Notch1 and Amyloid Precursor Protein Are Competitive Substrates for Presenilin1-dependent (cid:1) -Secretase Cleavage*

Proteolytic processing of the amyloid precursor protein (APP) by (cid:2) - and (cid:1) -secretases results in the production of a highly amyloidogenic A (cid:2) peptide, which depos-its in the brains of Alzheimer’s disease patients. Similar (cid:1) -secretase processing occurs in another transmembrane protein, Notch1, releasing a potent signaling molecule, the Notch C-terminal domain. It has been shown that both events are dependent on a presenilin-depend-ent protease. We now test the hypothesis that activated Notch1 and APP are competitive substrates for the same proteolytic activity in neurons. Treatment of neurons with the native Notch ligand, Delta, induces endogenous Notch1 intramembraneous cleavage and diminishes A (cid:2) production in a dose-dependent manner. Complementary experiments showed that the converse was also true. Overexpressing human APP (APP 695Sw ) in neurons leads to a decrease in endogenous Notch1 signal transduction, as assessed by a CBF1 luciferase transcription assay, by Notch C-terminal domain nuclear translocation in vitro and by analysis of Notch C-terminal domain generation and Notch1 staining in vivo . In summary, two complementary approaches suggest that APP and Notch1 are physiologically relevant competitive substrates for (cid:1) -secretase activity. anti-

Amyloid peptide (A␤) 1 is the major component of senile plaques in the brains of Alzheimer's disease (AD) patients. A␤ peptide derives from the sequential proteolytic processing of a single pass transmembrane protein, the amyloid precursor protein (APP) by ␤and ␥-secretases (1)(2)(3). Processing of APP by these secretases is a normal physiological process. Cleavage by ␤-secretase produces a secreted N-terminal APP␤ (sAPP␤) protein and a C-terminal C99 fragment, which can be cleaved by ␥-secretase to produce the 40 -42 amino acid A␤. It has been suggested that a presenilin1 (PS1)-associated enzymatic activity is responsible for the intramembraneous ␥-secretase cleavage of APP (4 -6). Whether PS1 is ␥-secretase itself or is a critical cofactor modulating ␥-secretase activity remains uncer-tain; however, photoaffinity labeling of PS1 by potent ␥-secretase inhibitors indicates that PS1 may contain the active site of ␥-secretase (7,8).
In the present study we explore the possibility that APP and activated Notch1 compete for ␥-secretase activity. We hypothesize that if both APP and Notch1 are processed by the same protease, then activated Notch1 may compete with APP for ␥-secretase cleavage with a consequent decrease in A␤ production. Conversely, we predict that conditions that lead to elevated A␤ production (such as APP 695Sw overexpression) will decrease Notch1 signaling. To test these predictions, we activated endogenous Notch1 by treatment of APP 695Sw -overexpressing neurons with the soluble Notch1 ligand Dl-Fc and showed that the level of A␤ in the conditioned medium was substantially decreased in a dose-dependent manner. In addition, we assessed the activity of the Notch1 signaling pathway in APP 695Sw -overexpressing neurons by counting neurons with nuclear Notch1 staining and by measuring the transactivation of a Notch1 target gene, a CBF1 luciferase reporter. We found that both measures of Notch1 activation were significantly diminished in comparison to that observed in non-APP-overexpressing neurons prepared from control non-transgenic littermates. Finally, we examined Notch1 cellular localization in the hippocampal formation of adult APP 695Sw -overexpressing mice and found a statistically significant decrease in nuclear localization of Notch1, suggesting that Notch1 activation is less efficient in the presence of increased APP expression in vivo. Taken together, these data support the idea that A␤ generation and Notch1 activation are competing biological processes.
were prepared as described previously (31,32). The neurons from individual embryos (E16-18, 10 6 cells/ml) were cultured separately, and tails of the embryos were dissected for genotyping to identify cultures from Tg (expressing human mutant APP) or non-Tg embryos. At 2-6 days in vitro (DIV), prior to treatment with Dl-Fc or control medium, the growth medium in the neuronal cultures was replaced with fresh medium and Dl-Fc or control medium were applied on a daily basis for 2 days.
CBF1 Luciferase Assay-The CBF1-luciferase assay was performed as described previously (11) using a CBF1 luciferase reporter construct (33) and ␤-galactosidase as an internal control for transfection efficiency. The assay was performed in triplicate using an LKB 1251 Luminometer.
Conditioned Medium (CM)-Conditioned medium containing a secreted form of the Notch ligand Delta, Dl-Fc, was collected from cells growing in hygromycin B-free medium for 2 days prior to collection (34). Collected CM was concentrated 4 -5-fold, incubated with an anti-Fc antibody to precluster Dl-Fc (11) and applied to primary neurons. CM from 293T cells, which do not secrete Dl-Fc, served as a negative control and was prepared in a similar way as Dl-Fc CM. 20, 40, or 60 l of CM (Dl-Fc or control) were applied to the primary neurons growing in 200 l of the growth medium daily. 24 h after the last treatment with Dl-Fc/control CM, the total volume of neuronal conditioned medium was adjusted with fresh growth medium to an equal amount in all treated groups. The CM was collected for ELISA (A␤) or Western blots (secreted APP, sAPP), and the cells were lysed to measure the total level of APP expression as well as the amounts of C-terminal (C83 and C99) APP fragments.
ELISA and Western Blots-Cell lysates and CM were adjusted to equal protein concentrations for the analyses. For A␤ ELISA we used 25 l of tissue culture-conditioned medium. The capture antibody was 22C4, (to the C terminus of A␤), and the detection antibody was biotinylated 6E10 (to A␤ residues 1-17) (35). For the Western blot analysis, the CM was electrophoresed on 10 -20% Tricine gel for A␤ analysis or on 4 -20% Tris-Glycine gel for sAPP. SDS-polyacrylamide gel electrophoresis of the cell extracts was carried out on 4 -20% Tris-glycine gel for total APP expression and for C-terminal fragments of APP (C83 and C99). The cells (brains) were lysed in a buffer containing 50 mM TRIS, 150 mM NaCl, 2 mM EDTA, and 1% Nonidet P-40. The immunoblotting was performed with the following anti-APP antibodies (C-terminal 13G8 and C8 (to residues 676 -695), N-terminal 8E5 (to residues 444 -592), A␤-specific 6E10, and ␤ cut-specific 192sw (36,37) or anti-Notch1 antibody, TC.
Antibodies-192sw, 13G8, and 8E5 antibodies were a gift from Elan Pharmaceuticals, San Francisco, CA. C8 antibody was a gift from D. Selkoe, Brigham and Women's Hospital, Boston, MA (38). 22C4 was a gift from R. Nitsch (University of Zurich, Switzerland) (35). 6E10 antibody was from Senetek Research, St. Louis, MO. An anti-biotin antibody was purchased from Roche Molecular Biochemicals, Germany. TC was a polyclonal antibody against the C terminus of Notch1 (TC, courtesy of J. Aster, Brigham and Women's Hospital, Boston, MA, Ref. 39).
Biotinylation-To label cell surface molecules, growing primary neurons (treated with either control or Dl-Fc-containing conditioned medium) were incubated with EZ-Link Biotin (EZ-Link Sulfo-NHS-LC-Biotin, Pierce) for 45 min at 4°C. The cells were lysed, biotinylated cell surface proteins were immunoprecipitated with an anti-biotin antibody, electrophoresed on a Tris-glycine gel, probed with 8E5, and developed using enhanced chemiluminescence (ECL, Amersham Pharmacia Biotech). Quantitative analysis was performed using Molecular Analyst Software on a Bio-Rad Densitometer.
Assessment of the Nuclear Translocation of Endogenous Notch1 upon Ligand Binding in APP 695Sw -overexpressing and Non-overexpressing Control Littermate Neurons in Vitro-Primary neuronal cultures were obtained from individual embryos of heterozygote Tg 2576 mice; genotyping subsequently identified cultures from transgene-positive or transgene-negative embryos. Neurons at 2-6 DIV were treated with control CM or CM containing preclustered Notch1 ligand Dl-Fc for 24 h, fixed in 4% paraformaldehyde for 10 min, permeabilized with 0.5% Triton X-100, and blocked with 1.5% normal goat serum for 1 h. Cells were incubated with Notch1-specific TC antibody and a neuron-specific monoclonal MAP2 antibody (Sigma) overnight at 4°C. Fluorochromelinked secondary antibodies (CY3 anti-rabbit and fluorescein antimouse (Jackson)) were applied for 1 h at room temperature. Cultures were washed with Tris-buffered saline, counterstained with Hoescht 33342 (Molecular Probes) for 5 min at room temperature to visualize nuclei, and coverslipped using GVA mounting solution (Zymed Laboratories Inc.).
The images of all neurons (confirmed by MAP2 staining) immuno-stained with anti-Notch1 antibody in 5-6 random visual fields per well were collected using ϫ 40 objective (zoom 2) on a confocal microscope (Bio-Rad 1024) mounted on a Nikon Eclipse TE300 inverted microscope. Four wells per condition were examined in each experiment, and data were collected from two independent experiments. The percent of neurons containing nuclear Notch1 staining was recorded by an observer unaware of transgene status of the culture and was then calculated for APP 695Sw -overexpressing and non-transgenic littermate control cultures. A total of 500 neurons were quantified.

Analysis of Notch1 Cellular Localization in Vivo-
The sections of five-month-old Tg 2576 and control non-Tg mouse brains were immunostained with Notch1 antibody (TC) and counterstained with DAPI to localize the nuclei of the neurons. The location of the nucleus was identified in the DAPI channel, then a "mask" was applied to the Notch1 channel and the intensity of fluorescence within the nucleus was measured using a computerized image analysis system (Bioquant, Nashville, TN). The location of neurons within the CA1 field of the hippocampus to be measured was chosen using a systematic random sampling scheme analogous to those used in stereology; ϳ2,000 neurons were examined using a ϫ 100 water immersion objective. The ratio of Notch1 nuclear fluorescence to total Notch1 fluorescence in the CA1 field was compared in three Tg2576 and three non-Tg control littermate mice using Student's t test analysis.

RESULTS
APP Competes with Notch1 for ␥-Secretase-To test the hypothesis that APP competes with Notch1 for ␥-secretase processing, we activated endogenous Notch1 signaling in primary neurons by treatment with different concentrations of the Notch ligand, Dl-Fc, for 2 days and measured the amount of A␤ in conditioned medium. Treatment with Dl-Fc, but not control CM, appeared to cause a significant dose-dependent decrease in total A␤, as assessed by Western blot analysis (Fig. 1A). To confirm this observation, we used a sensitive and specific sandwich ELISA to quantitate total A␤. We found consistent, correlative decreases of total A␤ after treatment with Dl-Fc in a dose-dependent manner, with the highest dose leading to a decrease of about 45% (p Ͻ 0.001) in A␤ production (Fig. 1B). Results are representative of nine independent experiments, each performed at least in duplicate.
We next examined whether the decrease in A␤ secretion after Notch1 activation with Dl-Fc was due to any effects of this treatment on APP expression, metabolism or trafficking. We performed Western blot analysis of cell lysates (for APP expression and ␥-secretase cleavage) and conditioned medium (for secreted APP␣ and APP␤). Three different antibodies to APP were used to analyze the level of APP expression: C-terminal 13G8, N-terminal 8E5, and A␤-specific 6E10 antibodies. There was no change in the level of total APP expression in Dl-Fc treated neurons, in comparison to that in cells treated with control medium. The result was replicated in four independent experiments ( Fig. 2A).
We also analyzed whether treatment with Dl-Fc alters trafficking of APP molecules to the cell surface. To label cell surface proteins we biotinylated the surface molecules of the primary neurons with EZ-Link Biotin. Cell lysates were immunoprecipitated with an anti-biotin antibody, and immunoblotted with 8E5 to detect cell surface APP. There was no difference in the amount of biotinylated APP in Dl-Fc-treated neurons in comparison to that in control CM-treated cells (Fig. 2E). We also measured the amount of secreted APP in the conditioned medium. The level of sAPP (total, ␤-, or ␣-sAPP) was not different in neurons treated with Dl-Fc or control CM (Fig. 2, B and C). This demonstrates that the activity of ␣and ␤secretases was not changed due to activation of Notch1 with its ligand Delta. The amount of APP C-terminal fragments (CTF) also was not significantly different between Delta and control mediumtreated cells (Fig. 2D). This observation was reproduced in five independent experiments. The discordance between A␤ production and APP CTF generation in this system is in agreement with observations on APP processing after manipulation of nicastrin (40) and PS1 (41).
Thus, the major effect of Notch1 activation on APP was reduced A␤ generation. This result suggests that activation of Notch1, which leads to ␥-secretase-mediated cleavage of its C terminus, effectively decreases ␥-secretase action on APP. Notch1 Competes with APP for ␥-Secretase-We next performed the converse experiment. We reasoned that if ␥-secretase activity is a limiting step, overexpression of APP might inhibit the efficiency of ligand-induced Notch1 cleavage/nuclear translocation/signal transduction by competing with Notch1 for the ␥-secretase activity. Thus, we compared the amount of endogenous Notch signaling in response to its physiological ligand, Delta, in primary neuronal cultures derived from APP 695Sw -overexpressing mice and non-transgenic littermate controls. Neurons in which Notch1 is activated show a predominantly nuclear pattern of staining reflecting translocation of the C-terminal proteolytic product, NICD, to the nucleus ((21) ; Fig. 3, A and B). The percent of neurons with nuclearactivated Notch1 was reduced by more than 50% (p Ͻ 0.001) in neurons from APP 695Sw transgenic mice compared with that in neurons from non-transgenic littermates (Fig. 3C). Interestingly, the total level of Notch1 immunoreactivity in APP 695Swoverexpressing neurons was slightly higher than that in normal control littermates, perhaps reflecting diminished processing (Fig. 3, A and B).
To confirm that Notch1 signaling was indeed down-regulated in APP-overexpressing neurons we also measured activation of a Notch1 downstream transcription factor, CBF1, in the neuronal cultures. We co-transfected neurons prepared from control and APP-overexpressing transgenic mice with a CBF1 luciferase reporter construct (33) and ␤-galactosidase, an internal control for transfection efficiency, and measured luminescence caused by Dl-Fc treatment 24 h post-transfection. There was a 50% reduction (p Ͻ 0.001) in CBF1 luciferase activity in neurons prepared from transgenic embryos overexpressing APP 695Sw in comparison to that in control, non-transgenic neurons (Fig. 3D).
NICD Generation and Notch1 Nuclear Translocation Are Decreased in the Brain of APP-overexpressing Mice-To test the plausibility of the hypothesis that APP and Notch1 compete in the brain, we performed double-immunofluorescent staining of the brain sections with anti-APP (8E5) and anti-Notch1 (TC) antibodies. As expected from studies of each of these proteins individually, there was substantial co-localization of Notch1 and APP in neuronal cells. To test whether evidence for competition between APP and Notch1 for ␥-secretase can be detected in vivo, we analyzed the NICD generation in the brains of APP Tg and non-Tg mice by Western blot analysis. To overcome the potential problem of uneven loading or protein degradation, the ratio of the optical density of the ␥-secretasecleaved NICD band to that of the furin-cleaved band (NICD/ furin ratio) was measured. There was a significant decrease in the NICD/furin ratio in the brain of APP-overexpressing mice, compared with that of control, non-Tg littermates (p Ͻ 0.02; Fig. 4A) To further test the hypothesis about competition between APP and Notch1 we measured the intensity of nuclear Notch1 immunostaining in the CA1 area of the hippocampus in APP Tg and non-Tg mice. The ratio of Notch1 fluorescence in the nucleus to total Notch1 fluorescence in the CA1 area was significantly lower in APP Tg compared with that in non-Tg mice (p Ͻ 0.001), which suggests that there is less Notch1 activation in APP-overexpressing neurons in vivo (Fig. 4B). DISCUSSION There is a striking similarity between the proteolytic processing of the AD-related transmembrane protein APP and the Notch1 receptor. The latter is an important protein involved in cell-fate decisions during development (14, 16, 42) but continues to be expressed in the adult brain (43), with effects on neuronal plasticity (20,21) and on glial (oligodendrocyte) differentiation (34). Both proteins undergo extracellular proteolytic event(s), which precede an intramembraneous cleavage. The last processing event appears to be carried out by an identical proteolytic activity, because both APP and Notch1 cleavage are highly dependent on PS1 activity, both are blocked by several APP specific ␥-secretase inhibitors (5,11,22), and both are affected by dominant negative PS1 mutations (5,11). We now address the question of whether APP and Notch1 compete for the same protease in a physiologically relevant way. We used two complementary approaches. 1) We compared the amount of ␥-secretase processing of APP (by analyzing the secretion of A␤ into the conditioned media) after activation of endogenous Notch1, and 2) we compared the response of the Notch1 signaling pathway to stimulation with a physiological ligand in neurons overexpressing human mutant APP (APP 695Sw ) and in normal, non-transgenic neurons from littermate controls.
Knowing that Notch1 becomes a ␥-secretase substrate only when activated with its ligand Delta, we used increasing doses of the ligand to modulate the amount of one ␥-secretase substrate (Notch1) leaving the amount of a competitive substrate FIG. 3. Notch1 signaling in APP 695Sw -overexpressing and control non-transgenic primary neurons. A and B, confocal microscope images of the primary neurons prepared from non-transgenic littermate (A) and APP 695Sw -overexpressing (B) mice, treated for 24 h with 60 l of Dl-Fc and immunostained with Notch1 antibody. Arrows indicate nuclear Notch1 localization. Bar represents 50 mkm. C, quantitative analysis of the neurons with distinct nuclear Notch1 immunoreactivity as a percentage of total cells. The bars represent mean Ϯ S.D. *, p Ͻ 0.001; Student's t test. D, Notch1 signaling is significantly reduced in huAPP 695Sw -overexpressing primary neurons. To measure transcriptional activation of a CBF1 luciferase reporter in neurons, the cells were co-transfected with 0.5 g/100 l CBF1 firefly luciferase reporter (33), and 0.5 g/100 l of constitutively active CMV-␤-galactosidase as an internal control of transfection efficiency. The ratio of firefly luciferase activity to ␤-galactosidase activity normalizes CBF1 luciferase activity to the efficiency of transfection in every experiment. The bars represent luciferase activation in APP transgenic and non-transgenic neurons treated with Dl-Fc (mean Ϯ S.D.; n ϭ 3, *, p Ͻ 0.01; Student's t test). Results are representative of three independent experiments.
(APP) constant. We found that stimulation of the endogenous Notch1 receptor results in a significant decrease in the level of total A␤ secretion in a dose-dependent manner. Control experiments showed that the effect was not mediated by any effect on APP levels or trafficking, because our data show no effect of Notch1 activation on the level of expression of total APP, cell surface APP, or ␣and ␤-secretase products. These results are in accord with data in another system showing that APP processing by ␣-and ␤-secretases is independent of PS1/␥-secretase activity (6). Moreover, whereas A␤ production was significantly impaired, there was only a subtle difference, not statistically significant, in the amount of C83 and C99 APP CTF(s) in the cell lysates as detected by Western blot analysis 24 h after the last application of Dl-Fc. The discordance between A␤ production and the accumulation of APP C-terminal fragments is similar to that observed for neuroblastoma cell lines stably transfected with PS1 deletion mutants (41). Recent data on manipulation of nicastrin, a presenilin interactor, showed changes in A␤ production without alteration of C83/C99 levels (40), also suggesting that A␤ production can be dissociated from C83/C99 changes, especially in the circumstances where the effect on A␤ production of the manipulation is more subtle then PS1 ablation or use of ␥-secretase inhibitors.
We do not rule out the possibility that there may be other substrates for ␥-secretase besides APP and Notch1 or that there could be other molecules involved (such as nicastrin, for example) that modulate the ␥-secretase cleavage of various substrates. There also could be indirect effects of Notch activation on APP processing or on A␤ clearance. However, our experiments clearly demonstrate that activation of Notch by its native ligand specifically resulted in diminished A␤, without major changes in APP amount or subcellular distribution. Similarly, overexpression of APP specifically affected Notch cleavage/nuclear translocation/signaling but did not significantly change the level of Notch expression or its subcellular distribution. These observations strongly suggest that Notch and APP are competitive substrates for ␥-secretase activity.
The complementary experiment is also consistent with the hypothesis that ligand-activated Notch1 competes with APP for ␥-secretase activity. Using both morphological (nuclear translocation) and physiological (CBF1 transactivation) quantitative assays we found that endogenous Notch1 signaling was significantly diminished in primary neurons overexpressing APP. In addition, analyzing nuclear (activated) Notch1 staining in the brain of APP overexpressing mice revealed that this effect can be detected in vivo. We suggest that an excess of APP shifts the enzymatic activity "equilibrium" toward APP processing and thus diminishes Notch1 from being cleaved and from activating the Notch signaling pathway; i.e. that activated Notch1 and APP are competitive substrates for ␥-secretase processing.
The relationships among APP, PS1-related ␥-secretase activity, and Notch1 are complex and in some ways have appeared to be contradictory, but may be clarified in the context of our current studies. PS1 familial AD (FAD) mutations are gain-offunction in terms of APP processing resulting in increased production of (especially) A␤ 42 peptide (44 -46). However, FAD mutations in PS1 show partial loss-of-function in terms of Notch1 function as assessed by measuring the generation of the C-terminal Notch1 signaling domain (NICD) (24,47), Notch1 signaling (CBF1-luc activation assay, Ref. 48), neurite outgrowth (31), or rescue of the sel-12 mutant phenotype in Caenorhabditis elegans (49 -51). PS1 mutations that increase A␤ production the most resulted in elimination of Notch1 proteolysis (47). In contrast, PS1 deficiency (6,27) or aspartate to FIG. 4. Analysis of Notch1 signaling in vivo in APP 695Sw -overexpressing mice. A, Western blot analysis of NICD generation (inset) reveals that the ratio of the NICD band (*) to the furin-cleaved band (**) is significantly decreased in the brains of APP-Tg mice compared with that in control, non-Tg littermates (graph). The NICD lane represents cells transfected with NICD portion of the Notch1 molecule as a control (mean Ϯ S.D.; n ϭ 3, *, p Ͻ 0.02; Student's t test). B, the intensity of nuclear Notch1 immunoreactivity in the neurons of the CA1 area of the hippocampus is represented as a ratio of photon counts in the nuclei to the photon counts in the CA1 area. Control mice show higher level of nuclear Notch1 than APP-Tg mice (mean Ϯ S.D.; n ϭ 2,000 neurons, **, p Ͻ 0.001; Student's t test).

FIG. 5. Schematic representation of the competition between APP and
Notch1 for PS1-dependent protease activity. A, control cell. B, activation of Notch1 cleavage/signaling after binding of Notch ligand, Delta, leads to a decreased APP processing. C, overexpression of APP 695Sw results in a decreased Notch1 cleavage/signaling. D, are FAD mutations in PS1 "gain-of-function" in terms of APP processing, and loss-of-function in terms of Notch intramembraneous cleavage/signaling? alanine interchanges at positions Asp-257 and Asp-385 in PS1 (4,11) resulted in a decrease of both A␤ production and NICD generation/Notch signaling (although substitution at Asp-257 in another system did not alter A␤ production but inhibited NICD generation from a constitutively active Notch1 construct, Ref. 52). Similarly, Petit et al. (53) have described ␥-secretase inhibitors that diminish A␤ without altering Notch cleavage (53).
Based on our current data we propose that Notch1 and APP are competitive substrates for ␥-secretase activity. If so, PS1 FAD mutations could be interpreted as altering the relative affinities of ␥-secretase for APP and Notch1 to favor APP over Notch1 (Fig. 5). This also predicts the possibility that some agents could differentially affect APP and Notch1 processing by ␥-secretase, e.g. by inhibiting access of one or the other substrate, providing a possible explanation for the recent results of Petit et al. (53). This formulation (Fig. 5) predicts that relative levels of Notch1 and APP expressed in cells, as well as the degree to which Notch1 is activated by ligand, could impact A␤ synthesis. Because Notch1 expression diminishes markedly with age (54), there may be a subtle increase in A␤ production with age as a result of the release of APP/␥-secretase (PS1) from this substrate competition. However, the present study (using APP-overexpressing mice) and the fact that mice heterozygous for PS1ϩ/Ϫ or Notch1ϩ/Ϫ have no obvious pathologic phenotype, suggest that 50% of Notch1 activity is enough to maintain a normal phenotype. This might be an important assumption for the use of drugs inhibiting A␤ production, which have inhibition of Notch1 signaling as a side effect.