Nuclear Factor-κB Regulates βAPP and β- and γ-Secretases Differently at Physiological and Supraphysiological Aβ Concentrations*

Background: NF-κB regulates BACE1 but there is little data suggesting βAPP and γ-secretase involvement. Results: NF-κB differentially regulates Aβ production at physiological and supraphysiological Aβ concentrations by modulating transactivation of βAPP and γ-secretase promoters, thereby controlling γ-secretase activity. Conclusion: Under physiological conditions, NF-κB regulates Aβ homeostasis while it contributes in increasing Aβ production in the pathological context. Significance: NF-κB may be seen as a potential therapeutic target. Anatomical lesions in Alzheimer disease-affected brains mainly consist of senile plaques, inflammation stigmata, and oxidative stress. The nuclear factor-κB (NF-κB) is a stress-activated transcription factor that is activated around senile plaques. We have assessed whether NF-κB could be differentially regulated at physiological or supraphysiological levels of amyloid β (Aβ) peptides. Under these experimental conditions, we delineated the putative NF-κB-dependent modulation of all cellular participants in Aβ production, namely its precursor βAPP (β-amyloid precursor protein) and the β- and γ-secretases, the two enzymatic machines involved in Aβ genesis. Under physiological conditions, NF-κB lowers the transcriptional activity of the promoters of βAPP, β-secretase (β-site APP-cleaving enzyme 1, BACE1), and of the four protein components (Aph-1, Pen-2, nicastrin, presenilin-1, or presenilin-2) of the γ-secretase in HEK293 cells. This was accompanied by a reduction of both protein levels and enzymatic activities, thereby ultimately yielding lower amounts of Aβ and AICD (APP intracellular domain). In stably transfected Swedish βAPP-expressing HEK293 cells triggering supraphysiological concentrations of Aβ peptides, NF-κB activates the transcription of βAPP, BACE1, and some of the γ-secretase members and increases protein expression and enzymatic activities, resulting in enhanced Aβ production. Our pharmacological approach using distinct NF-κB kinase modulators indicates that both NF-κB canonical and alternative pathways are involved in the control of Aβ production. Overall, our data demonstrate that under physiological conditions, NF-κB triggers a repressive effect on Aβ production that contributes to maintaining its homeostasis, while NF-κB participates in a degenerative cycle where Aβ would feed its own production under pathological conditions.

Anatomical lesions in Alzheimer disease-affected brains mainly consist of senile plaques, inflammation stigmata, and oxidative stress. The nuclear factor-B (NF-B) is a stress-activated transcription factor that is activated around senile plaques. We have assessed whether NF-B could be differentially regulated at physiological or supraphysiological levels of amyloid ␤ (A␤) peptides. Under these experimental conditions, we delineated the putative NF-B-dependent modulation of all cellular participants in A␤ production, namely its precursor ␤APP (␤-amyloid precursor protein) and the ␤and ␥-secretases, the two enzymatic machines involved in A␤ genesis. Under physiological conditions, NF-B lowers the transcriptional activity of the promoters of ␤APP, ␤-secretase (␤-site APPcleaving enzyme 1, BACE1), and of the four protein components (Aph-1, Pen-2, nicastrin, presenilin-1, or presenilin-2) of the ␥-secretase in HEK293 cells. This was accompanied by a reduction of both protein levels and enzymatic activities, thereby ultimately yielding lower amounts of A␤ and AICD (APP intracellular domain). In stably transfected Swedish ␤APP-expressing HEK293 cells triggering supraphysiological concentrations of A␤ peptides, NF-B activates the transcription of ␤APP, BACE1, and some of the ␥-secretase members and increases protein expression and enzymatic activities, resulting in enhanced A␤ production. Our pharmacological approach using distinct NF-B kinase modulators indicates that both NF-B canonical and alternative pathways are involved in the control of A␤ production. Overall, our data demonstrate that under physiological conditions, NF-B triggers a repressive effect on A␤ production that contributes to maintaining its homeostasis, while NF-B participates in a degenerative cycle where A␤ would feed its own production under pathological conditions.
Alzheimer disease (AD) 3 is the first cause of dementia, involving memory and cognitive deficit ultimately leading to the loss of patient autonomy (1). The molecular dysfunctions are not yet elucidated, and actual treatments are only symptomatic. AD-affected brains present extracellular deposits; the senile plaques that are mainly composed of a set of hydrophobic peptides called amyloid ␤-peptides (A␤) (2). These peptides are at the center of the amyloid cascade hypothesis (3) that proposes A␤ as the main etiological trigger of the neurodegeneration taking place at late stages of the pathology (4). Therefore, a strategy aimed at circumscribing A␤ overload appears as one of the main therapeutic tracks.
A␤ is a normal product of ␤-amyloid precursor protein (␤APP) processing (5) that results from the sequential cleavage of ␤APP by two distinct enzyme activities: the ␤-secretase ␤-site APP-cleaving enzyme 1 (BACE1) and the ␥-secretase that mainly consists of a high molecular weight protein complex including anterior pharynx defective-1 (Aph1), presenilin enhancer-2 (Pen2), presenilin 1 or 2 (PS1, PS2), and nicastrin (NCT) (6). * This work was supported by the Fondation pour la Recherche Médicale Several lines of evidence suggest that inflammation and oxidative stress could contribute to AD pathology (7,8). This may influence A␤ production to some extent, since BACE1 expression appears to be modulated by stress (9,10). Oxidative stress and inflammation are characterized by the release of cytokines and by reactive oxygen species, known to activate the nuclear factor-B (NF-B) (11,12).
NF-B is a dimeric transcription factor, the family of which is composed by the proteins p50, p52, p65(RelA), RelB, and c-Rel (13,14). Two pathways have been delineated for NF-B activation, the canonical and the alternative pathways. They are generally associated with distinct triggers, involve different signaling proteins, and are linked to structurally and functionally varying DNA-binding dimers (14) (Fig. 1A). Interestingly, an increase of NF-B expression is observed in the hippocampus and entorhinal cortex of AD patients, two cerebral areas altered in this pathology (15).
Several in vitro studies suggested that NF-B could be activated by A␤ peptides in primary cultured neurons (16,17). This observation is indirectly consistent with the fact that NF-B has been observed in cells surrounding or within the amyloid plaques (16 -19). These studies report on an activation of the canonical pathway (16 -19) involved in a feedback control by which A␤ activates NF-B, which, in turn, regulates the production of A␤ peptides (17,19).
Interestingly, the inhibition of NF-B activation reduces A␤ secretion in vitro (20 -22), and it was suggested that this could occur by interfering with ␤APP processing (22)(23)(24)(25)(26)(27). In this context, it is noteworthy that we and others previously showed that NF-B mediated the A␤-associated increase of BACE1 transcriptional activity (28,29) but to date, there is relatively little data concerning the putative control of the ␥-secretase build-up and activity or on the expression of ␤APP by NF-B; and, if so, if this control may be dependent on A␤.
Here we show an A␤ concentration-dependent control of ␤APP and of the members of the ␥-secretase complex by NF-B. We establish that under physiological conditions, NF-B lowers A␤ production by repressing the protein ␤APP and the ␤and ␥-secretase activities at a transcriptional level. Conversely, at supraphysiological concentrations of A␤ aimed at mimicking the pathological situation, NF-B activates A␤ production by increasing ␤APP and processing enzyme activities. This set of data delineates a differential control of ␤APP, and ␤and ␥-secretases by NF-B that depends upon physiological or supraphysiological conditions of A␤ production. Thus, our results indicate that NF-B could normally control A␤ homeostasis in physiological conditions or contribute to a degenerative cycle by which A␤ feeds its own production in the pathological context.

MATERIALS AND METHODS
Cell Culture and Transfections-Human embryonic kidney 293 (HEK293) cells stably overexpressing empty pcDNA3.1 vector, wild-type (wt-␤APP), or ␤APP harboring the Swedish double mutation (K670N/M671L; Sw-␤APP) were obtained and cultured as described previously (30). Human SH-SY5Y neuroblastoma cells (CRL-2266, ATCC) were cultured following the manufacturer's instructions. SH-SY5Y cells stably expressing pcDNA3.1 or Sw-␤APP were generated following standard protocols and maintained in the presence of 400 g of geneticin (Invitrogen). Mouse embryonic fibroblasts (MEF) cells devoid of the ␤APP gene (␤APP Ϫ/Ϫ ) were kindly provided by Dr. U. Muller (31). Transient and stable transfections of cDNAs were obtained with jetPRIME reagent (Polyplus) with an optimized protocol (half quantities of cDNAs, buffer, and reagent recommended by the manufacturer were used). Cells were collected 30 h or 48 h after transient transfection, or selected with geneticin antibiotic (Invitrogen) to obtain stable transfection.
Activation and Inhibition of NF-B-NF-B was activated by transfection of HA-tagged-IKK1SϾE cDNA (IKK1SE, mutations S176E and S180E on the IKK1 kinase) and Flag-tagged-IKK2SϾE cDNA (IKK2SE, mutations S177E and S181E on the IKK2 kinase), and inhibited by transfection of IB superrepressor (Myc-tagged IBSR cDNA, mutations S32A and S36A on IB) (32,33). Their expression was verified by Western blot using mouse monoclonal anti-HA (Eurogentec), anti-Flag (Sigma), and anti-Myc 9E10 (given by Dr. Luc Mercken) antibodies, respectively (Fig. 1B). Their functionality was verified by cotransfection with B-luciferase, a consensus sequence of fixation of the various dimers of NF-B in-frame with luciferase ( Fig. 1, A-C). Cells were then lysed and assayed as described below for transactivation of promoters.
Exogenous A␤ Treatment-Twenty-four hours after transfection of the B-luciferase construct, the medium was replaced with Opti-MEM medium (Sigma) containing 2% fetal bovine serum and 10 M phosphoramidon to prevent A␤ degradation. Cells were treated for 48 h with 0.3, 1, or 3 M synthetic A␤42 (Bachem).
Inhibition of ␥-Secretase Activity-Thirty hours after transfection of the B-luciferase construct, cells were treated for 18 h with 100 M DFK167 (34) or a corresponding amount of Me 2 SO.
Transactivation of Promoters-cDNA encoding human promoters of ␤APP (given by Dr. D. Lahiri), Aph1, Pen2, NCT (given by Dr. X. Xu), PS1 (given by Dr. M. Vitek), PS2 (given by Drs. P. Renbaum and E. Levy-Lahad), and rat promoter of BACE1 (given by Dr. S. Rossner) in-frame with luciferase were co-transfected with ␤-galactosidase transfection vector, to normalize transfection efficiency. CMV-␤-galactosidase construction was used to assess putative artifactual effect of IKK1SE, IKK2SE, and IB constructions on the CMV promoter. 30 h after transfection, cells were harvested with phosphate-buffered saline/EDTA (5 mM), pelleted by centrifugation 5 min at 1000 ϫ g, lysed with 50 l of lysis buffer (luciferase kit Promega), centrifuged for 5 min at 2000 ϫ g. Luciferase and ␤galactosidase activities were then analyzed as previously described (35), and protein concentration determined to normalize the luciferase activity.
BACE1 Fluorimetric Assay-Cells were collected 30 h after transfection and lysed with 10 mM Tris-HCl, pH 7.5, then homogenates were monitored for their BACE1 activity as described previously (38,39). Briefly, samples (30 g of proteins in 25 mM acetate buffer, pH 4.5) were incubated in a final volume of 100 l of the above acetate buffer containing BACE1 substrate (10 M (7-methoxycoumarin-4-yl)acetyl-SEVNL-DAEFR K(2,4-dinitrophenyl)-RRNH2; R&D Systems) in the absence or presence of ␤-secretase inhibitor I (50 M, Promo-Cell). BACE1 activity corresponds to the ␤-secretase inhibitorsensitive fluorescence recorded at 320 and 420 nm as excitation and emission wavelengths, respectively. The slopes of the initial linear phase were calculated and expressed as fluorimetric units/mg/h.
In Vitro ␥-Secretase Assay-48 h after transfection, cells were used for an in vitro ␥-secretase assay developed in the laboratory (40). Briefly, cells were lysed with Tris 10 mM, pH 7.5, and membranes were isolated by centrifugation (22,000 ϫ g for 1 h). An equal amount of membranes was incubated overnight with a Flag-tagged quenched fluorimetric substrate (JMV2660) mimicking the ␤APP sequence targeted by ␥-secretase. The products A␤ and AICD-Flag were detected by Western blotting on a 16.5% Tris/tricine gel and revealed simultaneously with anti-A␤ 2H3 and anti-Flag antibodies, respectively.
Analysis of A␤40 Production-30 h after transfection, cells were allowed to secrete A␤ in Optim-MEM medium (Sigma) for 16 h, in the presence of 10 M phosphoramidon to prevent A␤ degradation. The culture media were collected, supplemented with RIPA buffer (10 mM Tris-HCl, pH 8, EDTA 5 mM, NaCl 150 mM), and incubated overnight with a 100-fold dilution of FCA3340 (41). A␤40 was precipitated with protein A-Sepharose (Invitrogen), Western blotted on a 16.5% Tris/ tricine gel, and revealed with anti-␤APP 2H3 (provided by Dr. D. Schenk, Elan Pharmaceuticals). Cells were collected, and protein concentration was determined in order to normalize A␤ secretion.
Statistical Analysis-Statistical analysis was performed with Prism software (Graphpad, San Diego, CA) using either the unpaired Student's t test for pairwise comparison or the Tukey multiple comparison test for one-way analysis of variance.

NF-B Reduces A␤ Secretion, ␤APP and Secretase Expressions and Activity in Physiological
Conditions-Two pathways have been described for the activation of the transcription fac-tors NF-B (Fig. 1A). In canonical and alternative pathways (Fig. 1A), inactive NF-B dimers are sequestrated in the cytoplasm either by IB (inhibitor of NF-B) or p100, respectively. The phosphorylation of IB or p100 by distinct IB kinase (IKK) complexes releases active NF-B dimers that translocate into the nucleus where they regulate the transcription of their target genes (13). The contribution of these highly regulated cascades can be selectively evaluated by means of mutant inhibitory or activating constructions, respectively, targeting IB or upstream IKK1 or IKK2 kinases (Fig. 1A). Thus, IKK1SE and IKK2SE are constitutively activated IKKs, the transfection of which (Fig. 1B) leads to drastic increases of NF-B activity (respectively ϩ715.0 Ϯ 72.6% and ϩ7039.0 Ϯ 899.3% com-

FIGURE 1. NF-B inhibition of A␤ secretion in transiently transfected HEK293 cells.
A, NF-B is a dimeric transcription factor sequestered in the cytoplasm by the binding of its inhibitor IB (inhibitor of B) or p100. The pathway is activated by phosphorylation of the IB kinase (IKK) complex. In the canonical pathway, NF-B dimers, mainly p50-RelA, are bound to IB. When the pathway is activated, IKK2 phosphorylates IB, which dissociates from NF-B. Free dimers of NF-B translocate to the nucleus and regulate the expression of target genes. In the alternative pathway, IKK1 phosphorylates the inhibitory p100 subunit, p100 is processed into p52, and the dimer p52-RelB translocates to the nucleus. IKK1SE and IKK2SE are constitutively activated IKKs, and their transfection enhances IB and p100 phosphorylation and subsequent NF-B activation. IBSR is a dominant negative IB that cannot be phosphorylated, and its transfection blocks the canonical NF-B pathway by sequestering the specific dimers in the cytoplasm. B and C, 30 h after transient transfection of IKK1SE, IKK2SE, and IBSR, HEK293 cells were collected. The efficiency of the transfection is established by Western blot (B) as described under "Material and Methods." The functional effect of the transfection is measured by co-transfection with B-luciferase, a consensus sequence of fixation of the various dimers of NF-B in-frame with luciferase (C). Bars correspond to luciferase activity expressed as percent of that observed in mock-transfected cells and are the means Ϯ S.E. of 16 -20 independent determinations. **, p Ͻ 0.001. D, HEK293 cells were transiently transfected with NF-B activating (IKK1SE and IKK2SE) or inhibitory (IBSR) constructions, and the secreted A␤ was detected after immunoprecipitation with FCA3340 antibody and Western blotting as described under "Material and Methods." Bars correspond to the densitometric analysis of secreted A␤ immunoreactivity expressed as percent of that observed in mock-transfected cells and are the means Ϯ S.E. of 5-6 independent determinations. *, p Ͻ 0.01; ns, not statistically significant. pared with control, n ϭ 20, Fig. 1C) via constitutive phosphorylation/inactivation of IB. Conversely, IB superrepressor (IBSR) does not undergo phosphorylation and therefore, its overexpression lowers NF-B activity (Ϫ46.8 Ϯ 3.2% compared with control, n ϭ 16, Fig. 1C). By means of these constructions, we establish that NF-B activation reduces endogenous A␤ secretion (Ϫ37.5 Ϯ 6.3% and Ϫ35.9 Ϯ 6.5% for IKK1SE and IKK2SE-expressing cells, respectively, n ϭ 5-6, Fig. 1D) while IBSR did not alter A␤ secretion (Fig. 1D). Our results indicate that both canonical and alternative NF-B-mediated pathways exert a negative modulation of physiological A␤ secretion in HEK293 cells.
The above set of data obtained in transiently and stably transfected cells overall show that ␤APP, BACE1, Pen2, PS1, and NCT expression is negatively regulated by NF-B indicating that both substrates and components of enzymes involved in A␤ production are turned down by NF-B in physiological conditions. However, our data also show that there exists a distinct NF-B-mediated regulation of some members of the ␥-secretase complex. It was therefore of prime interest to establish the functional consequence of NF-B modulation on ␤-secretase and ␥-secretase activities. We have taken advantage of BACE1directed fluorimetric assay (38) and of a recently developed in vitro ␥-secretase assay in reconstituted membranes (40) that allows monitoring of A␤ and AICD produced from an exogenous recombinant substrate. This permits us to directly assess whether the reduction of NF-B on protein expression converts into an enzymatic deficiency without putative "artifactual" modulation of A␤ production via the alteration of the secretory process, ␤APP trafficking, or secretase mislocalization. Fig. 4 shows that fluorimetric recording of ␤-secretase activity is increased by stable transfection of IBSR (ϩ100.6 Ϯ 40.7%, n ϭ 10, Fig. 4A) while stable transfection of the alternative pathway kinase IKK1SE did not alter it. Furthermore, AICD and A␤ productions are reduced by IKK1SE (Ϫ27.8 Ϯ 12.8% and Ϫ26.1 Ϯ 4.2% respectively, n ϭ 6 -8, Fig. 4, B-D) and increased by IBSR (ϩ77.2 Ϯ 36.2% and ϩ26.4 Ϯ 7.4%, respectively, n ϭ 6 -9, Fig. 4, B-D). Overall, these data show that in physiological conditions, NF-Blinked negative modulation of ␤and ␥-secretase compo- nents indeed trigger functional consequences on their enzymatic activities.
The above data show that in human cells producing endogenous levels of A␤ (Fig. 1D), NF-B reduces ␤APP expression and the proteolytic machinery involved in A␤ production. To examine whether such a phenotype is dependent on A␤ levels, we examined the NF-B activity in varying conditions or cell systems where A␤ could be modulated. First, we established that ␤APP itself or a ␤APP-derived fragment indeed controls NF-B in mouse embryonic fibroblasts since NF-B activity is drastically lowered by ␤APP depletion, (Ϫ77.4 Ϯ 3.6%, n ϭ 3, Fig. 5A). By contrast, in HEK293 cells overexpressing either wild-type ␤APP or Swedish-mutated ␤APP (42), where A␤ production is exacerbated (Fig. 5B, upper panels), NF-B activity appears drastically enhanced (ϩ542.6 Ϯ 81.6% and ϩ3521.0 Ϯ 388.3%, respectively, n ϭ 16, Fig. 5B). Exogenous treatment of HEK293 cells with various concentrations of A␤42 dosedependently increased NF-B activity (ϩ35.4 Ϯ 8.2% and ϩ53.7 Ϯ 5.7% at 1 M and 3 M, respectively, n ϭ 4, Fig. 5C). Finally, the inhibition of A␤ production by treatment of HEK293 cells with the ␥-secretase inhibitor DFK167 did not modify NF-B activity, while DFK167 prevented the A␤-induced increase of NF-B activity in Swedish-mutated expressing HEK293 cells (ϩ85.7 Ϯ 6.1%, n ϭ 8, Fig. 5D). This suggests that only supraphysiological levels of A␤ could activate NF-B in HEK293 cells.

NF-B Activates A␤ Secretion and Secretases Activities in
Sw-␤APP HEK293 Cells-The above data led us to envision that, at supraphysiological levels of A␤, NF-B could function differently than under physiological conditions. We therefore manipulated NF-B activity by transfection of its modulators in Sw-␤APP-expressing cells. First, we established that these cells display expected responsiveness to NF-B modulation, since IKK1SE or IKK2SE expression (Fig. 6A) increases B activity that is reduced by IBSR (Fig. 6B). Second, unlike in naïve human cells where IKK1SE and IKK2SE reduce A␤ production (see Fig. 1D), these two kinases enhance A␤ production in Sw-␤APP cells (Fig. 6, C and D). We therefore examined whether this NF-B-mediated increase of A␤ levels could be accounted for by a modulation of ␤APP or secretase expression and/or activities. The promoter activities of ␤APP, BACE1, and PS1 are activated by NF-B (Fig. 7, A, B, E). IKK1SE and IKK2SE expression increases the promoter activity of ␤APP (ϩ30.1 Ϯ  48 h after transfection, luciferase and ␤-galactosidase activities were measured as described under "Material and Methods." Bars correspond to luciferase activity expressed as percent of that observed in nontreated HEK293 mock cells (Ct) and are the means Ϯ S.E. of eight independent determinations, respectively. ***, p Ͻ 0.001, ns, not statistically significant.  10.3% and ϩ108.7 Ϯ 26.4% respectively, n ϭ 14 -23, Fig. 7A), while IBSR inhibits the promoter activities of BACE1 (Ϫ17.2 Ϯ 7.3%, n ϭ 9, Fig. 7B) and PS1 (Ϫ32.8 Ϯ 4.1%, n ϭ 9, Fig. 7E). The activation of BACE1 promoter was confirmed with the human construction (kindly provided by Dr. W. Song, data not shown). Aph1 promoter activity is reduced by IKK1SE (Ϫ16.2 Ϯ 3.7%, n ϭ 12, Fig. 7C), Pen2 promoter activity is reduced by IKK2SE (-43.3 Ϯ 7.8%, n ϭ 9, Fig. 7D) and increased by IBSR (ϩ25.0 Ϯ 7.6%, n ϭ 12, Fig. 7D), nicastrin promoter activity is increased by IBSR (ϩ33.7 Ϯ 22.2% compared with control, n ϭ 6, Fig. 7G) while PS2 promoter transactivation is not affected by the modulation of NF-B activity (Fig. 7F). The examination of the changes in protein expression that could have resulted from their transcriptional modulation shows that only ␤APP and Pen-2 expressions are controlled by NF-B in Sw-␤APP-expressing cells (Fig. 8). Thus, ␤APP levels are increased by IKK1SE and IKK2SE (ϩ32.9 Ϯ 2.8% and ϩ106.7 Ϯ 28.7%, respectively, n ϭ 6, Fig. 8A) while Pen2 expression is inhibited by IKK1SE (Ϫ30.7 Ϯ 7.0%, n ϭ 8, Fig. 8D). The protein expression of BACE1, Aph1, PS1, PS2, and NCT is not affected by the transient transfection of the constructions in Sw-␤APP HEK293 cells. We ruled out any putative artifactual effect of IKK1SE, IKK2SE, and IBSR on the CMV promoter driving the Sw-␤APP used for the stable transfection by means of a CMV-␤-galactosidase construct. Thus, none of the above constructs affected galactosidase activity (supplemental Fig. S1,  A and B).

NF-B-mediated Regulation of ␤APP Expression, Secretases Activities, and A␤ Secretion under Physiological or Pathological
Conditions in SH-SY5Y Cells-To assess whether NFB-mediated control of ␤APP, secretases, and A␤ production could be cell-specific, we examined the influence of NFB modulation in a human neuroblastoma cell line, SH-SY5Y cells either stably mock-transfected or expressing Sw-␤APP (Fig. 10A). Transient transfection of SH-SY5Y cells with IKK1SE, IKK2SE, or IBSR cDNA (Fig. 10B) expectedly modulated B-luciferase activity (supplemental Fig. S1C) but not CMV-galactosidase activity (supplemental Fig. S1D), indicating first that all the NFB machinery taking place in HEK293 was also present in this neuronal cell line and second, that no interference of the constructs with CMV promoter occurs in these cells. NF-B-associated pathway mediated inhibition of ␤APP expression (see IKK1SE and IKK2SE lanes in Fig. 10C), ␤-(see IBSR lane in Fig. 10E) and ␥-secretase (see IKK1SE and IKK2SE lanes in Fig. 10G) activities and A␤ secretion (see IBSR lane in Fig. 10I) was observed in mock-transfected SH-SY5Y cells. Conversely, in Sw-␤APP cells, NF-B increases ␤APP expression (see IKK2SE lane in Fig. 10D), ␤-(see IKK1SE lane in Fig. 10F) and ␥-secretase (see IKK1SE and IKK2SE lanes in Fig. 10H) activities and A␤ secretion (see IKK1SE and IBSR lanes in Fig. 10J).

DISCUSSION
We previously established that BACE1 promoter transactivation could be triggered by overexpression of wild-type ␤APP, and to a higher extent, Sw-␤APP in human cells. This effect could be accounted for by enhanced production of A␤, since the pharmacological inhibition and mutational inactivation of ␥-secretase activity abolished the increase of BACE1 promoter transactivation while conversely, exogenous application of synthetic A␤42 promoted this effect (29). Our study showed that A␤42-mediated increase of BACE1 transactivation was NF-Bdependent (29). This agreed well with a study showing that NF-B could regulate neuronal BACE1 promoter activation (28). These data delineate a deleterious dysfunction by which overproduction of A␤ feeds its own production.
In vivo studies have suggested an NF-B-dependent regulation of A␤ production. In transgenic mice NF-B increases ␤APP levels (43), BACE1 promoter activity (44), expression (25, 45), and enzymatic activity (25, 43), ␥-secretase activity (43); and A␤ production (25, 43,46). Furthermore, the inhibition of NF-B reduced the plaque burden in mice (25, 44) and increased the learning and memory deficits of mice (43). However, to modulate NF-B activity, those studies used non-steroidal anti-inflammatory drugs (46), natural compounds (25, 43) or modulation of upstream receptors of NF-B activation pathway (44,47), which are nonspecific to NF-B activation. In the present study, we focused on the mechanisms of NF-B regulation of A␤ production; thus we transfected mutant construction of the proteins that directly activate or inhibit NFB dimers and are commonly activated by the various stimuli activating NFB.
BACE1 produces C99 that requires subsequent cleavage by ␥-secretase to yield A␤ (5). However, although NF-B-dependent regulation of ␥-secretase has been suggested (43,48,49), the direct regulation of the ␥-secretase components by NF-B has not been detailed. Furthermore, although two NF-B binding sites have been delineated on the ␤APP promoter, no data concern its putative control by NF-B in pathological conditions. Our study shows that in supraphysiological conditions, NF-B activation up-regulates some components of the high molecular weight ␥-secretase complex and the ␤APP protein in human cells from both renal and neuronal origin.
Our data demonstrate a coordinated regulation of the three principal elements of A␤ biogenesis, i.e. its precursor ␤APP, and the two enzymatic activities ␤and ␥-secretases responsible for its production. We also show, strikingly, that NF-B-dependent control of A␤ production depends on pathophysiological context, unlike the above experimental conditions aimed at mimicking the pathology. In physiological conditions, A␤-mediated NF-B-dependent pathway results in a consistent lowering of ␤APP expression as well as ␤-secretase and ␥-secretase members expression, ultimately leading to reduced enzymatic activity. Thus, in physiological conditions, NF-B contributes to maintain A␤ homeostasis at a physiological level while in pathological conditions, NF-B contributes to a vicious cycle by which A␤ self-feds its own production (Fig. 11).
It has been demonstrated that NF-B inhibits BACE1 promoter activity in neuronal and resting glial cells while it activates BACE1 promoter in A␤-exposed neuronal and activated glial cells (28). This can be explained by the nature of the NF-B dimers involved since p52/c-Rel occurs in resting conditions while, p50/p65, p52/p65, and p52/c-Rel are responsible for NF-B-mediated BACE1 transactivation in activated neurons and glia (28). In this context, it is noteworthy that Valerio et al. (17) demonstrated that A␤40 could activate NF-B by favoring the nuclear translocation of p50 and p65 subunits. Interestingly, it has been demonstrated that the recruitment of NF-B components and the build-up of the distinct heterodimeric complex could be related to A␤ levels. Thus, Arevalo et al. showed that a low concentration of A␤40 triggers NGF-like phenotype while a 40-fold higher concentration prevents NGFinduced activation of NF-B (50). Accordingly, it was reported that only low concentrations of A␤ could trigger NF-B-dependent protective phenotype (18). This set of data support our conclusion that physiological and supraphysiological levels of A␤ could differentially contribute to its own production by FIGURE 9. NF-B activation of ␤and ␥-secretase activities in Sw-␤APP HEK293 cells. A, Sw-␤APP HEK293 cells were transiently transfected with NF-B activating (IKK1SE and IKK2SE) or inhibitory (IBSR) cDNA and the ␤-secretase activity was fluorimetrically measured as described under "Material and Methods." Bars correspond to ␤-secretase activity expressed as percent of that observed in mock-transfected cells (Ct) and are the means Ϯ S.E. of 4 -6 independent determinations. *, p Ͼ 0.05, ns, not statistically significant. B and C, Sw-␤APP HEK293 cells were transiently transfected with IKK1SE, IKK2SE, or IBSR and an in vitro ␥-secretase assay was performed as described under "Material and Methods" (B) and A␤ and AICD were quantified as described under "Materials and Methods" (C and D). Bars are expressed as percent of that observed in mock-transfected cells (Ct) and are the means Ϯ S.E. of 4 -6 independent determinations. *, p Ͼ 0.05; ns, not statistically significant. modulating its precursor and secretases via an NF-B-dependent mechanism.
Our data do not establish whether NF-B-associated increase in promoter transcription is direct or involves intracellular intermediates. One or several consensus sequence sites have been identified on the ␤APP promoter (51), human (19), and rat (28) BACE1 promoters as well as on the PS2 promoter (52) that could mediate a physical interaction between NF-B and these promoters. An additional possibility lies on the control of the tumor suppressor p53 by NF-B. In this matter, the literature indicates that NF-B could either activate p53 (53) or antagonize its effect (54), leading to either antiapoptotic and/or proapoptotic phenotypes (55). Interestingly, several studies indicated that p53 represses PS1 (56 -59), and we have shown that, at physiological concentrations of A␤, p53 also lowers Pen-2 promoter transactivation (60,61). If one accepts the view that NF-B activates p53 (53), this agrees with our observation that in a physiological context, p53-mediated NF-B-linked reduction of PS1, Pen-2, and other contributors of A␤ genesis (see Figs. 2 and 3) ultimately leads to decreased ␥-secretase expression and activity. Alternatively, if one considers NF-B as a repressor of p53, then one can envision that NF-B-mediated p53-dependent effect could only occur in pathological conditions where PS1 promoter transactivation is enhanced (see Fig.  7). It is actually difficult to resolve this issue in the absence of data concerning the influence of A␤ concentration on the control of p53 by NF-B. But there again, this presumes a differential NF-B-p53-dependent control of A␤ in physiological and supraphysiological conditions. NF-B is an ubiquitous transcription factor activated by inflammation, oxidative, and others cellular stresses (11,13). This activation results in a protective response aimed at restoring cellular homeostasis, but can be deleterious when becoming chronic. This can be compared with NF-B-mediated control of A␤ production that protects against A␤ overload in physiological conditions while it contributes to perpetuate and even increase A␤ levels in a more pathological context.
Besides classical strategies aimed at reducing A␤ production by directly modulating either ␤or ␥-secretases with pharmacological probes or by neutralizing A␤-associated effects by a vaccinal approach (62,63), alternative therapeutic tracks targeting NF-B could be theoretically envisioned. One of these concerns is non-steroidal anti-inflammatory drug (NSAID) treatment. In these pathological conditions, our data show that NSAID-associated down-regulation of NF-B could have beneficial effects on A␤ load even if deleterious side effects of interfering with NF-B pathways should not be underestimated. FIGURE 11. Scheme of A␤-NF-B interplay in physiopathological conditions. A, endogenous A␤ activates NF-B which, in turn lowers ␤APP, BACE1, and ␥-secretase activity, thereby lowering A␤. This allows maintaining A␤ homeostasis in physiological conditions. B, at high A␤ concentrations, NF-B mostly increases ␤APP and thereby increases A␤ in regulatory loop favoring exacerbated A␤ production.