deposits of Aβ peptide generated by proteolytic processing of the transmembrane protein, APP (Price et al., 1995 ). Neurofibrillary tangles consist of intracellular, filamentous aggregates of the microtubule-binding protein, tau (Friedhoff et al., 2000 ). Aβ deposition and tau aggregation in Alzheimer's disease are likely the result of altered metabolism and posttranslational modification of APP and tau (Selkoe, 2001 ). Patients with Down's syndrome also develop Alzheimer's disease pathology, including neuritic plaques (Head and Lott, 2004 ). There is evidence that neurodegeneration and plaque formation in Down's syndrome and some early-onset cases of Alzheimer's disease is likely caused by increased APP levels (Rovelet-Lecrux et al., 2006 ; Salehi et al., 2006 ; Sleegers et al., 2006 ).
Tau aggregation is accompanied by extensive phosphorylation, done mainly by cyclin-dependent kinase 5 (Cdk5) and glycogen synthase kinase 3β (GSK3β; Friedhoff et al., 2000 ). APP is also hyperphosphorylated in Alzheimer's disease, but the nature of the kinases that perform this phosphorylation is uncertain. The phosphorylated residue is Thr668 (numbering for APP695); Cdk5 (Iijima et al., 2000 ), GSK3β (Aplin et al., 1996 ), and c-Jun NH2-terminal kinase (JNK; Standen et al., 2001 ; Muresan and Muresan, 2005a ) are the candidate kinases. It was proposed that APP phosphorylation alters its metabolism, leading to increased production of carboxy-terminal fragments (CTFs; Lee et al., 2003 ).
Phosphorylation of APP also occurs normally in differentiated neurons. This phosphorylation has been implicated in the transport of vesicular cargoes into neurites (Muresan and Muresan, 2005b ), in neurite extension (Ando et al., 1999 ; Muresan and Muresan, 2005a ), and in signaling to the nucleus (Muresan and Muresan, 2004 ; Kimberly et al., 2005 ; Chang et al., 2006 ). In normal, differentiated neurons, phosphorylation of APP occurs through a specific signaling pathway that uses JNK and the JNK-interacting protein-3 (JIP-3; Muresan and Muresan, 2005a ). Cdk5 and GSK3β may also contribute to APP phosphorylation, in conditions that remain to be established.
The mechanism of APP phosphorylation is only partially characterized. Recent reports indicate that a primary site of APP phosphorylation in Alzheimer's disease neurons is the endosome (Lee et al., 2003 ). By extension, it was suggested that the phosphorylated APP (pAPP) found in the neurites of normal neurons might also represent endosomal pAPP (or phosphorylated CTFs). However, we recently reported that pAPP is recruited to Golgi-derived vesicles and is transported by kinesin-1 into neurites (Muresan and Muresan, 2005b ). APP is also phosphorylated during cellular stress and during mitosis (a situation relevant to brain development), but little is known about these processes. Here, we characterize the subcellular compartmentalization and the signaling pathways that lead to the production of pAPP in normal neurons (during proliferation, differentiation, and stress) and in neurons that overexpress APP. This condition leads to degeneration of neurons in vivo (Nishimura et al., 1998 ) and in culture, possibly by increasing sensitivity of APP-overexpressing cells to neurotoxins (Hanson et al., 2003 ). Our study was conducted in CAD cells (Qi et al., 1997 ), a CNS-derived, catecholaminergic, neuronal cell line that has emerged as an excellent system for studying neuronal cell biology and pathology (Lee et al., 2003 ; Muresan and Muresan, 2005b , 2006 ).
Using CAD cells, we compared the mechanism of phosphorylation of APP during mitosis, normal differentiation, and stress and in cells showing neurodegenerative phenotypes caused by overexpression of APP. We used immunocytochemistry to detect global changes in the amount and localization of pAPP in mitotic and degenerating cells, compared with normal cells. We also used dominant negative approaches and a battery of specific kinase inhibitors to identify the kinases and to define the signaling pathways that lead to APP phosphorylation. We found that, depending on context, at least four mutually exclusive pathways lead to the phosphorylation of Thr668 of APP, with distinct consequences on the intracellular localization and function of pAPP.
The following primary antibodies were used in this study: rabbit anti-Cdk5 (C-8), rabbit anti-p35 (C-19), mouse anti-JIP-1, rabbit anti-JIP-3 (Santa Cruz Biotechnology, Santa Cruz, CA); rabbit anti-early endosome antigen 1 (EEA1; Affinity BioReagents, Golden, CO); mouse anti-Rab5, mouse anti-EB1 (BD Transduction Laboratories, BD Biosciences, Palo Alto, CA); mouse anti-α-adaptin (AP2; Affinity BioReagents); rat anti-CD71 (transferrin receptor; Leinco Technologies, Ballwin, MO); mouse anti-Golgi 58K protein (Sigma, St. Louis, MO); mouse anti-ubiquitin (StressGen Biotechnologies, Victoria, BC, Canada); rabbit anti-phospho-SAPK/JNK (Thr183/Tyr185; Cell Signaling Technology, Beverly, MA); mouse anti-Alzheimer precursor protein A4 (MAB348, Clone 22C11; Chemicon, Temecula, CA); mouse anti-human Aβ protein (clone 4G8; Signet, Dedham, MA; used in immunocytochemistry); rabbit anti-APP (no. 2452; raised against a polypeptide from the APP cytoplasmic domain; Cell Signaling Technology). The latter antibody recognizes total APP (phosphorylated and nonphosphorylated; Muresan and Muresan, 2004 ), and was used for immunoblotting. Antibody 44-336Z (BioSource International, Camarillo, CA) was used to detect pAPP. This antibody recognizes specifically the Thr668-phosphorylated, but not nonphosphorylated forms of APP (Muresan and Muresan, 2005a ,b ; see also the BioSource data sheet for this antibody, showing specificity strictly for pAPP). Mouse anti-α-tubulin (Clone B-5-1-1) was from Sigma. The mouse anti-lysosomal membrane glycoprotein (LAMP-1) antibody (1D4B) developed by Dr. Thomas August (Johns Hopkins University, Baltimore, MD) was obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa (Iowa City, IA). Green fluorescent protein (GFP) was detected with mouse (BD Living Colors; BD Biosciences) and chicken antibodies (Aves Labs, Tigard, OR). The anti-GFP antibodies cross-react with YFP.
The mouse CNS-derived, catecholaminergic cell line, CAD (Qi et al., 1997 ), was grown in 1:1 F12:DME medium, containing 8% fetal bovine serum and penicillin/streptomycin. Differentiation was induced by culturing cells in serum-free medium.
CAD cells were transfected with FuGene 6 (Roche Diagnostics, Indianapolis, IN). Constructs of human APP695 and GFP-tagged, dominant negative Cdk5 (dnk5-GFP; Niethammer et al., 2000 ) were obtained from Dr. Li-Huei Tsai (Massachusetts Institute of Technology, Howard Hughes Medical Institute, Cambridge, MA). A JIP-3 construct (pcDNA3-JIP-3-FLAG) was obtained from Dr. Roger Davis (University of Massachusetts Medical School, Howard Hughes Medical Institute, Worcester, MA; Kelkar et al., 2000 ). A pcDNA3/APP-YFP (human APP695) construct was obtained from Dr. Carlos Dotti (Katholieke Universiteit Leuven, Belgium) and Dr. Christoph Kaether (EMBL, Heidelberg, Germany; Kaether et al., 2000 ).
SDS-PAGE and Western blot analysis was done as previously described (Muresan and Arvan, 1997 ).
After transfections, CAD cells were plated in complete medium (to allow attachment), and then transferred to differentiation medium without serum. At 48 h after transfection, cells were fixed for immunocytochemistry. At this stage, CAD cells show extended processes with well-defined growth cones, and axonal transport is robust. Occasionally, CAD cells were fixed after 72–120 h of differentiation. In some cases, CAD cells were fixed without inducing differentiation, to capture cells during mitosis.
Kinase inhibitors were added directly to the culture medium of differentiated or nondifferentiated CAD cells (transfected or not) as follows: the JNK inhibitor, SP600125 (Biomol Research Laboratories, Plymouth Meeting, PA), 25 μM, 1–8 h (Bennett et al., 2001 ; Han et al., 2001 ); the Cdk5 inhibitor, roscovitine (Calbiochem, San Diego, CA;), 20 μM, 6 h or overnight (Niethammer et al., 2000 ); the GSK3β inhibitor SB415286 (Sigma), 100 μM, 1–1.5 h (Wang et al., 2002 ); and the GSK3β inhibitor, lithium chloride, 20 mM, 1–1.5 h. Control cultures received equal amounts of dimethyl sulfoxide (DMSO) or 20 mM potassium chloride. For the activation of the JNK stress pathway, cells were treated with sorbitol (0.4 M, 30 min at 37°C; Inomata et al., 2003 ).
Transfected or nontransfected CAD cells were fixed for 20 min in phosphate-buffered saline containing 4% formaldehyde and 4% sucrose, permeabilized with 0.3% Triton X-100 (20 min, 20°C), and processed for labeling with antibodies as previously described (Muresan et al., 2000 ). Secondary antibodies coupled to Alexa dyes (488 and 594) were from Molecular Probes (Eugene, OR). A fluorescein-labeled goat anti-chicken IgY was from Aves Labs. Transfected APP-YFP was detected via the YFP fluorescence. Occasionally, sensitivity of detection was increased by immunolabeling with anti-GFP antibodies. Digital images were obtained with an Olympus IX81 microscope (20×, 40× objectives; Tokyo, Japan) equipped with deep cooled ORCA ER monochrome CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan), and collected using Image-ProPlus software (Media Cybernetics, Silver Spring, MD). Images were processed for contrast and brightness with Adobe Photoshop (San Jose, CA).
The degree of colocalization of various antigen pairs (e.g., pAPP and an endocytic marker; see Figure 3) was quantified as follows: First, the mean pixel intensity for each pair of images was determined in Photoshop CS2 (Sabo et al., 2003 ). Images were then thresholded at the mean intensity, to eliminate most of the diffuse, nonparticulate labeling. Finally, the thresholded images were multiplied. The resulting image showed nonzero pixel intensities only in regions of colocalization, i.e., regions were pixel intensity differed from zero in both images. Percentages of colocalization were calculated by dividing the number of pixels of nonzero intensities in the product image by the number of pixels of nonzero intensities in the parental images. The degree of colocalization of pAPP with endocytic markers was confirmed by calculating coefficients of colocalization, using Image-ProPlus software (Media Cybernetics; Manders et al., 1993 ).
The distribution of immunolabeled, fluorescent particles along neurites was analyzed in thresholded, inverted, grayscale images, using the “plot profile” function of the NIH Image 1.63 software. This procedure was used to compare the distribution of two antigens (e.g., pAPP and EEA1; see Supplementary Figure 3) along the neurites.
The percentage of neurites that showed significant accumulation of pAPP at their terminals in cell cultures treated with kinase inhibitors (i.e., SP600125, roscovitine, SB415286, and lithium chloride) was estimated by comparison with DMSO- or potassium chloride–treated cells (see Figure 5 and Supplementary Figure 4). Micrographs, obtained under identical conditions of exposure and image processing, were thresholded in Photoshop CS2, to limit the input levels between 50 and 255. This processing eliminated low intensity labeling. Percentages of neurites that showed pAPP at terminals were derived from the thresholded images (see legend to Figure 5 and Supplementary Figure 4, D–F).
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. Between 57 and 150 cells in each experimental group were analyzed.
Our goal was to investigate APP phosphorylation pathways in neurons during normal function and in conditions of increased APP expression. The latter is a feature of Down's syndrome. Toward this end, we used the CNS-derived cell line, CAD (Qi et al., 1997 ), a well-characterized model system for studying APP phosphorylation (Muresan and Muresan, 2004 , 2005a ,b ), with relevance to neurodegenerative diseases that involve APP (Lee et al., 2003 ; Muresan and Muresan, 2006 , 2007 ).
Similar to primary neurons (Muresan and Muresan, 2005b ), in differentiated CAD cells endogenous pAPP is largely localized to post-Golgi vesicles that accumulate at neurite endings, and represents only a small fraction of the total, mature APP (Supplementary Figure 1). Because of its low level, this endogenous pAPP fraction is difficult to detect—and to quantify—in immunoblots.
In conditions of APP overexpression, the fraction of APP that becomes phosphorylated is easily detectable. As in normal cells, most of the pAPP, detected in cells transfected with APP, is present in the fully matured, sulfated, and glycosylated APP form (Supplementary Figure 1A). However, the pAPP has altered intracellular distribution. Although endogenous pAPP is seldom detected in the cell body, an increased number of CAD cells overexpressing APP show significant or predominant localization of pAPP in the cell body (Figure 1, A–K). Importantly, many CAD cells with increased levels of pAPP also show dystrophic morphology, such as extensive vesiculation, and short or degenerated neurites (Figure 1, C and G). Such neurites, when present, show increased tortuosity and numerous varicosities (Figure 1, D–G), similar to what has been previously described for cortical neurons treated with fibrillar Aβ (Grace et al., 2002 ). Cells that did not show obvious signs of degeneration did not have increased levels of pAPP in the cell body, in spite of the fact that they expressed high levels of APP-YFP (Figure 1, L–N). Quantitatively, we found a significant correlation between the presence of large amounts of pAPP (concentrated in the cell body) and cell degeneration, in CAD cells overexpressing APP. Thus, ~70% of cells that show vesiculation and dystrophic morphology also show increased levels of pAPP. By contrast, only ~6% of cells with normal, differentiated phenotype show increased pAPP in the cell body (Figure 1O).
Under normal culturing conditions, a small number of CAD cells degenerate and die. These are identified by their shape (i.e., rounded and detached from the substrate) and by their abnormal microtubule cytoskeleton, lacking the typical radial distribution of microtubules emanating from the centrosome. We found that these cells consistently show increased levels of pAPP in the cell body, as detected by immunocytochemistry (Figure 1, P–R).
The increased levels of pAPP detected by immunofluorescence in cells that show features of degeneration is not an artifact caused by the rounded shape of these cells, because other proteins for which we tested—including the microtubule plus-end–tracking protein, EB1 (Figure 2, A–C), the phosphorylated JNK, and JIP-3 (see Figure 6, F–J)—are not up-regulated. We also tested whether the cells with degenerated appearance were undergoing apoptosis. As shown in Figure 2, A–F, no nuclear fragmentation is detected in these cells, indicating that increased phosphorylation of APP likely occurs before the degenerating cells become apoptotic. Moreover, apoptotic CAD cells found occasionally in cell cultures do not show increased pAPP levels (Figure 2, G–I). However, we found that cells overexpressing APP, with degenerated appearance, showed increased levels of ubiquitinated proteins (Figure 2, J–L). Neuronal inclusions containing ubiquitinated proteins are commonly detected in many neurodegenerative diseases (Ross and Pickart, 2004 ). Taken together, these findings indicate that neurons undergoing degeneration show increased levels of phosphorylated APP that is abnormally localized to the cell body.
Next, we set to identify the subcellular compartments containing pAPP within the cell body of normal and degenerated CAD neurons. In normal, differentiated cells, pAPP is only occasionally detected in the cell body. In such cases, most of it colocalizes with the Golgi marker protein, 58K (Supplementary Figure 2), suggesting its presence within the Golgi apparatus. This result is consistent with the reported preferential phosphorylation of the fully matured, glycosylated and sulfated APP form (processed in the Golgi apparatus), compared with immature forms (Ando et al., 1999 ; see also Supplementary Figure 1A, lanes 3 and 4).
As previously reported, within neurites, pAPP colocalizes with the JNK scaffolding proteins, JIP-1 and -3; activated, phosphorylated JNK; and the anterograde motor, kinesin-1 (Muresan and Muresan, 2005a ). On the basis of these, and other results (Muresan and Muresan, 2005b ), we concluded that APP is phosphorylated in the cell body, and accumulates in the distal neurites by anterograde transport. However, a similar distribution pattern would be detected if APP were phosphorylated at the neurite terminals (in endosomes), and then transported retrogradely toward the cell body. Such a retrograde transport of APP, endocytosed along the axonal surfaces, has been previously described (Yamazaki et al., 1995 ; Marquez-Sterling et al., 1997 ). We therefore asked whether neuritic pAPP could represent endocytosed, rather than Golgi-derived, exported pAPP. First, we carried out double-immunolabeling experiments of pAPP with protein markers for the early endosomal compartments: AP2 (Ball et al., 1995 ), EEA1 (Stenmark et al., 1996 ), and the small GTPase, Rab5 (Gorvel et al., 1991 ). Figure 3, A–J, shows that pAPP and the endocytic markers are distributed throughout the neurites in a punctate, vesicle-like manner. Although the degree of colocalization of pAPP with the various markers of endocytosis was variable, generally there was minimal overlap between the immunolabeling of pAPP and of either EEA1 (Figure 3, A–C and Supplementary Figure 3), Rab5 (Figure 3, D–G), or AP2 (Figure 3, H–J). Second, pAPP does not colocalize with a bona fide recycling protein, the transferrin receptor (Harding et al., 1983 ; Figure 3, K–M). Because a large proportion of APP is metabolized in lysosomes, we also tested whether pAPP might be targeted to lysosomes. As shown in Figure 3N, pAPP does not colocalize with the lysosomal marker, LAMP1 (Chen et al., 1985 ), which is restricted to cell bodies. We conclude that most pAPP detected within neurites is associated with anterogradely moving vesicles, but not with early endosomes or lysosomes.
A fraction of CAD cells transfected with APP contains an enlarged endosomal compartment, suggesting increased endocytic activity in some cells that overexpress APP (Figure 4, A–J). These cells display increased level of pAPP in the cell body, with a substantial fraction of the pAPP being localized to regions that also contain endosomal markers (Figure 4, G–J). As described above, these cells also show morphological features of degeneration. Importantly, although the endosomal compartment is enlarged (Figure 4, A–J), the Golgi apparatus remains compact, and restricted to a small region adjacent to the nucleus (Figure 4, K–N). We conclude that pAPP localizes primarily to compartments of the secretory pathway in normal CAD cells, whereas it preferentially accumulates in enlarged endosomes within the cell body, in degenerating cells.
To identify the kinases responsible for APP phosphorylation in normal versus degenerating CAD cells, we used pharmacological inhibitors. We have previously shown that, in differentiated CAD cells, JNK activity is required to phosphorylate APP, and to initiate the vesicular transport that leads to accumulation of pAPP at the neurite terminals (Muresan and Muresan, 2005a ). Because other studies have implicated Cdk5 and GSK3β in APP phosphorylation in neurons, we re-examined the effect of inhibitors targeting these kinases on APP phosphorylation. As previously reported, the JNK inhibitor SP600125 efficiently inhibits APP phosphorylation, to the point that pAPP cannot be detected (Figure 5C). These results were reproduced with CAD cells cultured for both short and long time periods, and at various cell densities (our unpublished results). Confirming our previous report (Muresan and Muresan, 2005a ), the treatment of CAD cells with the Cdk5 inhibitor, roscovitine does not prevent phosphorylation of APP and accumulation of pAPP at neurite endings (Figure 5D). On the contrary, ~10% more CAD cells appear to accumulate pAPP at neurite terminals upon roscovitine treatment than without treatment. Such increase in the fraction of differentiated cells that accumulate pAPP within neurites is also detected in cells that express a dominant negative form of Cdk5, dnk5 (Figure 5, E–H). These data suggest that blocking Cdk5 activity normally enhances JNK-dependent APP phosphorylation and accumulation of pAPP at neurite terminals. However, the effect of Cdk5 inhibition on APP phosphorylation may be indirect.
Next, we tested the effect of GSK3β inhibition on APP phosphorylation in differentiated, normal CAD cells. As shown in Figure 5, A and B, inhibition of GSK3β by treatment with the specific inhibitor, SB415286 reduces the amount of neuritic pAPP, but does not eliminate it. A mostly similar result is obtained upon treatment of CAD cells with lithium chloride (Supplementary Figure 4). These results indicate that GSK3β facilitates JNK-mediated phosphorylation of APP, and suggest that GSK3β might be part of the signaling cascade that includes JNK, possibly through activation of a kinase upstream of JNK (Kim et al., 2003 ). We further tested the participation of GSK3β in the phosphorylation of APP using CAD cells transfected with JIP-3, a condition that stimulates APP phosphorylation via JNK (Muresan and Muresan, 2005a ). As shown in Supplementary Figure 5A, inhibition of GSK3β significantly reduces the number of JIP-3–transfected cells showing pAPP labeling, whereas JNK inhibition essentially eliminates APP phosphorylation and accumulation of pAPP at terminals. Confirming results obtained in normal CAD cells (Figure 5), inhibition of Cdk5 consistently increases APP phosphorylation (Supplementary Figure 5).
We next examined the effects of these kinase inhibitors on APP phosphorylation in CAD cells that show phenotypes of degeneration in nontransfected cells (Figure 5, I–Q), or in cells transfected with APP-YFP (Figure 5, R–Z′). When treated with either JNK (SP600125) or GSK3β (SB415286) inhibitors, degenerating CAD cells show accumulation of pAPP in the cell body similar to untreated, degenerating cells (Figure 5, I–N, R–W, and Z′; compare with relevant images in Figure 1). By contrast, the Cdk5 inhibitor, roscovitine, promotes a significant reduction of pAPP in the cell body of degenerating cells (Figure 5, O–Q, X–Z, and Z′). These results support the notion that pAPP in degenerating neurons is the result of the activity of Cdk5, rather than GSK3β or JNK.
Further, results of immunocytochemistry are consistent with the activation of Cdk5, but not JNK, in degenerating CAD cells. Indeed, slightly increased levels of Cdk5 and its activator, p35 (or p25; the antibody detects both p35 and p25) are detected in the cell body of degenerating cells that express APP-YFP (Figure 6, A–E). Most importantly, these cells also showed changes in the distribution of Cdk5 and p35/p25 within the cell body (see inset in Figure 6D). By contrast, no increase in the level of the active, phosphorylated form of JNK or of JIP-3 (which facilitates APP phosphorylation by JNK) is detected in such cells (Figure 6, F–J). Taken together, these results suggest that different kinases control APP phosphorylation in normal versus degenerating CAD cells. Specifically, a GSK3β/JNK pathway operates in normal cells leading to the transport of pAPP into neurites, whereas Cdk5-dependent APP phosphorylation prevails in degenerating neurons, and leads to accumulation of pAPP in endosomal compartments, within the cell body.
Besides the APP phosphorylation pathways described above, at least two other pathways can be activated in CAD cells. First, stress activates a typical JNK cascade that leads to phosphorylation of many substrates, including APP (Inomata et al., 2003 ; Muresan and Muresan, 2005a ). As shown in Figure 7, A–D, the pAPP generated by osmotic stress is distributed diffusely throughout the cell body, without an increase in the neuritically localized pAPP. On the contrary, generation of pAPP via this stress-activated pathway is accompanied by a significant reduction of pAPP accumulation at neurite terminals (Figure 7, E–I). This result suggests that, under conditions of stress, transport into neurites might be diminished, a hypothesis that remains to be confirmed in future studies.
Second, as previously reported (Suzuki et al., 1994 ), APP is phosphorylated during the G2/M phase of the cell cycle. In nondifferentiated, proliferating CAD cells, APP is phosphorylated only during mitosis. Figure 7, J–L, shows pAPP localized to large particles throughout mitotic CAD cells. In nonneuronal cells, mitotic phosphorylation of APP is apparently performed by the Cdc2 kinase (Suzuki et al., 1994 ). Consistent with this result, our preliminary data show that the level of pAPP in mitotic CAD cells is significantly reduced in the presence of inhibitors of cyclin-dependent kinases.
Taken together, our results indicate that the same residue in APP, Thr668, is phosphorylated through different kinase pathways that are preferentially activated during certain physiological or pathological conditions. We propose that the context of APP phosphorylation—timing and location—allows for distinct functional interactions of pAPP with other proteins, a hypothesis that remains to be tested in future studies.
This study addresses the problem of APP phosphorylation in neuronal cells, using CAD cells as model system, and is intended as a first step toward a more comprehensive investigation of the role of APP phosphorylation in neuronal function and dysfunction. Although the short cytoplasmic domain of APP contains several residues that can be phosphorylated (Oishi et al., 1997 ), most studies have concentrated on Thr668, mainly because its phosphorylation alters significantly the conformation of APP (Ramelot and Nicholson, 2001 ) and its interaction with physiologically relevant proteins, such as Fe65 and JIP-1 (Muresan and Muresan, 2005b ). This phosphorylation may also alter APP processing (Lee et al., 2003 ; Ryder et al., 2003 ). Although phosphorylation of APP at Thr668 is thus highly relevant to the pathobiology of Alzheimer's disease, the pathways that lead to this phosphorylation, and their regulation, are subject of controversy. Studies from various laboratories reported that, in differentiated neurons, APP is phosphorylated by Cdk5 (Iijima et al., 2000 ), GSK3β (Aplin et al., 1996 ), or JNK (Standen et al., 2001 ; Muresan and Muresan, 2005a ). Although the lack of consensus with regard to this problem may be due to differences between neurons, it is more likely that APP may be phosphorylated by multiple kinases acting either in concert, or separately, according to a particular physiological state of the cell. In addition, certain signaling pathways not normally active may be activated under pathological conditions that lead to, or are caused by, abnormal expression of APP. One aim of our study was to distinguish between signaling pathways that lead to APP phosphorylation in one cell type (i.e., CAD cells), in the normal versus the degenerated state. We found that, in cells that show morphological traits of degeneration, APP is hyperphosphorylated by Cdk5, and the generated pAPP localizes to endosomes within the cell body. This is different from the situation in normal, differentiated CAD cells, where a small fraction of total APP is phosphorylated by JNK through a signaling pathway that requires the participation of JIP-3 (Muresan and Muresan, 2005a ) and GSK3β (as shown in this study). This pAPP is transported into neurites and accumulates at their terminals (Muresan and Muresan, 2005b ).
We started by investigating APP phosphorylation in degenerating CAD cells that naturally occur within cultures of normal CAD cells (which serve thus as controls). Spontaneous degeneration in cultures of CAD cells is rare, but is readily detectable because of alterations in the cell shape and organization of the microtubule cytoskeleton. To increase the number of degenerating cells, we used overexpression of exogenous APP, a condition known to cause degeneration in postmitotic neurons in vitro and in vivo (Yoshikawa et al., 1992 ; Nishimura et al., 1998 ; Hanson et al., 2003 ). This experimental manipulation mimics defects present in Down's syndrome (Head and Lott, 2004 ) and some early-onset cases of Alzheimer's disease (Rovelet-Lecrux et al., 2006 ; Sleegers et al., 2006 ). Our criteria for classifying a CAD cell as undergoing neurodegeneration included vesiculation, abnormal microtubule cytoskeleton, and short (or absent) neurites that showed varicosities and increased tortuosity. These are generally accepted phenotypes of dystrophic neurons (Grace et al., 2002 ; Grace and Busciglio, 2003 ).
We note that a large fraction of CAD cells that overexpress APP show normal phenotype and distribution of pAPP. This indicates that high APP levels per se do not generally cause neuronal degeneration. However, high APP levels increase the incidence of neuronal degeneration, a phenotype that occurs concomitant with increased APP phosphorylation by Cdk5, and up-regulation of early endosomes. These results are consistent with a previous report showing that overexpression of APP increases the level of spontaneous degeneration and viability of neuronal cells in culture (Hanson et al., 2003 ). Future studies are required to determine why some neurons degenerate, whereas others are resistant to degeneration, when APP levels are increased.
An important finding of our study is the link between neuronal damage and APP phosphorylation. We found that cell damage is accompanied by 1) an overall increase in APP phosphorylation, 2) a switch in the pathway that leads to APP phosphorylation, and 3) a change of the subcellular compartment containing pAPP. Two main phosphorylation pathways are revealed by this study: the GSK3β/JNK pathway, active in normal, differentiated neurons, and the Cdk5 pathway, active in degenerating neurons. Although it is difficult to test whether the Gsk3β/JNK pathway is turned off when the Cdk5 pathway becomes activated, our data consistently show that inhibition of Cdk5 enhances the GSK3β/JIP-3/JNK pathway of APP phosphorylation, which supports this scenario. Thus, fluctuations in the activity of Cdk5 may fine-tune phosphorylation of APP at Thr668, via distinct pathways.
What is the significance of APP targeting to endosomes, and of phosphorylation in degenerating neurons? It is likely that this represents a mechanism of eliminating excess APP, when APP levels are increased. APP degradation in endosomes may involve the secretases, leading to generation of phosphorylated CTFs. Such a mechanism may account for the increased levels of phosphorylated CTFs detected in Alzheimer's disease brains (Lee et al., 2003 ) and for the overall increase of endocytosis in sporadic forms of Alzheimer's disease (Cataldo et al., 1997 ) and Down's syndrome (Cataldo et al., 2004 ).
The diagram in Figure 8 illustrates the three pathways of APP phosphorylation proposed to be active in normal, stressed, or degenerating neurons. Phosphorylation of APP via the GSK3β/JIP-3/JNK pathway commits the generated pAPP to transport into neurites. The large number of APP-binding, scaffolding proteins present in neurons may additionally regulate APP phosphorylation by stabilizing the protein against degradation, changing its localization, and recruiting specific signaling complexes. During stress, JNK phosphorylates many substrates (including APP), in a cascade that may ultimately block transport by kinesin-1 (Morfini et al., 2006 ). In degenerating neurons, increased Cdk5 activity and altered localization leads to hyperphosphorylation of APP at the endosome. The activation of Cdk5 in degenerating neurons is reminiscent of the activation of cyclin-dependent kinases at mitosis, and could result from a failed attempt of neurons to reenter the cell cycle, as previously proposed (Herrup et al., 2004 ). Taken together, our results indicate that increased activity and mis-localization of Cdk5 may be key events that accompany neuronal injury. Our work thus further validates the hypothesis that aberrant Cdk5 activation may lead to neurodegeneration (Patrick et al., 1999 ; Nguyen et al., 2001 ; Bu et al., 2002 ; Cruz and Tsai, 2004 ; Liu et al., 2004 ; Fischer et al., 2005 ). Although the exact relationship between Cdk5 activity, level of APP, and APP phosphorylation is not known, a recent study suggested that the soluble ectodomain of APP may inhibit Cdk5 activity and confer neuroprotection (Han et al., 2005 ). Future studies are required to clarify the role and significance of aberrant APP phosphorylation during neurodegeneration.
In conclusion, we have shown that phosphorylation of Thr668 of APP occurs under various normal and pathological circumstances, and is the result of activation of distinct signaling pathways and kinases. Although the goal of our study was to broadly define these signaling pathways, future work will reveal their precise molecular details. This will allow the selective blocking or activation of specific signaling cascades that target APP, with potential therapeutic value in Down's syndrome and Alzheimer's disease.
We thank Dr. Dona Chikaraishi (Duke University Medical School, Durham, NC) and Dr. James Wang (Cogent Neuroscience, Inc., Durham, NC) for kindly providing the CAD cell line; Dr. Roger Davis, Dr. Carlos Dotti, Dr. Christoph Kaether, Dr. Ming-Sum Lee (Harvard Medical School, Boston, MA), and Dr. Li-Huei Tsai for kindly providing antibodies and cDNA constructs. This work was supported by National Institutes of Health Grant 5R01GM068596 and March of Dimes Birth Defects Foundation Grant 1-FY04-122. The work was started at Case Western Reserve University School of Medicine, Cleveland, Ohio.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E06-07-0625) on July 18, 2007.
The online version of this article contains supplemental material at MBC Online (http://www.molbiolcell.org).