Tumor Necrosis Factor- , Interleukin-1 , and Interferon- Stimulate -Secretase-mediated Cleavage of Amyloid Precursor Protein through a JNK-dependent MAPK Pathway*

The deposition of the amyloid (A ) peptide in neuritic plaques plays a critical role in the pathogenesis of Alzheimer’s disease (AD). A is generated through the proteolysis of amyloid precursor protein (APP) by the sequential actions of and -secretases. Although recent evidence has unveiled much about the biochemical identity and characteristics of -secretase, the mechanism regulating endogenous -secretase activity remains elusive. To identify possible extracellular signals and associated signaling cascades that could regulate APP proteolysis by -secretase activity, we have developed a cell-based reporter gene assay by stably cotransfecting HEK293 cells with the Gal4-driven luciferase reporter gene and the Gal4/VP16-tagged C-terminal fragment of APP (C99-GV), the immediate substrate of -secretase. The cleavage of C99-GV by -secretase releases the transcription factor that activates luciferase expression, providing a quantitative measurement of -secretase activity. Using this reporter assay, we have demonstrated that interferon, interleukin-1 , and tumor necrosis factorcan specifically stimulate -secretase activity, concomitant with increased production of A and the intracellular domain of APP (AICD). The -secretase-dependent cleavage of Notch is also enhanced upon the stimulation of these cytokines. The cytokine-enhanced -secretase activity can be suppressed by a potent inhibitor of c-Jun N-terminal kinase (JNK). Furthermore, cells transfected with dominantpositive MEKK1, one of the most potent activators of the JNK cascade, exhibit increased -secretase activity, suggesting that the JNK-dependent mitogen-activated protein kinase pathway could mediate the cytokine-elicited regulation of -secretase. Our studies provide direct evidence that cytokine-elicited signaling cascades control A production by modulating -secretase activity.

disease (AD) (1). A␤ is generated through the proteolysis of the amyloid precursor protein (APP) by the sequential actions of ␤and ␥-secretases (2). APP can be first cleaved by ␤-secretase that cuts at the 28-residue N-terminal to the transmembrane domain of APP (3), releasing a large soluble fragment (␤-APPs) into the lumen/extracellular space and retaining a 99-residue C-terminal fragment (C99-CTF) in the membrane. This membrane-tethered C99 can then be further processed by ␥-secretase through an unusual cleavage that apparently occurs within the transmembrane domain of C99, thus generating the A␤ peptides and a cytoplasmic tail dubbed APP-intracellular domain (AICD). Accumulated evidence has demonstrated that causative mutations linked to APP and the presenilins (PS1 and PS2) found in familial early-onset AD patients all result in increased production of the more amyloidogenic 42-residue A␤ isoform (A␤42) versus the soluble 40-residue isoform (A␤40). Both ␤and ␥-secretases have thus become important therapeutic targets.
Tremendous progress has been made toward the identification of these secretases. Whereas ␤-secretase appears to be a novel membrane-associated aspartyl protease (4), accumulated evidence has supported the notion that PSs are obligatory components of ␥-secretase (5). The discovery that presenilins are probably the proteases catalyzing the last step in A␤ generation is consistent with the idea that all FAD-causing mutations are localized either to positions in APP near the cleavage sites for A␤ production or to genes encoding the protease that generates A␤ (presenilin/␥-secretase). Its pivotal role in A␤ generation and AD pathogenesis is further demonstrated by the analysis of FAD-associated PS missense mutations that have been shown to intimately correlate with the selective elevation of A␤42 versus A␤40 in human, cultured cells, and transgenic mice (6). Biochemical and genetic analyses further reveal that PSs are part of a multimeric ␥-secretase complex. PS heterodimers co-migrate with a high molecular mass of ϳ250 kDa through density gradients (7), and detergent-solubilized ␥-secretase activity can be eluted by size exclusion chromatography with an estimated molecular mass of Ն2 MDa (8). Blue-native PAGE analysis of the ␥-secretase complex has given molecular masses ranging from 250 to 500 kDa (9 -12). By using an inhibitor-immobilized matrix, ␥-secretase activity can be isolated and co-purified with PS heterodimers in addition to another PS-associated membrane protein called nicastrin (13,14). Most recently, two independent studies using genetic screening in Caenorhabditis elegans have identified two other membrane proteins, Aph-1 and Pen-2, as novel obligate members of ␥-secretase (15,16). Furthermore, compelling evidence has shown that the full spectrum of ␥-secretase activity can be reconstituted by the co-expression of human presenilin, nicastrin, Aph-1, and Pen-2 (9,17), providing definitive proof for the minimal required constituents of the physiologically active ␥-secretase. These presenilin cofactors have been shown to play distinct functions in the formation of the high molecular weight enzyme complex, as Aph-1 stabilizes the presenilin holoprotein in the complex and Pen-2 sustains ␥-secretase activity by enforcing the endoproteolysis of presenilin (18). The identification of these presenilin cofactors presents additional objects for the intervention of ␥-secretase activity and A␤ production.
The endogenous mechanism regulating ␥-secretase activity remains elusive, despite the enormous progress in biochemical characterization of this protease in recent years. Previous studies have gathered substantial evidence demonstrating the close correlation between inflammation and A␤-centered pathogenesis of AD. A variety of inflammatory mediators, including interleukin-1␤ (IL-1␤), IL-6, tumor necrosis factor-␣ (TNF-␣), and transforming growth factor-␤, are up-regulated in AD patients versus nondemented individuals and have been suggested to promote the synthesis and processing of APP. In addition, marked increase of A␤40 and A␤42 is observed in primary astrocytes or astrocytoma cells following stimulation with a combination of interferon-␥ (IFN-␥) and TNF-␣ or IFN-␥ and IL-1␤ (19,20), suggesting that inflammatory cytokines could regulate the processing of APP through cytokine-elicited signaling pathways. Indeed, a recent study has demonstrated that platelet-derived growth factor can enhance the ␤-␥-secretase-mediated proteolysis of APP through a Src-Rac-dependent pathway (21), suggesting that endogenous brain A␤ levels could be under stringent regulation mediated by a variety of inflammatory molecules. Here, we developed an efficient and quantitative assay and use it to obtain evidence for the first time that proinflammatory cytokines, including IFN-␥, IL-1␤, and TNF-␣, are potent stimulators of ␥-secretase, resulting in the increased production of A␤ through a pathway involving the c-Jun N-terminal kinase (JNK)-dependent MAPK pathway.

EXPERIMENTAL PROCEDURES
Reagents-BCA protein assay reagent kit and SuperSignal West Femto reagents were purchased from Pierce. Rabbit anti-Notch(Val 1744 ) antibody, anti-stress-activated protein kinase/JNK MAP kinase antibody, and mouse anti-phospho-stress-activated protein kinase/JNK antibody were from Cell Signaling Technology, Inc. (Beverly, MA). Horseradish peroxidase-conjugated anti-rabbit IgG was from Santa Cruz Biotechnology, Inc. Horseradish peroxidase-conjugated anti-mouse IgG and ECL Western blotting detection reagents were from Amersham Biosciences. FuGENE 6 transfection reagent, Expand long template PCR system, and the PCR nucleotide mixture were from Roche Applied Science. Dual luciferase assay reagents, Steady-Glo luciferase assay reagents, and pRL-TK vector were from Promega. The QuikChange site-directed mutagenesis kit was from Stratagene. The human A␤40 colorimetric ELISA kit was from BioSource International Inc. Interferon-␥, interleukin-1␤, and tumor necrosis factor-␣ were from PeproTech Asia (Rehovot, Israel). All other reagents were at least reagent grade and obtained from standard suppliers.
To generate a Gal4-driven luciferase reporter gene construct that is suitable for stable transfection, pFR-Luc vector was first cleaved with StyI, and the protrusive ends were converted to blunt ends by Taq DNA polymerase. An additional cleavage of the linearized vector by NdeI released the complete open reading frame of firefly luciferase along with the upstream promoter plus five tandem repeats of Gal4 binding motif that was then purified and ligated with pBudCE4.1 vector (Invitrogen) previously cleaved by NdeI and ScaI. The residual cytomegalovirus promoter sequence carried over from the original pBudCE4.1 vector was removed by the cleavage of NdeI and SpeI, followed by the conversion of protrusive ends to blunt ends by Taq DNA polymerase and subsequent self-ligation to generate a novel Gal4 promoter-driven luciferase reporter construct (Gal4-Luc) containing a zeocin-resistant selection marker. In transient transfection, a vector constitutively expressing Renilla luciferase, pRL-TK, was included as an internal control to normalize the transfection variation.
A Notch expression construct encoding the mouse Notch1 (residues 1704 -2184) was derived from a vector expressing constitutively activated membrane-bound Notch (N⌬E) in which a methionine (Met 1727 ) residue within its transmembrane domain was replaced with valine (a gift of Dr. Raphael Kopan, Washington University, St. Louis, MO (22,23)) by PCR. The sequences of the forward and reverse primers used in PCR were 5Ј-CCCTCTAGACAGGCATGCCACGGCTCCTGACGCCCC-TTC-3Ј (Notch-S10-XbaI) and 5Ј-CCCGGATCCCGACGAGCTGTCCAA-CAGCCAGCCCTTGCC-3Ј (Notch-A3-BamHI), respectively. Underlined bases denoted the appended restriction sites. PCR amplimers were purified, cleaved with BamHI and XbaI, and subcloned into pBudCE4.1 vector to generate the N⌬E-pBud vector.
Cell Culture-Human embryonic kidney cells (HEK293) and COS7 were maintained in DMEM supplemented with 10% fetal bovine serum (FBS) and 0.1 mg/ml penicillin and streptomycin. T-REx293 cells were purchased from Invitrogen and cultured in DMEM supplemented with 10% FBS and 5 g/ml blasticidin. Cells were incubated in a humidified incubator at 37°C in 5% CO 2 .
Transient Transfection of Mammalian Cells-Transfection of HEK293, COS, and T-REx293 cells was performed using FuGENE 6 transfection reagent. Cells were plated onto 6-well microplates and grown to about 60 -80% confluency prior to transfection. On the day of transfection, the culture medium was replaced with fresh 2 ml/well of DMEM supplemented with 10% FBS. C99-GV (0.5 g), pFR-Luc (0.5 g), and pRL-TK (0.1 g) were diluted in 100 l of serum-free DMEM and mixed with 3.3 l of FuGENE 6, followed by incubation at room temperature for 15 min. Transfection mixtures were added dropwise into cell culture medium and incubated at 37°C for 24 h. Transfected cells were harvested by PBS containing 20 mM EDTA and lysed with 100 l of 1ϫ passive lysis buffer (Promega). Cell lysates were clarified by centrifugation, and subjected to luciferase assay using the Dual luciferase assay reagent kit. Protein content of lysates was determined by the BCA protein assay reagent kit.
Generation of Stably Transfected Cell Lines-T-REx293 cells were grown in 10-cm dishes until 50% confluency as described above. On the day of transfection, the culture medium was replaced with 8 ml of DMEM containing 10% FBS. T-REx293 cells were transfected with equal amounts of C99-GV (5 g) and Gal4-Luc (5 g) by FuGENE 6 transfection reagent according to the manufacturer's instructions. Transfected cells were cultured in DMEM supplemented with 10% FBS, 200 g/ml hygromycin, 250 g/ml zeocin, and 5 g/ml blasticidin (DMEM-HZB), and single colonies resistant to antibiotic selection were isolated individually using cloning cylinders. Each of independent cell lines was screened for the tetracycline-induced expression of C99-GV and corresponding luciferase signals that can be attenuated by ␥-secretase inhibitors. Individually isolated cell lines were plated into 96-well microplates in DMEM-HZB containing 5 g/ml tetracycline in the presence or absence of 10 M compound E, a potent ␥-secretase inhibitor, for 24 h. Cell lysis and the addition of reagents for luciferase assay were completed simultaneously using the Steady-Glo luciferase assay system (Promega) as described in the manufacturer's instructions. Luciferase signals were determined by a MLX microplate luminometer (DYNEX Technologies, Inc.). The optimal cell line was the one exhibiting the highest tetracycline-induced luciferase signal that can be significantly suppressed by compound E down to the basal level comparable with the luciferase signal obtained in the absence of tetracycline induction. Based on this criterion, clone T20 was selected and optimized for the cell-based ␥-secretase assay (Fig. 1).
To generate cells constitutively expressing N⌬E, HEK293 cells were transfected with N⌬E-pBud (10 g) by FuGENE 6 transfection reagent as described above. Transfected cells were selected by DMEM supplemented with 10% FBS and 250 g/ml zeocin (DMEM-Zeo), and single colonies were isolated individually using the cloning cylinder method. Lysates of selected cell lines were generated and analyzed for the overexpression of N⌬E by SDS-PAGE and Western blotting using anti-Notch(Val 1744 ) polyclonal antibody. Clone N7 was shown to express significant amounts of N⌬E that can be cleaved by ␥-secretase, and the treatments of ␥-secretase inhibitors, both compound E and DAPT, dramatically suppressed this cleavage, demonstrating the efficiency of this Notch-based assay for ␥-secretase (Fig. 4).
Cell-based ␥-Secretase Assays-To examine cytokine-elicited effects on ␥-secretase specifically, T20 cells stably transfected with C99-GV and Gal4-Luc were trypsinized and washed with serum-free DMEM prior to plating onto 12-well microplates in 1 ml/well serum-free DMEM at 5 ϫ 10 5 cells/well. Following the incubation at 37°C overnight, cells were treated with 50 ng/ml IFN-␥, IL-1␤, or TNF-␣ in DMEM containing 1 g/ml tetracycline, the inducer of C99-GV expression, and incubated at 37°C for various intervals as specified. For dose-response studies, various concentrations of cytokines were included in the medium as specified. Cells incubated with serum-free DMEM containing 1 g/ml tetracycline alone were used to define the basal level of ␥-secretase activity at each time point, whereas cells treated with DMEM without tetracycline were used to estimate the nonspecific background emission of the luciferase signal. To terminate reactions, cells were harvested with PBS containing 20 mM EDTA and lyzed in 100 l of 1ϫ passive lysis buffer (Promega). Cell debris was removed by centrifugation at 13,200 ϫ g for 5 min, and luciferase activity in clarified lysates was determined by mixing 20 l of lysates and 20 l of Steady-Glo luciferase assay reagent in a 96-well LumiNunc microplate (Nunc). Following incubation at room temperature for 5 min, emitted luminescence in individual microwells was determined by a MLX Microplate luminometer and subsequently normalized by the protein content of lysates. The protein content of clarified lysates was determined by the BCA protein assay kit as described in manufacturer's instructions. The normalized luciferase signal emitted by T20 cells in serum-free DMEM without tetracycline was referred to as 1-fold of activation. For the time course studies, T20 cells were treated with 50 ng/ml of the respective cytokines in serum-free DMEM containing 1 g/ml of tetracycline, incubated at 37°C for various intervals, and harvested for luciferase assay. For treatments of chemical compounds, such as compound E, DAPT, MAP kinase inhibitors, compounds were added into medium to the final concentration of 10 M or as specified.
To examine cytokine-elicited stimulation on ␥-secretase-dependent cleavage of Notch, N7 cells constitutively expressing N⌬E were plated onto 12-well microplates in 1 ml/well of serum-free DMEM at 5 ϫ 10 5 cells/well. Following incubation at 37°C overnight, cells were treated with 5 or 50 ng/ml IFN-␥, IL-1␤, or TNF-␣ in DMEM and incubated at 37°C for 9 h. Treated cells were harvested by PBS containing 20 mM EDTA and dissolved in 50 l of 1ϫ passive lysis buffer, followed by centrifugation at 13,200 ϫ g for 5 min to remove cell debris. The protein concentrations of clarified supernatants were determined by the BCA protein assay reagent kit, and cell extracts containing equivalent amounts of proteins were resolved by SDS-PAGE and analyzed by Western blotting using anti-Notch(Val 1744 ) polyclonal antibody as described below.
A␤ ELISA-To determine cytokine-stimulated A␤ production, T20 cells (5 ϫ 10 5 cells/well) were seeded onto 12-well tissue culture plates and treated with various amounts of cytokines in serum-free DMEM containing 1 g/ml tetracycline, followed by incubation at 37°C for 9 h. Conditioned media were harvested, clarified by centrifugation, supplemented with the Complete® protease inhibitor mixture, and stored at Ϫ80°C until ready for assay. Levels of secreted A␤40 in conditioned media were determined using a quantitative human A␤40 sandwich ELISA kit as described in the manufacturer's instructions. A␤40 measurements were conducted in triplicate, and the medium of T20 cells treated with serum-free DMEM alone was included as the blank.
Site-directed Mutagenesis of MEKK1-The MEKK1 expressing vector (pCMV-MEKK1) encoding the catalytic kinase domain of wild-type mouse MEKK1 was purchased from Clontech. To generate the dominant-negative forms of MEKK1 (MEKK1-T560A or MEKK1-T572A (24)), two essential threonine residues within its catalytic domain, Thr 560 and Thr 572 , were replaced with alanine, respectively, using the QuikChange site-directed mutagenesis kit according to the manufacturer's instructions. Two complementary oligonucleotides containing the desired mutations, flanked by unmodified nucleotide sequences, were synthesized. The paired primers for the generation of mutant MEKK1-T560A were T560A-forward (5Ј-TTGGCATCAAAAGGAGCCGGTGCAGGAGAGT-3Ј) and T560A-reverse (5Ј-ACTCTCCTGCACCGGCTCCTTTTGATGC-CAA-3Ј), whereas the ones for MEKK1-T572A were T572A-forward (5Ј-GGACAGTTACTGGGGGCAATTGCATTCATGG-3Ј) and T572Areverse (5Ј-CCATGAATGCAATTGCCCCCAGTAACTGTCC-3Ј). Underlined nucleotides denote the single based changes to incorporate the desired missense mutations. The dominate-negative kinase activity of mutants T560A and T572A was verified by using the In Vivo kinase assay (Clontech) according to the manufacturer's instructions.
SDS-Polyacrylamide Gel Electrophoresis and Western Blot Analysis-Clarified cell extracts containing equivalent amounts of proteins were mixed with equal volumes of 2ϫ SDS sample buffer (125 mM Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, 2% dithiothreitol, and 5% ␤-mercaptoethanol) and boiled at 100°C for 5 min. Denatured proteins were resolved in Tris glycine polyacrylamide gels (10 or 12%). Separated proteins were transferred electrophoretically to polyvinylidene difluoride membranes (Immobilon-P, Millipore). Membranes were then treated with blocking buffer (5% nonfat dry milk, 0.1% Tween 20 in PBS (PBST)) at room temperature for 1 h, followed by a brief rinse with PBST. Blocked membranes were probed with appropriate dilutions of primary antibody in PBST at room temperature for 1 h. The unbound primary antibody was removed by extensive washes with PBST. Subsequently, washed membranes were incubated with appropriate horseradish peroxidase-conjugated secondary antibody in PBST at room temperature for 1 h. Following extensive washes with PBST, antibodyreacted proteins were visualized by chemiluminescence using the ECL Western blot detection reagents or SuperSignal West Femto reagents. The autoradiography of x-ray film and the band intensity were processed using Quantity One software version 4.5 (Bio-Rad). The antibodies used and their dilutions were as follows: anti-Gal4 (Santa Cruz), 1:1000; anti-Notch(Val 1744 ), 1:1000; anti-phospho-stress-activated protein kinase/JNK and anti-stress-activated protein kinase/JNK (Cell Signaling Technology), 1:1000; horseradish peroxidase-conjugated antimouse IgG (Amersham Biosciences) 1:10,000; and horseradish peroxidase-conjugated anti-rabbit IgG (Santa Cruz), 1:1000.

RESULTS
The Generation and Characterization of a Cell-based Reporter Gene Assay Specific for ␥-Secretase-To specifically address the regulation of ␥-secretase activity, we first generated a cell line (T20) stably transfected with the Gal4-luciferase reporter gene and C99-GV, the immediate substrate of ␥-secretase C-terminal tagged with Gal4/VP16 that was derived from a yeast transcription factor and a viral transcription activator. The rationale of this cell-based luciferase reporter assay for ␥-secretase activity is depicted in Fig. 1A. To optimize the responsiveness of this reporter assay to ␥-secretase-mediated proteolysis, a tetracycline-regulated mammalian expression was employed so that the expression of C99-GV would only be induced when the detection of ␥-secretase activity was needed, reducing the background emission of luminescence because of constitutive expression of C99-GV. In the isolated stable line T20, the ␥-secretase-dependent cleavage of C99-GV expressed upon tetracycline induction releases the Gal4/VP16-tagged APP intracellular domain (AICD-GV) from the membrane, and the transcription factor then activates the expression of the firefly luciferase reporter gene under the control of five tandem repeats of Gal4 cis-elements in the Gal4-Luc vector, resulting in a ϳ30-fold increase of luminescence emission in comparison to the background emission exhibited in the absence of tetracycline induction (1-fold of activation) (Fig. 1B). Such ␥-secretasedependent emission of luminescence can markedly be attenuated by a known ␥-secretase inhibitor, compound E (25), in a dose-dependent manner (Fig. 1B), demonstrating the specificity of this cell-based reporter gene assay for ␥-secretase. However, residual luminescence (ϳ30% to control) can still be observed upon treatment with 10 M compound E. We suspected that a small portion of the chimeric C99-GV might be improp-erly localized to the nucleus and eluded from membrane-bound ␥-secretase, resulting in the basal expression of luciferase reporter that generated the residual luminescence. To determine whether the nuclear mislocalization of chimeric C99-GV could account for the source of residual luminescence, the partially purified nuclear fraction of T20 cells treated with or without tetracycline was isolated, followed by Western blot analysis of C99-GV using anti-Gal4 antibody. We found trace amounts of uncleaved chimeric C99-GV in tetracycline-induced T20 cells, but not in uninduced T20 and T-REx293 host cells (data not shown). In addition, complete inhibition of A␤ production in T20 cells treated with DAPT, another potent inhibitor of ␥-secretase (26), was observed (Fig. 1C), suggesting that the residual luminescence is characteristic of background expression of luciferase reporter gene independent of ␥-secretase activity. Thus, the luminescence emitted by T20 cells would be largely dependent on ␥-secretase-mediated cleavage of membrane-bound chimeric C99-GV, providing a highly efficient, sensitive, and quantitative measurement of ␥-secretase activity.
␥-Secretase Activity Can Be Stimulated by IFN-␥, IL-1␤, and TNF-␣-Marked increases in A␤40 and A␤42 were observed in both neuronal and non-neuronal cells following stimulation with combinations of IFN-␥ and TNF-␣ or IFN-␥ and IL-1␤ (19,20), but no details were provided as to the molecular mechanism underlying such cytokine-stimulated production of A␤. We therefore examined whether or not IFN-␥, IL-1␤, and TNF-␣ can individually stimulate the ␥-secretase-mediated cleavage of APP, the last proteolytic step in the generation of A␤, by using the newly established cell-based reporter assay for ␥-secretase. We found that cytokine-treated T20 cells displayed dramatically enhanced ␥-secretase activity ( Fig. 2A). Among the cytokines examined, TNF-␣ exhibited the most potent stimulatory effect, inducing an ϳ100% increase of ␥-secretase activity after a 9-h treatment in comparison with the basal enzyme activity exhibited by the untreated control T20 cells. Such cytokine-elicited regulation of ␥-secretase also exhibited a dosedependent stimulation when T20 cells were stimulated with various concentrations of respective cytokines for 9 h (Fig. 2B).
To determine whether these cytokine-triggered stimulations of ␥-secretase can be mediated through specific ligand-receptor interactions, a recombinant soluble TNF receptor (sTNFR) fragment was found to effectively antagonize TNF-␣-elicited activation of ␥-secretase (Fig. 2C), demonstrating the specificity of a cytokine-elicited effect on ␥-secretase. Our data suggested that these inflammatory cytokines could trigger distinct downstream signaling cascades through which ␥-secretase activity can be controlled. These stimulatory effects were not because of nonspecific stimulation of luciferase as demonstrated by the virtually unaffected luciferase signal upon treatments with respective cytokines in a control cell line constitutively expressing the luciferase reporter gene (Fig. 2D), substantiating that the observed up-regulation of ␥-secretase activity was a cytokine-dependent effect. The cytokine-stimulated ␥-secretase activity was further evidenced by the increased production of secreted A␤ and cognate AICD in conditioned media and cell lysates of cytokine-treated T20 cells (Fig.  3). Consistently, TNF-␣ exhibited the most potent effect on stimulating the ␥-secretase, resulting in an ϳ1.6-fold increase of both A␤ and AICD in the presence of 50 ng/ml TNF-␣. Together, our results suggest that IFN-␥, IL-1␤, and TNF-␣ could augment APP processing through cytokine-dependent stimulation of ␥-secretase, despite the varied degree of potency of respective cytokines on ␥-secretase activity.
␥-Secretase-mediated S3 Cleavage of Notch Can Be Potentiated by IFN-␥, IL-1␤, and TNF-␣-We next examined whether or not the cytokine-elicited up-regulation of ␥-secretase also affects the ␥-secretase-mediated S3 cleavage of Notch. To do so, we treated a stable cell line (N7) constitutively expressing N⌬E that was a Notch mutant protein lacking its extracellular domain but retaining its membrane-spanning region. As depicted in Fig. 4A, the recombinant N⌬E was expressed as a membrane-tethered protein that can be cleaved directly by ␥-secretase independent of its ligand activation (22). By Western blotting analysis using the anti-Notch(Val 1744 ) antibody, we demonstrated that the production of NICD in N7 cells resulting from ␥-secretase-mediated cleavage of N⌬E was significantly inhibited by compound E and DAPT, two potent ␥-secretase inhibitors (25,26) (Fig. 4B). This anti-Notch(Val 1744 ) antibody only detected the cytosolic NICD that was cleaved by ␥-secretase between Gly 1743 and Val 1744 residues, but cannot recognize the intact recombinant N⌬E, the endogenous full-length The membrane-tethered C-terminal fragment of APP (C99) was derived from the APP695 isoform and fused with a C-terminal tag of Gal4/VP16. This chimeric construct (C99-GV) thus served as the immediate substrate for ␥-secretase. Scheme I, the cleavage of C99-GV by ␥-secretase activates the expression of the luciferase reporter gene. Scheme II, the ␥-secretase inhibitor blocks the proteolysis of C99-GV and prevents expression of the luciferase reporter gene. The solid circle denotes the Gal4/VP16 tag. Oval, rectangle, and diamond represent the extracellular domain, transmembrane domain, and intracellular domain of APP695, respectively. The arrow indicates the membrane-localized ␥-secretase, and secreted A␤ is indicated accordingly. B and C, dose response of T20 cells to ␥-secretase-specific inhibitors, compound E and DAPT. Clonally isolated T20 cells were treated with various concentrations of either compound E (B) or DAPT (C) in culture medium for 24 h at 37°C. ␥-Secretase activity in clarified lysates and secreted A␤40 in conditioned media were determined as described under "Experimental Procedures." Background luminescence emitted by T20 cells treated with culture medium in the absence of tetracycline was referred as 1-fold of activation. Results from a representative experiment are expressed as the mean (ϮS.D.) of triplicate measurements. Notch1, or Notch1 cleaved at other positions, thus making the N7 cell line an ideal alternative cell-based system to examine endogenous ␥-secretase activity. Using the N7 cell line, we found that, in the presence of IFN-␥, IL-1␤, or TNF-␣, ␥-secretase-catalyzed cleavage of N⌬E was stimulated as indicated by the increased production of NICD (Fig. 4C). When added in higher dosage (50 ng/ml), IFN-␥, IL-1␤, and TNF-␣ stimulated the production of NICD by 12, 15, and 64%, respectively. This cytokine-stimulated NICD production is consistent with the cytokine-dependent potentiation in the ␥-secretase-mediated cleavage of C99-GV (Figs. 2 and 3), although in a smaller extent. It is possible that the constitutive overexpression of N⌬E has saturated N7 cells with NICD prior to cytokine stimulation, making N7 cells a less responsive system than T20 cells. Our data suggest that these proinflammatory cytokines can also promote the ␥-secretase-mediated proteolysis of Notch.
Cytokine-elicited Stimulation of ␥-Secretase Is Blocked by JNK Inhibitor SP600125-We next determined through which intracellular signaling cascade the cytokine-triggered regulation of ␥-secretase was transduced. It has been well documented that cellular responses induced by TNF-␣ and IL-1␤ can result in the activation of JNK (27,28). We thus hypothesized that cytokine-elicited signaling might converge on JNK through which ␥-secretase activity can then be regulated. The possible engagement of the JNK-dependent MAP kinase pathway in cytokine-elicited regulation of ␥-secretase was first explored by treating T20 cells with a potent cell-permeable JNK inhibitor, SP600125 (29), in the presence of respective cytokines. We found that the cytokine-dependent stimulation of ␥-secretase activity was completely suppressed by SP600125 down to the basal level, concomitant with the decreased production of A␤40 in conditioned media of SP600125-treated T20 cells (Fig. 5, A and B). The ␥-secretase in control T20 cells without cytokine stimulation exhibited its basal activity at 5-fold of activation that was only slightly suppressed by SP600125. Neither a potent p38 MAP kinase inhibitor FIG. 2. The stimulatory effects of TNF-␣, IL-1␤, and IFN-␥ on ␥-secretase. A, the regulatory effects of cytokines on ␥-secretase. T20 cells in a 12-well microplate (5 ϫ 10 5 cells/well) were maintained in serum-free DMEM overnight at 37°C. Cytokine was added into serum-free DMEM containing 1 g/ml tetracycline to a final concentration of 50 ng/ml. Following incubation at 37°C for 9 h, ␥-secretase activity in treated cells was determined. B, dose-dependent stimulation of ␥-secretase activity by cytokines. T20 cells (5 ϫ 10 5 cells/well) in 12-well microplates were treated with 5 or 50 ng/ml of the respective cytokines in serum-free DMEM containing 1 g/ml tetracycline at 37°C for 9 h, followed by the determination of ␥-secretase activity in the clarified lysates of treated cells. Solid bar, IFN-␥ treatment; shaded bar, IL-1␤ treatment; striped bar, TNF-␣ treatment. C, blocking effect of soluble sTNFRI fragment on TNF-␣-stimulated ␥-secretase activity. T20 cells (5 ϫ 10 5 cells/well) in 12-well microplates were treated with 50 ng/ml TNF-␣ in the presence or absence of 500 ng/ml recombinant sTNFR1 at 37°C for 9 h, followed by the determination of ␥-secretase activity in the clarified lysates of treated cells. Solid bar, control (without TNF-␣); striped bar, TNF-␣ treatment. Results from a representative experiment are expressed as the mean (ϮS.D.) of triplicate measurements and analyzed by Student's t test. **, p Ͻ 0.001. Background luminescence emitted by T20 cells treated with serum-free DMEM alone was referred to as 1-fold of activation. D, luciferase activity was not affected by cytokines. GL-T12 cells constitutively expressing firefly luciferase were plated in 12-well microplates (5 ϫ 10 5 cells/well) and treated with 50 ng/ml of the respective cytokines in serum-free DMEM containing 1 g/ml tetracycline at 37°C for 9 h, followed by the determination of ␥-secretase activity in the clarified lysates of treated cells. Luminescence emitted by GL-T12 cells that were not treated by cytokines (Control) was defined as 100% relative luminescence.
Constitutively Active MEKK1 Potentiates ␥-Secretase, whereas Dominant-negative Mutants of MEKK1 Attenuate Its Activity-To ascertain the pivotal role of JNK on the cytokineelicited stimulation of ␥-secretase, T20 cells were transfected with either a constitutively active construct of MAPK kinase kinase 1 (MEKK1), one of the most potent activators of JNKdependent cascade (30), or a dominant-negative mutant of MEKK1 (T572A). Subsequently, ␥-secretase activity in transfected cells was then treated with the JNK inhibitor SP600125. As shown in Fig. 6A, the transfection of constitutively active MEKK1 in T20 cells dramatically enhanced ␥-secretase activity, an ϳ20-fold increase, whereas cells transfected with the dominant-negative mutant T572A exhibited much less stimulation of ␥-secretase. We suspected that the incomplete attenuation of JNK activation might be because of the residual kinase activity of this dominant-negative mutant and the remaining endogenous MEKK1. Consistently, transfection of the constitutively active MEKK1 in N7 cells can stably sustain the A, the A␤40 contents in the conditioned media of cytokine-treated T20 cells. T20 cells (5 ϫ 10 5 cells/well) in 12-well tissue culture plates were treated with 5 or 50 ng/ml IFN-␥, IL-1␤, or TNF-␣ in serum-free DMEM containing 1 g/ml tetracycline for 9 h at 37°C. The A␤40 level in the conditioned medium obtained from T20 cells treated with serum-free DMEM containing 1 g/ml tetracycline was determined to estimate the basal level of secreted A␤ without cytokine stimulation. A␤40 levels in clarified conditioned media were determined by the human A␤40 ELISA colorimetric kit as described under "Experimental Procedures." Solid bar, IFN-␥ treatment; shaded bar, IL-1␤ treatment; striped bar, TNF-␣ treatment. Data are shown as the mean (ϮS.D.) of triplicate measurements from a representative experiment. B, Western blot analysis of intracellular AICD. T20 cells were treated with 50 ng/ml IFN-␥, IL-1␤, or TNF-␣ in serum-free DMEM containing 1 g/ml tetracycline for 9 h at 37°C as described above. Cytokine-treated cells were harvested and dissolved in 1ϫ passive lysis buffer. Clarified lysates containing equal amounts of proteins (15 g) were resolved by SDS-PAGE. Gal4-tagged AICD was analyzed by Western blotting using an anti-Gal4 polyclonal antibody, followed by horseradish peroxidase-conjugated anti-rabbit IgG antibody. Antibodyreacted proteins on the immunoblot were visualized by chemiluminescence using ECL Western blot detection reagents. T20 cells treated with serum-free DMEM containing 1 g/ml tetracycline alone were included as the control. The arrow indicates that the Gal4-tagged AICD resulted from ␥-secretase cleavage of C99-GV in T20 cells. The band intensities of AICD from different samples were quantified and normalized by the band of AICD from control. The numbers below each lane denote the normalized AICD band intensities from different samples. TNF-␣-activated ␥-secretase activity for up to 48 h, whereas both dominant-negative mutants of MEKK1 failed to retain TNF-␣-elicited activation of ␥-secretase in the same duration of time as evidenced by the dramatically reduced levels of NICD (Fig. 6B, upper panel). The effectiveness of both dominantnegative mutants on blocking JNK activation was demonstrated by the significantly reduced levels of phosphorylated JNK after a 48-h treatment of TNF-␣, whereas the levels of total JNK remained unaffected (Fig. 6B, middle and bottom  panels). Using a Chinese hamster ovary line stably coexpressing all four constituents of ␥-secretase (␥-30) (9), we further verified that A␤ production in cells transfected with the dominant-negative mutant T572A was significantly attenuated by 25% in comparison to the mock transfected control (Fig. 6C). The present data thus delineated an appealing ␥-secretaseregulatory signaling cascade that can be initiated by extracellular proinflammatory cytokines, including IFN-␥, IL-1␤, and TNF-␣, and transduced intracellularly through the MEKK1-JNK-dependent MAPK pathway. DISCUSSION The generation of A␤ through the proteolytic processing of APP, by sequential actions of ␤and ␥-secretases, is considered to be the most critical step in the pathogenesis of AD. The involvement of secretase-mediated cleavage of APP in the disease process is further substantiated by the identification and characterization of these secretases, including ␣-, ␤-, and ␥-secretases. In particular, the catalytic component of ␥-secretase, presenilin, is mutated in autosomal-dominant familial forms of AD. Despite recent advances toward understanding the molecular mechanisms underlying the proteolysis of APP, how the homeostasis of this proteolytic processing is maintained in vivo remains mostly elusive. Here, we provide evidence that the ␥-secretase-dependent cleavage of APP is under stimulatory control by such proinflammatory cytokines as IFN-␥, IL-1␤, and TNF-␣ through a signaling cascade mediated by MEKK1 and JNK. This finding is further corroborated by the increase of ␥-secretase-mediated cleavage of N⌬E upon the stimulation with these proinflammatory cytokines. Our results also provide evidence that cytokine-elicited signaling cascades converge on specific intracellular mediators, such as JNK, through which the ␥-secretase processing of APP can thus be modulated.
Despite accumulated evidence suggesting the close correlation of inflammatory responses with A␤-centered pathogenesis of AD (31), there has been little evidence illustrating how these inflammatory responses can affect the proteolytic processing of APP and the consequent A␤ deposition in the neuritic plaques, the most critical pathological hallmark of the AD brain. Our results not only confirm the stimulated production of A␤ by such proinflammatory cytokines as IFN-␥, IL-1␤, and TNF-␣ (19,20), but also provide a possible molecular mechanism for how the up-regulated expression of IFN-␥, IL-1␤, and TNF-␣ can promote A␤ production in AD patients. Interestingly, whereas Blasko et al. (19,20) previously showed that only the combinations of IFN-␥ plus IL-1␤ and IFN-␥ plus TNF-␣, but not individual cytokines, can stimulate A␤ production in neuronal and non-neuronal cells, we clearly demonstrated that IFN-␥, IL-1␤, and TNF-␣ can each alone stimulate A␤ production, although in varied degrees (Fig. 3A). We suspect that this discrepancy might be because of the assay systems used. Whereas Blasko et al. (19,20) measured the production of A␤ from sequential ␤/␥ cleavage of endogenous APP and addressed the proteolytic processing of APP as a whole in the presence of cytokines, we, in contrast, adopted a C99-transfected stable line T20 whose production of A␤ is largely originated from ␥-secretase-dependent cleavage of ectopically expressed C99 and selectively examined the single ␥-secretase cleavage of C99. It is thus possible that A␤ production in our system becomes highly responsive to cytokine stimulations. Although our reporter assay for ␥-secretase does not distinguish the ␥ cleavage that produces the C termini of A␤ peptides from the ⑀ cleavage that generates the N termini of AICD, previous studies have shown that the ␥/⑀ cleavages of APP and the S3 cleavage of Notch are presenilin/␥-secretase-dependent (32,33), suggesting that these distinct cleavages could still be mediated by the same catalytic machinery and are under the control by the same repertory of intracellular pathways. In addition, the duration of the cytokine treatment might also be a factor. Our present results show that the optimal stimulation of ␥-secretase activity can be reached at 9 h post-exposure to cytokines, in contrast to the 24-h treatments of cytokines employed by Blasko et al. (19,20). Furthermore, using two different cell-based ␥-secretase assay systems, the APP-based T20 (Figs. 2 and 3) and Notch-based N7 lines (Fig. 4), we found that endogenous ␥-secretase activity can be enhanced consistently upon cytokine stimulations toward its cleavage of different substrates, C99 and N⌬E. The cytokine-triggered increase of A␤ levels in conditioned media (Fig. 3), concomitant with augmented production of AICD, unequivocally demonstrates that signaling cascades elicited by these proinflammatory cytokines all converge onto ␥-secretase-catalyzed proteolysis, including ␥/⑀ cleavage of APP and S3 cleavage of Notch. The present data thus reveal an attractive mechanism explaining how inflammatory mediators can accelerate AD pathogenesis. Given that the ectodomain shedding in ␥-secretase substrates has been regarded as a prerequisite for ␥-secretase-mediated intramembrane proteolysis (34), our data lead us to postulate that the secretase-mediated proteolysis of APP can be subject to multiple levels of regulation by intracellular pathways, providing a coordinated proteolysis of APP for the stringent production of A␤ in physiological conditions. This notion is thus in accordance with previous reports demonstrating that platelet-derived growth factor can enhance the ␤␥-secretase-mediated proteolysis of APP (21) and that the combinatorial effects of IFN-␥ plus IL-1␤ and IFN-␥ plus TNF-␣ can stimulate A␤ production (19,20), suggesting that equally tight control of the ectodomain shedding and the ␥-secretase-mediated intramembrane cleavage could be crucial for the homeostasis of endogenous A␤ production. To better understand the control of APP proteolysis, quantitative measurements of ␣and ␤-secretase upon cytokine treatments, respectively, might be required to fully reveal the relative contributions of ectodomain shedding and the enhanced ␥-secretase cleavage in the proteolysis of APP and the regulation of amyloidogenesis.
Our discovery of the cytokine-dependent stimulation of ␥-secretase activity becomes more evident as a JNK-specific antagonist SP600125 and a potent JNK activator MEKK1 can, respectively, inhibit and sustain ␥-secretase activity (Figs. 5 and 6). The involvement of the JNK-dependent MAPK pathway on the regulation of ␥-secretase is further supported by decreased production of A␤ and NICD in the presence of SP600125 or a dominant-negative mutant of MEKK1 (T572A) and increased production of NICD in the presence of constitutively active MEKK1, providing the basis for identifying novel reagents aimed to lower A␤ by using JNK as a pharmacological target. This concept has been tested by the application of JNK inhibitors in neurodegenerative diseases (35). Interestingly, overexpression of PS1 has been shown to negatively regulate JNK-dependent signaling pathways, whereas mutant PS1 can potentiate JNK-induced apoptosis (36,37). In transgenic mice overexpressing a mutant PS1, the activation of JNK in neurons neighboring to amyloid plaques and the ones containing intracellular accumulation of A␤ is evident (38,39). In addition, the activation of MAP kinases, including JNK, has been documented in susceptible neurons of AD patients and correlates with AD pathogenesis (40 -43). It is thus plausible that signaling cascades elicited by proinflammatory cytokines could result in the activation of the JNK-dependent MAPK pathway through which ␥-secretase-mediated production of A␤ can then be stimulated. Giving that cytokine-elicited activation of JNK could stimulate ␥-secretase and increase the production of A␤, which in turn could induce the expression of mutant PS1 in the AD brain, furthering the production of A␤ and possibly constituting a positive feedback of A␤-induced pathogenesis. We are currently investigating whether or not the cytokine-elicited regulation of ␥-secretase is because of alterations in the JNKdependent phosphorylation of this protease. Alternatively, cytokine-elicited activation of JNK may lead to the phosphorylation of APP at Thr 668 , the phosphorylation site previously shown to be utilized by JNK (44), and expedite the secretasemediated processing of APP via enhanced interaction between the membrane-tethered C99 and ␥-secretase. Thus, the cytokinedependent regulation of ␥-secretase by signaling through the JNK pathway might play a pivotal role in the inflammationassociated AD pathogenesis.
Increased levels of TNF-␣ and IL-1␤, the two potent stimulators of ␥-secretase, have been correlated with AD and other neurodegenerative defects (45,46). In addition to the JNK-de-pendent activation of transcription factor AP-1, signaling elicited by TNF-␣ and IL-1␤ can also activate NFB-dependent transcription of genes involved in chronic and acute inflammatory responses (47,48). The possibility thus exists that the expression of AP-1/NFB-responsive genes could play a role in regulating ␥-secretase, implicating the presence of endogenous genetic modifiers of this protease. Given that these two cytokines tend to induce each other (49), it is likely that AP-1-and NFB-dependent transcription might regulate the expression of a joint repertory of genes whose products could then regulate ␥-secretase activity, underlying the pathogenesis of various neurodegenerative diseases. As to IFN-␥ on regulating ␥-secretase, although IFN-␥-elicited intracellular signaling is primarily mediated by the JAK/STAT kinase cascade (50), it has also been found to induce the phosphorylation of p42/p44 MAPK (ERK1/2) through Jak1 and Raf (51). Because both JNK and ERK1 have been shown to phosphorylate APP at Thr 668 (44), it is thus likely that TNF-␣, IL-1␤, and IFN-␥ may also promote the phosphorylation of APP, resulting in more efficient processing by ␥-secretase and increased production of A␤. Moreover, recent evidence has demonstrated that fibrillar A␤ can stimulate microglia through a cell surface receptor complex, leading to increased production of TNF-␣ and IL-1␤ that are responsible for A␤-induced neurotoxicity and neuronal death (52,53). Thus, our present study could then provide a functional linkage between inflammation-associated activation of ␥-secretase-dependent A␤ production and A␤-elicited neurotoxicity via microglial activation characteristic of the AD brain.
Taken together, our data suggest the possibility that the activation of MEKK1 and JNK through signaling cascades triggered by TNF-␣, IL-1␤, and IFN-␥ could be critical to the generation of A␤ by ␥-secretase in AD brain and that new therapeutic targets for the intervention of AD pathogenesis can be identified along these signaling cascades. Thus, intrinsic inflammatory mechanisms may provide such a delicate regulatory system that is capable of influencing the homeostasis of APP processing and A␤ production. Furthermore, the experimental systems described in this article could be of use in finding novel molecules that attenuate the cytokine-MEKK1-JNK-induced activation of ␥-secretase and that, in turn, could be beneficial for the development of anti-AD drugs.