Mutant Presenilin 1 Increases the Expression and Activity of BACE1*

Mutations of the presenilin 1 (PS1) gene are the most common cause of early onset familial Alzheimer disease (FAD). PS1 mutations alter the activity of the γ-secretase on the β-amyloid precursor protein (APP), leading to selective overproduction of β-amyloid (Aβ) 42 peptides, the species that forms oligomers that may exert toxic effects on neurons. Here we show that PS1 mutations, expressed both transiently and stably, in non-neuronal and neuronal cell lines increase the expression and the activity of the β-secretase (BACE1), the rate-limiting step of Aβ production. Also, BACE1 expression and activity are elevated in brains of PS1 mutant knock-in mice compared with wild type littermates as well as in cerebral cortex of FAD cases bearing various PS1 mutations compared with in sporadic AD cases and controls. The up-regulation of BACE1 by PS1 mutations requires the γ-secretase cleavage of APP and is proportional to the amount of secreted Aβ42. Aβ42, and not AICD (APP intracellular domain), is indeed the APP derivative that mediates the overexpression of BACE1. The effect of PS1 mutations on BACE1 may contribute to determine the wide clinical and pathological phenotype of early onset FAD.

The ␤-amyloid peptide (A␤) 2 that accumulates in vulnerable brain regions in Alzheimer disease (AD) is released from the ␤-amyloid precursor protein (APP) by sequential cleavages by ␤-secretase and ␥-secretase. A single protein called BACE1 is responsible for ␤-secretase activity, whereas ␥-secretase involves at least four proteins including a catalytic subunit called presenilin-1 (PS1) (1,2). Mutations of PS1 are the most common cause of early onset familial AD (FAD). The known effect of PS1 mutations on APP processing is the increased production of A␤ species ending at residue 42, which aggregates faster than the A␤40 isoform and accumulates in the brain in the state of soluble low molecular weight oligomers (3,4). Small, soluble, and diffusible aggregates composed of a mixture of full-length and N-terminal-truncated A␤42 species appear early in the cerebral cortex of subjects at risk of Alzheimer disease pathology (5). The rate of accumulation as well as the properties of aggregation and toxicity of cerebral soluble A␤ depend on the ratio of the three major A␤ species, 1-42, pyroglutamate 3-42, and pyroglutamate 11-42 (6). We have shown that, in the cerebral cortex of FAD cases with mutations of PS1, the relative percentage of the two N-terminal-truncated A␤ species is significantly increased in comparison to sporadic AD cases (7). A relative increase of N-terminal-truncated A␤ peptides also occurs in the brain of transgenic mice bearing a double PS1 mutation (8). When overexpressed in cell lines, BACE1 increases the production of A␤ 11-x (9) and A␤ 3-x peptides (10) in vitro. BACE1 has been found to be transcriptionally regulated by several different mechanisms involving various transcription factors and pathways (11)(12)(13)(14)(15), and given the complexity and richness in transcription factor recognition sites of its gene promoter, it is likely to be a highly regulated protein (16). We and other research groups have shown that oxidative stress up-regulates the expression of BACE1 through the activity of the ␥-secretase (14,17). Moreover, we showed that the presence of presenilins is necessary for obtaining the increase of BACE1 expression under stress conditions (17,18). On this basis, we predicted that mutant PS1 might augment the expression of BACE1 by altering the activity of the ␥-secretase.
icillin, 100 g/ml streptomycin, 2 mM L-glutamine (Euroclone, Milano, Italy). HEK-293 APPwt stable cells (provided by Dr. Luciano D'Adamio, Albert Einstein College of Medicine, Bronx, NY) were cultured similarly, except for the addition of puromycin at 5 g/ml. Mouse embryonic fibroblast (MEF) wt, PS1 Ϫ/Ϫ , and PS dko (provided by Dr. Bart De Strooper, Center for Human Genetics, VIB4 and KULeuven, Leuven, Belgium) (19), APP/APLP2 dko (provided by Dr. Ulrike Mueller, Max Planck Institute for Brain Research, Frankfurt, Germany) (20,21), and Fe65 ko (provided by Dr. Tommaso Russo, CEINGE, Napoli, Italy) were cultured in low glucose Dulbecco's modified Eagle's medium with 10% (v/v) fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM L-glutamine (Euroclone, Milano, Italy). SH-SY5Y and SKNBE(2c) cells were cultured in RPMI 1640 supplemented with 10% (v/v) fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM L-glutamine, 1% nonessential amino acids, and 1% sodium pyruvate (all from Invitrogen). M17 cells were obtained from Dr. Xinglong Wang and Dr. Xiongwei Zhu at the Department of Pathology, Case Western Reserve University, Cleveland, Ohio 44106 and cultured in Opti-MEM with 10% fetal bovine serum and 1% penicillin/streptomycin and with 50 g/ml G418. Transient transfection of empty vector (pcDNA3.1, Invitrogen), wild type, and mutant PS1 cDNA (M146V, S170F, L392V, and D385A) and BACE1 in HEK-293 and HEK-293 APPwt cells was carried out with polyethyleneimine (Sigma) at 12 l/20 g of DNA according to the manufacturer's instructions. Transient transfection of empty vector, wild type, and mutants PS1 cDNA (M146V, S170F, L392V, and D385A) in MEFs and SKNBE(2c) was carried out with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. In particular, to improve transfection efficiency in MEFs, cells were transfected in penicillin/streptomycin-free medium at 70% optical confluence. Efficiency of transfection was checked with the use of a green fluorescent protein coding plasmid. APP reconstitution in MEFs APP/APLP2 dko was performed by transfecting a human APP 695 construct together with pcDNA3.1, PS1 wt, or PS1 S170F using Lipofectamine 2000 as described above. For mRNA analysis, cells were collected 4 h after transfection; for protein analysis, cells were collected 12 h after transfection. The ␥-secretase inhibitor XIX (Calbiochem) was added at 25 nM to the cells at the time of transfection and replenished every 4 h for 12 h. The BACE1 Inhibitor IV (22) was added to the cells overnight before trans-FIGURE 1. PS1 mutations induce the transcription and activity of BACE1. A-C, HEK-293 APPwt cells were transfected with empty vector, wild type PS1, or mutant PS1 constructs. BACE1 mRNA expression (A), BACE1 protein levels (B) and activity (C) were analyzed; also, indication of BACE1 activity is given by the analysis of ␤APPs on cell culture media. D-F, MEFs deficient for PS1 were transfected with control, wild type, and mutant plasmids. Quantification of BACE1 mRNA (D), protein levels (E), and activity (F) was performed. Data represent the mean and S.E. of three or more experiments; significance is relative to empty vector control. Statistical analysis was performed with ANOVA and Bonferroni post-test.
fection and after transfection until cell harvest at 15 nM (Calbiochem).
Generation of SH-SY5Y PS1 S170F Stable Cell Lines-Cells were transfected with SspI linearized and purified pcDNA3.1, PS1 wt, and PS1 S170F constructs using Lipofectamine 2000 according to the manufacturer's protocol. G418 was used as mean of selection up to 800 g/ml in culture medium until resistant clones could be identified. Clones were picked, trypsinized, and transferred to 96 wells in triplicate. DNA was then extracted by NaCl-ethanol precipitation. PCR was conducted with FIGURE 2. PS1 mutations stably expressed in neuronal cell lines determine BACE1 up-regulation. SH-SY5Y, stably transfected with pcDNA3.1, PS1 wt, and PS1 mutant S170F, were analyzed for BACE1 mRNA (A), BACE1 protein levels (B), and BACE1 activity (C). Similarly, M17 neuronal cells, stably expressing PS1 A246E mutant, PS1 wt, and pcDNA3.1, were analyzed: D, BACE1 mRNA; E, BACE1 protein levels; F, BACE1 activity. G and H, sandwich enzyme-linked immunosorbent assay for A␤40 and -42, performed analyzing conditioned medium of SH-SY5Y stable cells (G) and M17 stable cells (F). The A␤42/40 ratio is also presented. Data represent the mean and S.E. of three or more experiments, performed on two different stable cell clones; significance is relative to empty vector control (CTR); for enzyme-linked immunosorbent assay, p values are specified in the figure. Statistical analysis was performed with ANOVA and Bonferroni post-test. the following primers: PS1 277-S 5Ј-gcc acc ctg agc cat tat ct-3Ј and Universal primer bovine growth hormone primer sequence as antisense primer. Positive clones were sequenced to ascertain the presence or not of the mutation using the primer PS1 291-313 S: 5Ј-ggc tac cat taa gtc agt cag c-3Ј. At least two different clones for each stable cell line were analyzed.
Evaluation of BACE1 activity, PS1, BACE1, and APP Protein Levels-BACE1 enzymatic activity was measured in frozen samples of human, mouse brain tissues, and in cultured cells using a fluorometric reaction kit (R&D Systems, Inc., Minneapolis, MN) according to the manufacturer's protocol. More details are in the supplemental "Experiment Procedures." For quantification of PS1, BACE1, and APP protein levels, 50 g of total proteins (obtained by cell lysis R&D buffer, R&D Systems, according to the suggested protocol) were loaded on 9% Trisglycine SDS-PAGE gels (unless otherwise specified) and transferred onto polyvinylidene difluoride membranes (Amersham Biosciences). Blots were probed with an anti-PS1 polyclonal antibody recognizing a band of about 20 kDa, corresponding to the C-terminal fragment of the protein (1:100; Cell Signaling Inc., Beverly, MA) with an anti-BACE1 polyclonal antibody (1:300; Calbiochem) and with a monoclonal antibody anti-APP (clone 22C11, 1:250 Chemicon International, Temecula, CA); the reactive bands were revealed with ECL plus (Amersham Biosciences). To normalize protein levels of PS1, APP, and of the mature form of BACE1 (75 kDa), membranes were stripped with Restore Western blot stripping reagent (Pierce) and probed with a monoclonal antibody against ␤-actin (1:6000; Sigma). Western blot determination of PS1 CTFs and A␤ peptides was performed by running protein samples on 10 -18% Tris-Tricine gel and transferring the proteins on a 0.20 m nitrocellulose membrane (Bio-Rad). The density of the specific bands was quantified using Quantity One software system (Bio-Rad).
RNA Interference-APP 695 RNA silencing was accomplished using the SMARTpool siRNA (Dharmacon, Lafayette, CO) consisting of four duplexes, all designed to target distinct sites within the APP gene. A non-targeting siRNA pool was used as a negative control. siRNAs (100 nM) were transfected in HEK-293 cells according to the manufacturer's instructions using DharmaFect One transfection reagent (Dharmacon); cells were harvested 48 h after siRNA transfection for mRNA analysis and after 72 h for protein analysis. To verify the inhibition rate after RNA, silencing levels of APP were evaluated by reverse transcription-PCR (forward, 5Ј-gtt ccg agg ggt aga gtt tg-3Ј; reverse 5Ј-cct ggg aca ttc tct ctc gg-3Ј) and by Western blot (monoclonal antibody 22C11; Chemicon International).

Presenilin 1 Mutations Determine an Increase in BACE1
Activity, Based on Its Augmented Transcription-HEK-293 cells, stably overexpressing APP wild type, were transiently transfected (10, 23) with empty vector control (pcDNA3.1), PS1 wild type, and constructs coding for three PS1 mutations  (4 males and 4 females) and agematched littermates (4 males and 3 females) were analyzed. mRNA was extracted from one dissected brain hemisphere, whereas protein extraction was performed on the other hemisphere. A, BACE1 mRNA. B, BACE1 protein levels. C, an aliquot of brain protein lysates was also used to perform BACE1 activity assay. D and E, frozen brains from 11 FAD patients carrying 10 different mutations, 10 sporadic AD (SAD) patients, and 12 control (CTR) subjects were processed likewise. D, BACE1 mRNA levels; E, BACE1 activity. Statistical analysis was performed with t test (A-C) and ANOVA (D-E) and the Bonferroni post-test.
We then asked whether the presence of endogenous wild type PS1 could affect the extent of BACE1 gene transcription. Mouse embryonic fibroblasts lacking PS1 (MEFs PS1 Ϫ/Ϫ ) transfected with PS1 mutants showed an increase of BACE1 mRNA (42-93%) compared with cells transfected with wild type PS1, higher than that observed in HEK-293 APPwt cells, indicating that the up-regulation of BACE1 determined by the PS1 mutations was not compensated by the presence of the wild type PS1; also, they showed an increase of BACE1 protein levels (65-311%) and activity (32-73%) compared with cells transfected with wild type PS1 (Fig. 1, D-F). Again, the PS1 aspartate mutant did not have any influence on BACE1.
As shown in Fig. 1, B and E, the PS1 D385A CTF is barely visible, as this mutant undergoes little endoproteolysis (28). Full-length D385A PS1 was detected, ascertaining the efficiency of transfection for this mutant as well (not shown).
To verify the specificity of BACE1 antibody on Western blot, we transfected HEK-293 cells either with the empty vector pcDNA3.1 or with a plasmid coding for the full-length BACE1. Western blot analysis (supplemental Fig. 1A) showed the same pattern for transfected and endogenous BACE1. Also, to verify the specificity of the BACE1 activity assay, we pretreated HEK-293 APPwt cells with a BACE1 inhibitor (22), then transfected them with either wild type PS1 or the S170F mutant. As predicted, the inhibitor prevented the PS1-mediated increase of BACE1 activity observed in non-treated cells (supplemental Fig. 1B).
Several reports have recently shown how BACE1 levels can be controlled by post-translational and post-transcriptional mechanisms (11-15, 17, 18, 29 -33). To further confirm the A, HEK-293 APPwt cells were treated with a transition analogue inhibitor of ␥-secretase at the moment of transfection with empty vector control, PS1 wild type, and PS1 S170F and replenished twice every 4 h; BACE1 protein levels were analyzed. B, MEFs PS1 Ϫ/Ϫ were treated similarly, and BACE1 protein levels were analyzed. Basal levels of BACE1 mRNA (C), protein levels (D), and activity (E) were evaluated in MEFs PS1 Ϫ/Ϫ , all resulting lower than in wild type cells. F, MEFs deficient for both PS1 and PS2 were transfected with empty vector control or with PS1 and PS2: reconstitution of PS1/2 expression determined an increased expression of BACE1; PS1/2 CTFs are produced as expected. Data represent the mean and S.E. of three or more experiments; significance is relative to the empty vector control or to control cells. Statistical analysis was performed with ANOVA (A and B) and Bonferroni post-test and t test (C-F).

PS1 Mutations Control BACE1 Production
effect of mutant PS1 on BACE1 expression, we generated a line of stable neuroblastoma cells (SHSY-5Y), expressing wild type PS1 and the S170F mutation (34). Also, we utilized a similar pre-established line of stable cells, M17, carrying wild type PS1 and the A246E mutation. PS1 A246E stable cells showed (Fig.  2H) a little increase of soluble secreted A␤40 and a much higher increase of A␤42 (ϩ24 and 115%, respectively). PS1S170F and A246E mutations determined a significant increase in the levels of BACE1 mRNA (ϩ82.74 and ϩ94.14%), protein levels (ϩ196 and ϩ40.7%), and activity (ϩ40 and ϩ 81%) compared with control cells. (Fig. 2, A-E). As shown by Western blot analysis, SH and M17 stable clones show adequate levels and pattern of PS1 expression and maturation (Fig. 2, B and E). Of note, stable SH-SY5Y cells have levels of PS1 mRNA, measured by Taqman real time PCR, of 1.3-1.9ϫ those of non-transfected cells; similarly, M17 show levels of 1.9 -3.0ϫ those of wild type M17 cells (data not shown). PS1 S170F stable cells showed (Fig. 2G) an increase of soluble secreted A␤40 and -42 (ϩ34 and ϩ71%, respectively), as also reported before (35). Both cell lines showed an increase of the A␤42/40 ratio (S170F, ϩ27%; A246E, ϩ73%), although this was less evident in the S170F mutant, where A␤40 levels are higher.
Mutant PS1 Up-regulates BACE1 in Vivo-To investigate if the same effect observed in vitro is also present in vivo, we first evaluated BACE1 parameters in the brains of mutant PS1 knock-in mice (PS1 M146V KI) that express physiological levels of the human mutant protein. The levels of BACE1 mRNA and activity were significantly increased by 41 and 51% in the brain of mutant PS1 knock-in mice compared with the wild type littermates (Fig. 3, A and C), whereas the 47% elevation of BACE1 protein levels did not reach statistical significance (Fig. 3B). Up-regulation of BACE1 was also detected in brains of FAD subjects bearing various PS1 mutations (supplemental Table 1). Levels of BACE1 mRNA and activity were significantly increased on average by 80 and 36%, respectively, in cerebral cortices of 11 FAD cases bearing 10 different PS1 mutations compared with age-matched normal controls (Fig. 3, D and E). Interestingly, in sporadic AD, levels of BACE1 activity, and not mRNA, significantly differ (ϩ24%) from controls, as previously reported (10,36,37). supplemental Fig. 1, C and D, show BACE1 mRNA and activity levels in each FAD case described, expressed as relative values of the mean of the control cases average.
The Up-regulation of BACE1 Requires the Presenilins and the Activity of the ␥-Secretase-As we have recently shown (17,18), PS1/2 and the activity of the ␥-secretase are required to obtain the up-regulation of BACE1 induced by oxidative stress. Thus, we asked if the action of the PS1 mutants on BACE1 is also mediated by their involvement in the ␥-secretase activity. In HEK-293 APPwt (Fig. 4A) and MEFs PS1 Ϫ/Ϫ (Fig. 4B) cells, transiently transfected with PS1 mutant S170F, BACE1 protein levels were not altered in the presence of a transition state analogue inhibitor of the ␥-secretase, suggesting that the up-regulation of BACE1 induced by the PS1 mutations depends on the activity of the ␥-secretase. Indeed, in MEFs PS1 Ϫ/Ϫ the basal levels of BACE1 mRNA, protein levels, and activity are lower than in wild type MEFs (Ϫ30, Ϫ38, and Ϫ20%, respectively) ( Fig. 4, C-E). As shown before (Fig. 4B, empty vector versus PS1 wt in cells not treated with ␥-inhibitor), reconstitution of PS1 in these cells reestablishes the basal values of BACE1 protein levels. Furthermore, in MEFs lacking both presenilins (MEFs PS dko), BACE1 protein levels increase upon reconstitution of PS1/2 by transient transfection (Fig. 4F), as shown by the detection of PS1/2 CTFs.
The Up-regulation of BACE1 Requires APP-Because PS1 mutations increase the expression of BACE1 by a mechanism that requires the ␥-secretase activity, we asked whether this effect depends on the presence of APP, one major substrate of ␥-secretase. MEFs deficient for APP and APP-like protein 2 (MEFs APP/APLP2 dko, or APP dko), also naturally lacking APLP1, were transfected with empty vector, wild type PS1, and PS1 mutants. They did not show any variation of BACE1 mRNA, protein levels, and activity compared with wild type MEFs (Fig. 5, A-C, first and second set of columns). When APP 695 was reconstituted into MEFs APP dko cells, the concomitant transfection of PS1 S170F mutant determined a significant increase of BACE1 parameters. (Fig. 5, A-C, third set of columns). The requirement of APP for the up-regulation of BACE1 was confirmed by another approach. Silencing of APP by RNA interference in HEK-293 cells, subsequently transfected with the PS1 S170F mutant, prevented the increase of BACE1 mRNA, protein levels, and activity (Fig. 5, D-F).
A␤42 Is the APP Derivative That Mediates the Up-regulation of BACE1-The APP derivatives that result from the ␥-secretase cleavage are the A␤ peptides and the AICD. We investigated which derivative is responsible for BACE1 up-regulation. We first analyzed the role of AICD, which is generated by the ␥or ⑀-cleavage of APP together with the A␤ peptides (38). We cloned several AICD fragments into a pcDNA3.1 vector; AICD 57 and 59 are the theoretical APP derivatives resulting form the ␥-cleavage of APP and corresponding to A␤42 and -40, respectively; AICD 50 and 51 are the ⑀-cleavage derivatives resulting from the processing of APP (39 -41). Transfection of several cell lines (SKNBE-2c, SH-SY5Y, HEK-293, and MEFs; data from at least three experiments for each cell line, all showing similar outcomes, were pulled together) with constructs coding for AICD 50, 51, 57, and 59 determined no change of BACE1 mRNA compared with the empty vector control (Fig. 6A). To test for AICD expression, we performed a semiquantitative PCR, starting from the cDNA of the same cells transfected with AICD. The lower panel in A shows how AICD is amplified only in transfected cells, whereas the endogenous APP-derived FIGURE 5. The activation of BACE1 by mutant PS1 requires APP. MEFs from APP/APLP2-deficient mice were transfected with empty vector control, wild type, and all the mutants considered in Fig. 1. Furthermore, MEFs APP DKO were transfected with human APP 695, reconstituting its expression, and with a representative PS1 mutant (S170F). Quantification of BACE1 mRNA (A), protein levels (B), and activity (C) were carried out as previously described. D-F, APP knock down (E, Western blot shown, performed with APP N-terminal-specific antibody) was performed on HEK-293 cells. Cells were transfected with nontargeting siRNA or Homo sapien APP-specific siRNA in 2 pulses, at times 0 and 24 h later; empty vector control, PS1 wild type, and PS1 S170F mutant were transfected subsequently, and quantification of BACE1 mRNA (D), protein levels (E), and activity (F) were performed. Data represent the mean and S.E. of three or more experiments; significance is relative to empty vector control. Statistical analysis was performed with ANOVA and the Bonferroni post-test. CTR, control. amplicon appears as a faint band. More permissive PCR conditions allowed for the APP template to be amplified to more detectable levels (not shown). AICD is thought to act, whether in the nucleus or not, through the binding with chaperones Fe65 and Tip60 (42,43). We tested whether MEFs lacking Fe65 showed different behavior upon transfection with PS1 mutant S170F compared with wild type cells. As shown in Fig. 6B, both cell lines yielded an increase in BACE1 mRNA levels upon transfection with the cited mutant, thus ruling out a role for the AICD-Fe65 complex in the control of BACE1 transcription. A further confirmation that AICD does not have a role in BACE1 transcription control resulted from the analysis of brain tissue of transgenic mice (tg) overexpressing AICD. Mice were generated on an FVB background, with the transgene placed under the control of calmodulin kinase II-␣ promoter (driving the expression in the forebrain) and have been recently described and characterized (44). Seven tg and 7 littermates carrying the AICD 50, 57, or 59 transgene were analyzed at 4 weeks of age for forebrain BACE1 mRNA expression; we did not detect any difference between the tg and littermate mice (Fig. 6C).
We, thus, concentrated our attention on A␤ peptides. It is thought that the most toxic A␤ species are low molecular weight oligomers, which form when soluble monomers are let in aqueous milieu at for a certain period of time (45). Before using A␤ on our cells, we checked the aggregation state of the used preparations. As shown in supplemental Fig. 2A  Transfected AICD transcript was detected by PCR conducted on the same cDNAs used for BACE1 real time PCR experiments, and a representative panel of semiquantitative PCR is shown. B, wild type and Fe65-deficient MEFs were transfected with empty vector, PS1 wt, or PS1 S170F. BACE1 transcription levels were determined. Data represent the mean and S.E. of three or more experiments; significance is relative to empty vector control. C, seven 4-week-old AICD transgenic mice (tg) and their littermates were analyzed for BACE1 mRAN expression. AICD length (50, 57, or 59 amino acids), mouse sex (F or M) and tg line (1-5, 5-1, 4-4) are indicated. Data from tg mice are compared with each corresponding littermate control (each control (CTR) ϭ 100). Statistical analysis was performed with ANOVA and Bonferroni post-test.
SH-SY5Y and SKNBE-2c cells (data from at least three experiments for each cell line, all showing similar outcome, were pulled together) were treated with the same peptides (Fig. 7B); again, only A␤-(1-42) determined a significant increase (ϩ62%) in BACE1 mRNA.
To rule out a possible toxic effect of A␤, we have evaluated cell viability (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay) and oxidative stress (reactive oxygen species production, GSH depletion) in SH-SY5Y cells exposed to A␤ species under the same experimental conditions which determine the up-regulation of BACE1. We employed 20 M H 2 O 2 as positive controls for oxidative stress production and cell damage. Compared with A␤ free controls, at 1 h of incubation we detected only a 13% decrease of cell viability with A␤42 and a 31% decrease with H 2 O 2 ; A␤42 determined a doubling of reactive oxygen species production but only 15% decrease of FIGURE 7. A␤-(1-42) peptide is the APP derivative responsible for the up-regulation of BACE1. A, primary neurons, obtained from a 17.5 pregnant FVB mouse, and cultured for 21 days (similar data for 14 days in vitro neurons not shown) were treated for 1 h with freshly dissolved 1 M A␤ peptides or with medium alone. BACE1 transcription levels were determined. Nine wells for each condition from three different neuronal cultures were analyzed as single experimental points. B, a series of neuronal cell lines (SKNBE(2c) and SH-SY5Y) were treated similarly at least 18 h after plating with 1 M A␤ peptides or with medium alone for 1 h. mRNA was extracted as described, and TaqMan Real Time PCR for BACE1 was performed. Data are a pool of several experimental repeats, performed at least in three repeats for each different cell line, which yielded analogous results; they represent the mean and S.E. C and E, HEK-293 APPwt cells were transfected with empty vector, wild type PS1, and mutant PS1 constructs. A␤ species were quantified in conditioned media by enzyme-linked immunosorbent assay. D and F, BACE1 activity was also measured in these cells and correlated to the levels of A␤40 and A␤42. Statistical analysis was performed with ANOVA and the Bonferroni post-test and the study of linear correlation (D and F). CTR, control.
GSH, as compared with a 4ϫ increase in reactive oxygen species and 80% decrease of GSH induced by H 2 O 2 . Non-significant modifications from controls were obtained with A␤40 and A␤42s. These data argue against a role for toxicity and oxidative stress in A␤42-induced BACE1 up-regulation. Further data confirmed these results. HEK-293 APPwt cells were transfected with the three PS1 mutants, and secreted A␤ was measured by enzyme-linked immunosorbent assay. PS1 mutants showed altogether an increased secretion of A␤42, which significantly correlated with BACE1 activity (Fig. 7, C and D). Of note, the levels of A␤40 did not correlate with those of BACE1 activity (Fig. 7, E and F). As PS1 mutations determine an increase of A␤-(1-42) secreted in the extracellular medium (Fig. 7C), it is likely that secreted A␤ is responsible for the effect on BACE1. We, thus, predicted that an antibody directed toward A␤42 could sequester the peptide so as to inhibit its signaling/pathologic action, as already suggested (46). Indeed, HEK-293 APPwt transfected with PS1 S170F mutant did not show an increase of BACE1 mRNA upon incubation with an antibody specific for A␤42, whereas neither anti-A␤40 nor anti ␤-actin antibodies showed this effect (supplemental Fig. 2D).

DISCUSSION
The known effect of PS1 mutations is the alteration of the ␥-secretase cleavage on APP, leading to the relative increased production of A␤42 species (3,4). We showed that PS1 mutations also augment the expression of BACE1 and that this effect is dependent on the presence of APP and is proportional to the amount of A␤42 produced and secreted in the extracellular milieu. This novel effect of PS1 mutations implies the existence of a positive feedback loop from the ␥to the ␤-secretase cleavages of APP in which A␤42 is the APP derivative that acts on BACE1 transcription, as previously hinted under stress conditions (14,17).
The activity of BACE1 is increased in familial early onset (the present study) as well as in sporadic late-onset AD (32,36,37,47). In the latter conditions, various factors such as oxidative stress (10,48,49), hypoxia (50,51), and ischemia (52) have been proposed as causes of BACE1 up-regulation, and different underlying mechanisms have been described, including the activation of hypoxia-inducible factor 1 (53), the decrease of GGA3 activity (30), and the loss of micro RNA specific clusters (32).
Additional non-transcriptional mechanisms have been hypothesized to account for increased BACE1 protein levels and activity in sporadic AD: maturation of BACE1 may be controlled by PS1 itself (33); expression of a BACE1 antisense transcripts (54), which respond to cellular stresses and A␤ itself, seem to stabilize BACE1 mRNA, thus increasing its protein expression; alternative splicing of BACE1 pre-mRNA has been found to be a control system as well (31); under ischemia, stabilization of BACE1 protein levels, because of defective degradation, is obtained by depletion of GGA3 (30). In all these cases BACE1 mRNA production is not necessarily augmented.
We show that the primary mechanism by which PS1 mutants act is through an increase of BACE1 transcript. The main actor seems to be A␤42, but the signaling pathway, starting form the action of secreted A␤42 down to nuclear effectors, has still to be elucidated. We have previously shown that a potential pathway is a c-Jun NH 2 -terminal kinase (JNK)/AP1 cascade, at least in stressful conditions (18). Hypothetically, soluble oligomers of A␤42 may act through the binding with a cellular receptor, located either on the plasma membrane or at intracellular sites (55)(56)(57), activating a BACE1-inducing signaling, or more in general, membrane damage generated by oligomers (58,59) may trigger a stress cellular response leading to enhanced BACE1 expression.
FAD PS1 mutations result in a wide and heterogeneous clinical phenotype that includes atypical presentations, such as ataxia, paraparesis, and epilepsy (35,60,61). The up-regulation of BACE1 determined by PS1 mutations may significantly contribute to determine the phenotype of FAD. Indeed, the overexpression of BACE1 increases the production of N-terminaltruncated A␤ species (9). We and others (62) have shown that the composition of soluble A␤ reflects the pathological and clinical phenotype of A␤ amyloidosis and that the prevalence of N-terminal-truncated A␤42 peptides is correlated with the rate of aggregation and with the degree of toxicity of the mixture of A␤ species (6). Moreover, recent data obtained with different animal models of A␤ amyloidosis support the hypothesis that the composition of A␤ species, indicated as "A␤ strains" (63)(64)(65), dictate the conformation of A␤-soluble aggregates, which in turn produces different pathological phenotypes.