BRI2 Protein Regulates β-Amyloid Degradation by Increasing Levels of Secreted Insulin-degrading Enzyme (IDE)*♦

Background: The British precursor protein (BRI2) influences amyloid precursor protein metabolism. Results: BRI2 lowers β-amyloid peptide levels by increasing levels of secreted insulin-degrading enzyme (IDE) in both cells and mice. Conclusion: BRI2 as a receptor protein regulates IDE levels and in turn promotes β-amyloid degradation. Significance: Targeting the regulation of IDE may lead to new approaches to therapeutically address sporadic Alzheimer disease. The amyloid precursor protein (APP) is one of the major proteins involved in Alzheimer disease (AD). Proteolytic cleavage of APP gives rise to amyloid-β (Aβ) peptides that aggregate and deposit extensively in the brain of AD patients. Although the increase in levels of aberrantly folded Aβ peptide is considered to be important to disease pathogenesis, the regulation of APP processing and Aβ metabolism is not fully understood. Recently, the British precursor protein (BRI2, ITM2B) has been implicated in influencing APP processing in cells and Aβ deposition in vivo. Here, we show that the wild type BRI2 protein reduces plaque load in an AD mouse model, similar to its disease-associated mutant form, ADan precursor protein (ADanPP), and analyze in more detail the mechanism of how BRI2 and ADanPP influence APP processing and Aβ metabolism. We find that overexpression of either BRI2 or ADanPP reduces extracellular Aβ by increasing levels of secreted insulin-degrading enzyme (IDE), a major Aβ-degrading protease. This effect is also observed with BRI2 lacking its C-terminal 23-amino acid peptide sequence. Our results suggest that BRI2 might act as a receptor protein that regulates IDE levels that in turn influences APP metabolism in a previously unrecognized way. Targeting the regulation of IDE may be a promising therapeutic approach to sporadic AD.

Alzheimer disease (AD) 3 is a disease of progressive dementia and neuron loss characterized by the deposition of ␤-amyloid plaques and the formation of neurofibrillary tangles in the brain (1). The main components of amyloid plaques, the amyloid-␤ (A␤) peptides, are derived by sequential proteolytic cleavage of the amyloid precursor protein (APP) (2). Although mutations in the APP or the PS1 or PS2 genes, leading to alterations in APP processing, are genetically linked to familial cases of AD (3), increased levels of amyloidogenic A␤ peptides are believed to play a major role in disease pathogenesis (4). However, besides A␤ generation, A␤ clearance is of equal importance in maintaining A␤ steady-state levels. The metalloproteases neprilysin, endothelin-converting enzyme (ECE-1 and ECE-2), and insulin-degrading enzyme (IDE) are the main A␤-degrading enzymes in the brain (5,6) and reduced clearance of A␤ peptides may be more related to cases of sporadic AD (7).
Recently, the British precursor protein (BRI2, ITM2B, and E25) (8 -10) has been implicated to influence APP processing in cells and A␤ aggregation in vivo (11)(12)(13)(14)(15). Although mutations in the BRI2 gene have been linked to familial Danish dementia and familial British dementia (10,16), the physiological function of wild type BRI2 is not known. The 266-aa type II transmembrane protein is processed by several proteases to shed a 23-aa peptide via furin-like cleavage (17), and a fragment, including the BRICHOS domain (18), into the extracellular space (19). In familial Danish dementia, an 11-aa longer protein, ADan precursor protein (ADanPP), is expressed due to a 10-nucleotide duplication before the stop codon of the BRI2 gene causing a frameshift and elongation of the open reading frame (16). Furin-like cleavage of ADanPP yields a 34-aa peptide prone to amyloid formation (16). BRI2 can interact with APP and has been proposed to inhibit secretase cleavage of APP (11,(13)(14)(15). Other studies report an interaction of the 23-aa BRI peptide, or the BRICHOS domain, with A␤ in vitro, and a potential inhibition of A␤ aggregation in vivo (12,20).
We previously showed that even the disease-related form of BRI2, ADanPP, reduces plaque deposition in mouse models of AD (21). Given that these two dementia-related proteins can interact and influence each other, this prompted us to further investigate the underlying mechanisms of the observed A␤ plaque reduction. We suspected that not only A␤ generation but also A␤ clearance by degrading proteases may be affected, because A␤ as well as BRI2-derived peptides can be degraded by IDE (22), one of the major A␤-degrading enzymes in the brain.

EXPERIMENTAL PROCEDURES
Plasmids-The cDNA encoding the human WT form of BRI2 in vector pCR2.1 (provided by R. Vidal, Indianapolis, IN) was mutated by standard mutagenesis using PfuUltra High Fidelity DNA polymerase (Stratagene) to introduce the Danish mutation (10-nucleotide insertion TTTAATTTGT). BRI2 and ADanPP cDNA inserts were liberated by digest with BamHI and introduced blunt-ended into the EcoRV cloning site of pcDNA3.1/Zeo(ϩ) (Invitrogen). To generate the BRI2⌬ construct, a stop codon was introduced following the triplet encoding aa 243 of BRI2 by site-directed mutagenesis of the pcDNA3.1/Zeo(ϩ)-BRI2 construct. All cDNA constructs were verified by sequencing. The CMV wild type ␤APP695 cDNA construct (23) was used to express APP695 (24). A human PS1 cDNA mutated at L166P (25) was inserted into pcDNA3.1/Zeo(ϩ) (Invitrogen). pEGFP-C1 (Clontech) was used to express EGFP.
Generation of Transgenic Mice Expressing Wild Type Human BRI2 (wtBriPP)-BRI2 cDNA (see above) was introduced into the blunt-ended SalI cloning site of the cosmid-based Syrian Hamster prion protein expression vector (provided by S. Prusiner, San Francisco). After removal of vector sequences by NotI digestion, microinjections of the purified construct into C57Bl/6 pronuclei yielded several putative founders C57Bl/6N-Tg(SHaPrP-BRI2). Two founders were further bred with C57Bl/6J mice to produce stable transgenic lines. All mice included in the analyses are of generation F2 or higher.
Transgenic Mice Used for Cross-breeding-The APPPS1-21 mice used for cross-breeding with wtBriPP transgenic mice have been described previously (25). APPPS1-21 mice have been generated and maintained on a pure C57Bl/6J background. All mice analyzed were hemizygous for the transgene(s) of interest. All animal experiments were performed in accordance with the current German animal welfare law and licensed by the local veterinary authority (Regierungspraesidium Tuebingen).
Histology and Immunohistochemistry-Brains were removed upon sacrifice and immersion-fixed in 4% paraformaldehyde. Immunohistochemistry was done on 25-m thick coronally cut cryoprotected free-floating frozen sections, using standard immunoperoxidase procedures with Elite ABC kits (Vector Laboratories) with Vector SG (Vector Laboratories, Burlingame, CA). The CN3 polyclonal antibody to A␤ was used as described previously (26).
Stereology and Quantification of Pathology-Stereological analysis was performed using a microscope equipped with a motorized x-y-z stage coupled to a video microscopy system (Systems Planning and Analysis, Inc., Alexandria, VA). Neocortical brain regions were defined using a standard mouse brain atlas (51). Quantification was done on the left hemisphere. Analysis of A␤ amyloid load was done on a series of coronally cut 25-m free-floating sections (every 24th section for the neocortex). Thus, all analyses included 8 -10 sections per animal. The amyloid load (percentage) was determined by calculating the areal fraction occupied by immunoreactive A␤, in two-dimensional sectors at a single focal plane at 20ϫ/0.45 numerical aperture.
Cell Culture, Transfection, and Treatment-HEK293 cells (ATCC), HEK293 cells stably expressing wild type ␤APP695 cDNA (HEK293-APPwt cells) (23), and HeLa cells (ATCC) were cultured in DMEM 4.5 g/liter glucose with L-glutamine (Lonza) supplemented with 10% fetal calf serum (Invitrogen) and 1% penicillin/streptomycin (Invitrogen). Transient transfection of cells was carried out using FuGENE HD (Roche Diagnostics) or Lipofectamine LTX with Plus Reagent (Invitrogen) according to the manufacturer's instructions. Cells transfected with pcDNA3.1/Zeo(ϩ) vector served as a control. One day after transfection, cells were supplemented with fresh medium and incubated for 24 h to obtain conditioned cell media. 10 mg/ml insulin solution (human), cell culture tested (Sigma), was added to the cell culture medium at 10 M concentration during the incubation period as indicated. After gathering conditioned cell media, the relative number of cells was measured using AlamarBlue cell viability reagent and fluorescent detection according to the manufacturer's protocol (Invitrogen), to adjust for differences in cell numbers.
Antibodies and Western Blot Analysis-The following antibodies were used: monoclonal antibody 6E10 specific to human A␤ (crude ascites; Covance); anti-APP C-terminal rabbit polyclonal antibody A8717 (Sigma); monoclonal antibody 8G4 to GAPDH (HyTest Ltd.); chicken polyclonal antibody to ITM2B raised against amino acids 1-60 (Abcam); rabbit polyclonal antibody to insulin-degrading enzyme (IDE) (ab25970, Abcam), used for Western blot detection of IDE; rabbit polyclonal antibody IDE-1 (27), used for detection of IDE in cerebral spinal fluid; monoclonal anti-IDE antibody (ab25733, Abcam), used for immunoprecipitation of IDE; 56C6 anti-neprilysin monoclonal antibody (Novocastra); monoclonal antibody NT1 raised against the N-terminal region of PS1 (28); and HRPconjugated secondary antibodies (Santa Cruz Biotechnology). Synthetic CTF-50, a 50-amino acid peptide resulting from the ␥-secretase cleavage of the C terminus of APP at Leu 720 -Val 721 (Calbiochem) was used as a molecular weight control for AICD. For detection of proteins from mouse brains, brains were freshfrozen, homogenized at 10% (w/v) in homogenization buffer (0.32 M sucrose in PBS, including complete protease inhibitor tablets (Roche Diagnostics)), aliquoted, and stored at Ϫ80°C until further use. For detection of proteins from cell lysates, cells were lysed in RIPA buffer supplemented with protease inhibitors (10 mM Tris, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 1ϫ complete inhibitor mixture (Roche Diagnostics), 5 mM EDTA, 2 mM 1,10-phenanthroline (Sigma)) followed by centrifugation at 16,000 ϫ g for 15 min. For purification of cellular membranes, cells were incubated in hypotonic buffer (10 mM Tris, pH 7.4, 1 mM EDTA, 1 mM EGTA, 1ϫ complete inhibitor mixture, 2 mM 1,10-phenanthroline) for 10 min on ice. Cells were homogenized, and nuclei and debris were removed by centrifugation at 1500 ϫ g for 10 min. The supernatant was then centrifuged at 100,000 ϫ g for 1 h. The membrane pellet was resuspended in RIPA buffer. Protein concentrations were determined using BCA protein assay reagent (Thermo Scientific), and equal amounts of protein were analyzed by SDS-PAGE. Conditioned cell media were centrifuged at 900 ϫ g for 5 min at 4°C to eliminate cells, and supernatants were supplemented with 1ϫ complete protease inhibitor mixture (Roche Diagnostics) and stored at Ϫ80°C. APP, IDE, neprilysin, wild type, and mutated forms of BRI2 (ITM2B) were analyzed by 10 or 12% Tris glycine SDS-PAGE. To detect total A␤, equal amounts of conditioned cell media were analyzed on 12% NuPAGE gels using MES running buffer (Invitrogen). For analysis of APP C-terminal fragments, separation of A␤40/A␤42 and Western blot detection of proteins was carried out as described previously (29). Representative blots from at least three independent experiments are shown. Densitometry was performed with ImageJ software to quantify the intensity of the individual bands.
Immunodepletion of IDE from Conditioned Cell Media-A mouse monoclonal antibody against IDE (ab25733, Abcam) was coupled to M-280 sheep anti-mouse IgG Dynabeads (Invitrogen). Conditioned cell media from BRI2-transfected HEK293 cells were incubated with IDE-coupled Dynabeads or uncoupled Dynabeads as a control. Immunoprecipitations were carried out according to the manufacturer's instructions. Subsequently IDE in cell media and precipitated IDE were analyzed by Western blot. Media were transferred to HEK293-APPwt cells for 24 h, and secreted A␤ was analyzed by Western blot and immunoassay.
Immunoassay to Measure Human A␤ in Conditioned Cell Media-Conditioned cell media were analyzed for A␤38, A␤40, and A␤42 levels where indicated by an electrochemiluminescent-based sandwich immunoassay using MS6000 human (6E10) A␤-3-Plex kit and the Sector Imager 6000 (Meso Scale Discovery) according to the manufacturer's instructions. Each sample was measured in duplicate, and the mean was taken. Values were adjusted for the relative number of viable cells measured by AlamarBlue cell viability assay. The concentration obtained for controls was set to 100%, and % A␤ relative to controls was calculated. To measure human A␤ in wtBriPP/APPPS1 mice, frozen brain hemispheres were homogenized in 50 mM Tris, pH 8, 150 mM NaCl, 4 mM EDTA containing protease inhibitors complete (Roche Diagnostics) and 2 mM 1,10-phenanthroline (Sigma). A␤ was extracted with diethanolamine (DEA) as described previously (21), and A␤ levels were measured by immunoassay as described above.
Mass Spectrometry Analysis-Molecular masses of intact A␤(1-40) peptides and the proteolytic products upon degradation of A␤ by recombinant or cell-derived IDE were determined using the matrix-assisted laser desorption ionization-time of flight mass spectrometer (MALDI-TOF MS, Biflex III, Bruker Daltonics GmbH, Germany). Snap-frozen samples from A␤ degradation assays were thawed and desalted using C18 Zip-Tips (Millipore) following the manufacturer's instructions. Samples were mixed with ␣-cyano-4-hydroxycinnamic acid (Sigma) spotted onto a ground steel MALDI target plate (Bruker Daltonics GmbH, Germany). Spectra were recorded in the linear mode at a laser frequency of 20 Hz within a mass range from 1,000 to 6,000 Da. Each spectrum is the result from an average outcome of at least 300 laser shots collected in 30 shot steps. FlexAnalysis 1.0 software (Bruker Daltonics GmbH, Germany) was used for visual estimation of the mass spectra performing smoothing and base-line subtraction.
Statistical Analysis-Results were analyzed by analysis of variance followed by Bonferroni's multiple comparison test to compare multiple groups or unpaired t test to compare two groups using GraphPad Prism 5 statistical software. The level of statistical significance was set at p Ͻ 0.05, two-tailed. In all graphs included, errors bars indicate one S.E.

BRI2 Reduces Plaque Load in the APPPS1 Mouse Model-We
have recently established and described a mouse model for familial Danish dementia overexpressing ADanPP, a mutant form of BRI2. When these mice were crossed to the APPPS1 mouse model of ␤-amyloid deposition (25), we found a significant reduction in plaque load in the double transgenic mice expressing ADanPP as compared with control APPPS1 single tg littermates (21). These results suggest that ADanPP expression may influence A␤ generation or deposition. A potential influence of the wild type BRI2 protein on APP processing has been described by other groups (11,13,14). To compare the effect of BRI2 and ADanPP in the same AD mouse model, we crossed transgenic mice expressing the wild type BRI2 protein (wtBriPP mice) to APPPS1 mice. We found that BRI2 leads to a 53.3% reduction in A␤ plaque load (Fig. 1, A-C), similar to the 68% reduction previously found with ADanPP (21). Western blot analyses of whole brain homogenates confirmed a decrease in total A␤ in double tg mice compared with single tg APPPS1 mice, whereas levels of APP and the APP C-terminal fragment C99 were not significantly changed (Fig. 1, D and E). In a next step we analyzed if BRI2 affects the metabolism of APP and A␤ in 1.4-month-old APPPS1 mice prior to amyloid deposition. Although levels of APP and C99 were similar in single and double tg predepositing mice (Fig. 1, F and G), the analysis of A␤ by immunoassay revealed decreased levels of A␤40, A␤42, and A␤38 in the BRI2-expressing double tg APPPS1 predepositing mice (Fig. 1I). In double tg compared with single tg APPPS1 mice, A␤40 was reduced by 41.5 Ϯ 16.8% (p ϭ 0.0298); A␤42 was reduced by 62.5 Ϯ 23.3% (p ϭ 0.0280), and A␤38 was reduced by 42.8 Ϯ 17.0% (p ϭ 0.0273). No changes in ␣or ␤-cleaved APP C-terminal fragments (CTFs) or AICD were detected by Western blot with an APP C-terminal specific antibody (Fig. 1H).
Taken together, these results suggest an influence of BRI2 on A␤ generation or turnover that occurs prior to amyloid deposition. Because levels of C99, C83, and AICD remained unchanged, and levels of all A␤ species were reduced, we suspected that rather than having an effect on APP processing by secretases, BRI2 reduced A␤ levels post-processing, potentially by enhanced A␤ clearance. To further investigate the molecular mechanism of the influence of BRI2 and ADanPP on APP, and A␤ metabolism in particular, we conducted cell culture experiments.
BRI2 and Its Mutant Form ADanPP Reduce Levels of Secreted A␤ and Increase Levels of C99 in HEK293-APPwt Cells-HEK293 cells stably expressing APP695 (HEK293-APPwt cells) were transfected with constructs to co-express either BRI2 or ADanPP. We found that both BRI2 and ADanPP clearly reduced levels of secreted A␤ in transfected cells, whereas levels of APP remained unchanged ( Fig. 2A). Quantification of A␤ peptides from conditioned cell media by immunoassay revealed a decrease in both A␤40 and A␤42. BRI2 and ADanPP significantly reduced A␤40 to levels 46.0 and 52.2% that of controls and A␤42 to levels 27.1 and 31.0% that of controls, respectively (Fig. 2B). No significant differences were found between the two constructs in their effect on A␤. To test the influence of both proteins on APP processing, C-terminal fragments of APP were analyzed. We found a strong increase in the ␤-cleavagederived C99 fragment with both constructs, in line with previously published results for BRI2 in cell culture (11,13). No major changes in levels of the ␣-secretase cleavage product C83 nor in levels of the ␥-secretase cleavage product AICD were observed ( Fig. 2C and supplemental Fig. 1). Therefore, the decrease in secreted A␤ does not seem to result from an inhibition of ␥-secretase cleavage of the C99 fragment. Still BRI2 might modulate ␥-secretase activity specifically, such that A␤ and AICD generation are affected differently. As an example, certain nonsteroidal anti-inflammatory drugs selectively lower A␤42 and increase A␤38 levels without inhibition of AICD generation (31,32). To investigate this possibility, we specifically analyzed levels of all three of the major secreted A␤ forms by immunoassay in conditioned cell media of BRI2 and control transfected cells. BRI2 leads to a decrease of all three A␤ forms. Levels of A␤38 in BRI2-transfected cells dropped below the detection limit, and no selective increase in A␤38 was observed (Fig. 2D). Taken together, these results do not point to an inhibition or modulation of ␥-secretase cleavage of APP by BRI2.
Effects of BRI2 on A␤ and C99 Are Independent of the BRI Peptide-In investigating plaque load reduction by BRI2, it has been proposed that the shed 23-aa BRI peptide may inhibit extracellular A␤ aggregation and deposition (12). We therefore asked whether the reduction in A␤ levels observed in cells was due to the involvement of the BRI peptide. BRI2 was cloned and FIGURE 1. BRI2 reduces amyloid plaque deposition in the APPPS1 mouse model of AD. A and B, A␤ immunostaining of 4-month-old wtBriPP ϩ /AP-PPS1 ϩ double tg mice compared with single tg wtBriPP Ϫ /APPPS1 ϩ littermates reveals a decrease in A␤ plaque deposition. Higher magnification is shown in the right panels. C, stereological quantification of neocortical A␤ load reveals a remarkable decrease in A␤ deposition in the double tg mice (**, p Ͻ 0.001; all females; n ϭ 11 single tg, n ϭ 7 double tg). D, Western blotting of APP, C99, and A␤ in wtBriPP Ϫ /APPPS ϩ and wtBriPP ϩ /APPPS ϩ mice reveals no change in levels of APP and C99 but a decrease in A␤ in double transgenic mice. Shown are three mice for each genotype. GAPDH serves as a loading control. E, quantification of C99 levels by densitometric analysis of band intensities normalized to GAPDH (p ϭ 0.683; n ϭ 5 single tg, n ϭ 5 double tg). Scale bars, 500 and 100 m. F-I, analysis of APP, C99, and A␤ levels in 1.4month-old pre-depositing wtBriPP ϩ /APPPS1 ϩ double tg mice compared with single tg wtBriPP Ϫ /APPPS1 ϩ littermates. F, Western blot analysis with antibody 6E10 reveals no differences in APP or C99 levels. Shown are three mice for each genotype, all females. GAPDH serves as a loading control. G, quantification of C99 levels by densitometric analysis of band intensities normalized to GAPDH and APP (p ϭ 0.2201; n ϭ 5 single tg, n ϭ 5 double tg, all females). H, Western blot analysis with antibody A8717 reveals no differences in ␣or ␤-cleaved APP C-terminal fragments (APP-CTFs) (upper panel) or AICD (lower panel, stronger exposure of the same blot). The two bands represent the nonphosphorylated and phosphorylated form of AICD (AICD/p-AICD). I, immunoassay of DEA extracted brain homogenates shows a significant decrease in A␤40, A␤42, and A␤38 comparing double tg with single tg littermates (*, p Ͻ 0.05; n ϭ 9 wtBriPP ϩ /APPPS1 ϩ ; n ϭ 5 wtBriPP Ϫ /APPPS1 ϩ , all females). OCTOBER 28, 2011 • VOLUME 286 • NUMBER 43 modified, such that a shorter BRI2 protein lacking the peptide sequence (aa 244 -266) was expressed. In transient transfection experiments, this BRI2 variant, BRI2⌬, led to a similar decrease in secreted A␤ and increase in C99 as seen for the expression of full-length BRI2 and ADanPP (Fig. 3). These results suggest an influence of BRI2 on A␤ levels, which is independent of its secreted peptide.

BRI2 Promotes A␤ Degradation by IDE
BRI2 Increases Levels of Secreted IDE-The observed A␤ reduction did not seem to result from inhibition or modulation of ␥-secretase cleavage by BRI2 or sequestration of soluble A␤ by the BRI peptide. The balance between anabolism and catabolism of A␤ determines its steady-state levels, such that a reduction in A␤ may also be caused by enhanced A␤ degradation. Two major A␤-degrading enzymes in the brain are neprilysin (33) and IDE (34). Thus, we next analyzed the expression of neprilysin and IDE in BRI2-transfected cells to test for a possible involvement in increased A␤ degradation. Levels of neprilysin were not significantly changed in the membrane fraction of HEK293-APPwt cells transfected with BRI2 compared with controls (Fig. 4A). IDE can be secreted from cells, where it can degrade A␤ (35) and also stay in the cytoplasm, where it is involved in the degradation of AICD (34). To discriminate between these different pools, we analyzed IDE levels in cell supernatants and cell lysates of cells transfected with the different BRI2 constructs. Expression of BRI2, ADanPP, and BRI2⌬ led to a significant increase in secreted IDE in conditioned cell media by 2.6-, 3.2-, and 2.9-fold, respectively, compared with controls. Levels of intracellular IDE were not sig- HEK293-APPwt cells were transfected with BRI2 or BRI2⌬, a construct that expresses BRI2 lacking the C-terminal peptide (aa 244 -266). Both BRI2 and BRI2⌬ lead to a decrease in secreted A␤ compared with controls (co). Western blot detection with the antibody A8717 shows an increase in C99 but no change in AICD or APP levels. Expression of the BRI2 proteins was confirmed by Western blot using the ITM2B antibody. Detection of GAPDH serves as a loading control. or ADanPP expression plasmids or pcDNA3.1 vector as a control (co). A, secreted A␤ was prominently decreased by BRI2 and ADanPP, whereas APP levels remained unchanged. Expression of BRI2 constructs was confirmed by Western blot with the ITM2B antibody, and detection of GAPDH serves as a loading control. B, quantification of secreted A␤40 and A␤42 by immunoassay from four independent experiments reveals a significant reduction in BRI2-and ADanPP-transfected cells relative to controls (***, p Ͻ 0.0001; BRI2 versus ADanPP p Ͼ 0.05). C, C-terminal fragments of APP were analyzed in cell lysates. An increase in levels of C99 fragments (middle panel, short exposure) but no change in levels of AICD (lower panel, longer exposure of the same blot) are observed. APP levels remain unchanged. D, HEK293-APPwt cells were transfected with BRI2 or pcDNA3.1 vector as a control (co). A␤40, A␤42, and A␤38 from conditioned cell media were analyzed by immunoassay from three independent experiments. All three A␤ forms are decreased in cell media of BRI2 transfected cells. Secreted A␤38 from BRI2 transfected cells is below the detection limit.
nificantly affected by expression of these constructs (Fig. 4, B and C).
As an additional control, for overexpression of an unrelated transmembrane protein, we transfected cells with the presenilin 1 mutant PS1L166P. This mutant associated with familial AD leads to increased A␤42 generation (36). In contrast to BRI2, the overexpression of PS1L166P did not lead to increased IDE secretion (supplemental Fig. 2, A and B) or a decrease in A␤ (supplemental Fig. 2C) showing that this effect is specific for BRI2 proteins. Expression and functionality of PS1L166P were confirmed by the expected effect of increased A␤42 generation (supplemental Fig. 2C). Cell viability was not affected by overexpression of BRI2, ruling out a possible increase in levels of IDE due to cell death (supplemental Fig. 2D). As a second control, overexpression of an unrelated protein, EGFP, did not influence levels of secreted IDE or A␤ (supplemental Fig. 2, E  and F). Taken together these results suggest that the BRI2-me-diated reduction in A␤ may be due to increased degradation by IDE.
Insulin Treatment Inhibits the BRI2-mediated A␤ Decrease-To test if the BRI2-mediated A␤ decrease could indeed be caused by increased levels of IDE, we analyzed if the effect could be blocked by the addition of insulin. Because insulin is a substrate with higher affinity for IDE than A␤, its addition in excess inhibits A␤ degradation by IDE (35). To this end we transfected HEK293-APPwt cells with BRI2 or vector control and compared A␤ levels in conditioned cell media with or without the addition of insulin during the incubation period. We found that the presence of insulin in the culture medium of BRI2-transfected cells significantly blocked the BRI2-mediated reduction in A␤, as shown by Western blot and immunoassay measurements of A␤40 and A␤42 (Fig. 5, A and B). BRI2 transfection significantly reduced secreted A␤40 and A␤42 to levels 11.3 and 11.2% that of the control (p Ͻ 0.0001, n ϭ 5). Insulin significantly blocked this reduction (p Ͻ 0.0001, n ϭ 5), such that only a minor reduction to 80.6 and 66.7%, A␤40 and A␤42, was observed (p ϭ 0.0019/p ϭ 0.0015). Thus, inhibition of A␤ degradation by IDE repressed in large part the A␤-lowering effect of BRI2 expression. The addition of insulin did not change levels of APP or APP C-terminal fragments (Fig. 5, A and C). Interestingly, the BRI2-mediated increase in C99 was not blocked by insulin (Fig. 5C), indicating that the A␤ decrease and C99 increase are not linked to each other.

Transfer of Conditioned Cell Medium from BRI2-transfected HEK293 Cells to Untransfected HEK293-APPwt Cells Causes a
Decrease in Secreted A␤-The above results strongly suggested a major role for secreted IDE in the A␤ reduction by BRI2. The secreted form of IDE is contained in the cell medium. Thus, we asked if the transfer of conditioned cell medium from BRI2transfected cells would have a similar effect on HEK293-APPwt cells as overexpression of BRI2 in these cells (for an overview of the experiment see Fig. 6A). To do so, we first transfected HEK293 cells with BRI2 or control vector and harvested conditioned cell media 48 h post-transfection. BRI2 expression in these cells and an increase in secreted IDE in the conditioned cell media were confirmed by Western blot (Fig. 6B). Equal amounts of conditioned cell medium from control or BRI2transfected HEK293 cells were then transferred to HEK293-APPwt cells and incubated for 24 h to allow A␤ secretion. In addition, insulin was added to one set of cells to inhibit A␤-degradation by IDE. Secreted A␤ was clearly reduced in the BRI2 media-treated cells compared with the controls (Fig. 6, C and  D). Quantification by immunoassay revealed a significant reduction in secreted A␤40 and A␤42 to 55.9 and 63.0% that of control levels (p Ͻ 0.0001, n ϭ 4). Thus, the transfer of conditioned cell medium of BRI2-transfected cells (BRI2-CM) leads to a similar effect as BRI2 overexpression. In addition, when insulin was added to the conditioned cell medium during incubation with the HEK293-APPwt cells, its A␤ reducing effect was completely blocked (p Ͻ 0.0001, n ϭ 4) (Fig. 6, C and D). APP expression in HEK293-APPwt cells remained unchanged by the addition of BRI2-CM or insulin (Fig. 6H). In a second set of experiments, we immunodepleted IDE from the conditioned cell medium prior to its addition to HEK293-APPwt cells. Immunoprecipitation with an IDE-specific antibody effectively FIGURE 4. BRI2 increases levels of secreted IDE. HEK293-APPwt cells were transfected with BRI2, ADanPP, or BRI2⌬ expression plasmids or empty vector as a control (co). A, levels of neprilysin in the membrane fraction of cells remained unchanged by BRI2. APP in the membrane fraction is shown as a loading control. B, secreted IDE was increased in the conditioned cell media from all cells transfected with one of the BRI2 constructs. In contrast, intracellular IDE detected in cell lysates remained unchanged by BRI2 expression. C, quantification of secreted IDE by densitometric analysis of band intensities. The 3-fold increase in secreted IDE relative to control (co) from three independent experiments is shown (***, p Ͻ 0.0001 relative to control; BRI2 versus ADanPP, BRI2 versus BRI2⌬, ADanPP versus BRI2⌬, p Ͼ 0.05). OCTOBER 28, 2011 • VOLUME 286 • NUMBER 43 removed secreted IDE from the conditioned cell medium of BRI2-transfected cells (BRI2-CM) compared with the control immunoprecipitation (Fig. 6E). Although BRI2-CM following control immunoprecipitation reduced secreted total A␤ to the same extent as untreated BRI2-CM, immunodepletion of IDE significantly prevented the BRI2-CM-mediated A␤ decrease (Fig. 6, F and G).

BRI2 Promotes A␤ Degradation by IDE
Taken together, these results suggest that a secreted factor in the medium of BRI2-transfected cells confers the BRI2-mediated A␤ decrease and that this soluble factor is IDE, because the effect can be blocked by insulin and, more specifically, by immunodepletion of IDE. Next, we were interested if the transfer of conditioned cell medium would also lead to an increase in levels of the C99 fragment in the recipient cells. As shown in Fig. 6H, C99 levels were not changed in these cells, although A␤ was reduced. Thus, as seen in the experiment above, the observed A␤ decrease and C99 increase are independent effects.
BRI2-mediated A␤ Decrease by IDE in HeLa Cells-Next, we used HeLa cells to confirm our results in a different cell line. HeLa cells had been used previously to study the effects of BRI2 on APP processing (13). We co-transfected APP695 and BRI2 into HeLa cells and analyzed levels of APP cleavage products and IDE. Again, BRI2 led to a reduction in secreted A␤ (supplemental Fig. 3A) paralleled by an increase in secreted IDE, with no changes in cell viability (supplemental Fig. 3, C and D). Similar to that in HEK293-APPwt cells, an increase in C99 levels was observed, while AICD levels remained the same (supplemental Fig. 3, B and H). To test if the increased levels of secreted IDE also caused the A␤ decrease in this cell line, we included the addition of insulin as variable in the experiment. As in the HEK293-APPwt cells, in HeLa cells almost all of the A␤-lowering effect was blocked when A␤ degradation by IDE was inhibited by the addition of insulin (supplemental Fig. 3, E and F). BRI2 transfection significantly reduced secreted A␤40 and A␤42 to levels 38.1 and 12.7% that of the control, as measured by immunoassay (p ϭ 0.0021/p ϭ 0.0019, n ϭ 3). Insulin significantly blocked this reduction (p ϭ 0.0002/p ϭ 0.0003, n ϭ 3), as only a minor reduction to 79.4 and 82.5% A␤40 and A␤42 compared with control levels was observed (p ϭ 0.0029/p ϭ 0.0301) (supplemental Fig. 3, E and F). As a control, levels of APP and APP C-terminal fragments were examined and shown to remain the same following the addition of insulin (supplemental Fig. 3, G and H). These results confirm the major influence of IDE in the BRI2-mediated A␤ decrease.
Enhanced IDE-mediated Degradation of A␤  in Cell Media of BRI2-transfected HEK293-APPwt Cells-To specifically test for IDE-mediated degradation of A␤, we compared the A␤ proteolysis pattern obtained after its degradation by recombinant IDE and by conditioned cell media from control and BRI2-transfected HEK293-APPwt cells. Following incubation with synthetic A␤(1-40), very similar MALDI-TOF spectra were obtained for conditioned cell media from BRI2-transfected cells (Fig. 7, E and F) and recombinant IDE (Fig. 7, A and  B). Both contained peaks corresponding to individual A␤ fragments generated after its degradation by IDE (Table 1) (30). Incubation with cell media from control-transfected cells did not show an increase in IDE-specific A␤-degradation products (Fig. 7, C and D). This evidence strongly supports that BRI2 mediates increased A␤ degradation through secreted IDE.
Increased IDE Levels in Cerebral Spinal Fluid of wtBriPP Mice-Next, we wanted to test if the BRI2-mediated increase in secreted IDE may also be involved in the observed A␤ decrease and plaque load reduction in the transgenic mice. As expected, no significant differences in IDE levels were detected in the FIGURE 5. Insulin treatment inhibits BRI2-mediated A␤ decrease. BRI2-or vector control (co)-transfected HEK293-APPwt cells were supplemented with fresh medium 1 day after transfection and incubated for 24 h to generate conditioned cell medium. Where indicated, 10 M insulin was added. A and B, without addition of insulin BRI2 expression prominently decreases levels of secreted A␤. Inhibition of IDE activity with insulin prevents this A␤ decrease almost completely. A, levels of secreted A␤ are shown in two exposures of the same blot. Secreted IDE is increased with BRI2 expression and remains unchanged by the addition of insulin. In addition, APP expression remains unaltered by insulin. GAPDH serves as a loading control. B, A␤40 and A␤42 from conditioned cell media were analyzed by immunoassay from five independent experiments. Graphs show % remaining A␤ relative to controls in BRI2-transfected cells and BRI2-transfected cells treated with insulin. Insulin significantly blocked the BRI2-mediated reduction in A␤40 and A␤42 (***, p Ͻ 0.0001). C, addition of insulin does not alter levels of APP C-terminal fragments. total brain homogenates of wtBriPP-tg mice compared with controls (Fig. 8, A and B). Here, both the secreted and the intracellular pools of IDE are mixed together such that changes in only the secreted form may be masked by the relatively overwhelming level of intracellular IDE. To specifically analyze secreted IDE, we collected cerebral spinal fluid from wtBriPP-tg and non-tg littermates. Western blot detection of IDE revealed significantly higher levels of IDE in the cerebral spinal fluid of tg mice compared with the non-tg littermate controls (Fig. 8, C and D).

DISCUSSION
The British precursor protein BRI2 is argued to influence APP processing and A␤ deposition in cells and in vivo by FIGURE 6. Transfer of conditioned cell medium from BRI2-transfected HEK293 cells to untransfected HEK293-APPwt cells causes a decrease in secreted A␤. A, schematic overview of the experiment. HEK293 cells were transfected with BRI2 or empty control vector (co). Conditioned cell medium from these cells was transferred to untransfected HEK293-APPwt cells with or without addition of 10 M insulin as indicated and incubated for 24 h. B, parallel to BRI2 expression, increased levels of IDE were detected in conditioned cell media from BRI2-transfected HEK293 cells compared with controls. C and D, HEK293-APPwt cells treated with conditioned cell medium from BRI2-transfected cells (BRI2-CM) show decreased levels of secreted A␤ compared with cells treated with control medium (co-CM). This effect is inhibited by the addition of insulin (ϩIns). C, detection of secreted A␤ by Western blot. D, analysis of secreted A␤40 and A␤42 by immunoassay from four independent experiments. Graphs show % remaining A␤ relative to control for cells incubated with BRI2-CM compared with BRI2-CM-incubated cells additionally treated with insulin. BRI2-CM significantly reduced secreted A␤40 and A␤42 relative to controls (p Ͻ 0.0001), while insulin blocked this reduction completely (***, p Ͻ 0.0001). E, IDE in conditioned cell medium from BRI2 transfected HEK293 cells was immunodepleted using an IDE-specific antibody (IDE-IP) or only beads as a control (co-IP). IDE levels in immunodepleted or control cell media as well as the precipitated IDE from control and IDE-specific immunoprecipitation is shown (precip.). F and G, conditioned cell media were then transferred to HEK-APPwt cells, incubated for 24 h, and secreted A␤ analyzed by Western blot. Immunodepletion of IDE reverts the BRI2-mediated decrease in A␤. G, quantification of total A␤ by immunoassay from three independent experiments shows that immunodepletion of IDE from the cell media of BRI2 transfected cells blocks its effect on A␤ decrease (IDE-IP-CM versus co-IP-CM, **, p ϭ 0.0098; BRI2-CM versus co-CM, ***, p Ͻ 0.0001). H, treatment with BRI2-CM did not lead to an increase in levels of the C99 fragment in HEK293-APPwt cells. The addition of insulin does not alter levels of APP, APP C-terminal fragments, or AICD. GAPDH is shown as a loading control.
varying mechanisms (11)(12)(13)(14)(15). Although the previous studies have suggested that mutations in the BRI2 protein may relate to a loss of BRI2 function and its effect on APP (37,38), we have recently described that the mutant form of BRI2, ADanPP, decreases A␤ plaque load in transgenic mouse models of AD (21). In this study, we show that wild type BRI2 causes a similar plaque load reduction in the APPPS1 mouse model. We also confirmed previous results that overexpression of BRI2 decreases levels of cell-secreted A␤ (11,13) and showed in addition that the mutated form ADanPP reduces A␤ levels to a similar extent in HEK293-APPwt cells. Not only A␤ production but also A␤ degradation plays an important role in the regulation of A␤ levels (5,6). Because we found no evidence for changes in APP processing in mice, we investigated the potential role of BRI2 and ADanPP in regulating A␤ degradation. Although neprilysin levels were not changed in BRI2-transfected cells, we found that the reduction in A␤ was strikingly paralleled by an increase in cell-secreted IDE. This effect was specific for BRI2 proteins because it was not found with the overexpression of PS1-L166P or EGFP used as controls. With additional experiments, we demonstrated that the observed increase in secreted IDE was responsible for most, if not all, of the A␤ decrease. By simply transferring conditioned cell medium of BRI2-transfected HEK293 cells, the A␤ reduction could be induced in nontransfected HEK293-APPwt cells to a similar degree as with BRI2 transfection. This shows that a factor secreted from BRI2-transfected cells, which is most probably IDE, must mediate the effect. This is strongly supported by the result that addition of insulin, which inhibits A␤ degradation by competing for IDE (35), completely abolished the A␤-lowering effect of both BRI2 transfection and the transfer of conditioned cell media from BRI2-transfected HEK293 cells. Furthermore, the more specific removal of IDE, by immunoprecipitation from the conditioned media, abolished the A␤-lowering effect on the recipient cells, ruling out a major contribution of other soluble factors from the conditioned cell media. In addition, incubation of synthetic A␤(1-40) with conditioned cell medium from BRI2 transfected cells showed enhanced A␤-derived degradation fragments very similar in their MALDI-TOF spectrum to those derived from incubation with recombinant IDE, strongly supporting that BRI2 mediates A␤ degradation by secreted IDE.
As an alternative mechanism, it has been proposed that BRI2 could lower A␤ levels through direct inhibition of ␥-secretase cleavage of the APP C99 fragment (11,13,14). In line with this hypothesis, we also found increased amounts of the C99 fragment in cells. However, in our experiments levels of the ␥-secretase cleavage product AICD remained unchanged. These results are contradictory to previously published results (13) and may be due to the direct detection of the AICD fragment in our experiments, as opposed to use of an indirect method of measurement (13). In our cell experiments, the main factor responsible for the decrease in A␤ was contained in conditioned media and could be inhibited by insulin. Our results indicate that the increase in C99 is independent of the A␤ reduction, because it was not observed in cells that received the conditioned cell medium from BRI2-transfected cells. In addition, we did not observe changes in C99 levels in vivo, in APPPS1 mice crossed with wtBriPP mice. This is in line with the described increase in A␤ in BRI knock-out mice without changes in APP C-terminal fragments C83 and C99 (14).
In mouse models, the 23 amino acid long BRI peptide, which is shed by furin-like cleavage, is proposed to inhibit A␤ aggregation and thus reduce plaque load in transgenic mice (12). With a shorter BRI2 protein lacking the peptide sequence (BRI2⌬244 -266) we still observed increased levels of secreted IDE and decreased A␤. Thus, this function of BRI2 seems to be independent of the BRI peptide sequence. This observation is in line with results from others reporting that aa 46 -106 of BRI2 were sufficient for the reduction of A␤40 levels in vitro (11).
IDE is a metalloprotease that cleaves insulin as well as other peptide substrates (39). The enzyme is primarily located in the cytosol, but a fraction of the enzyme is found in peroxisomes and on the plasma membrane. In addition, the enzyme is secreted by an unconventional pathway (40), can be found in the cell media of several cell lines, and is shown to degrade extracellular A␤ (27,35). Here, we show that BRI2 influences only the cell-secreted form of IDE, whereas intracellular levels of IDE remain relatively unchanged. To date, it is not known exactly how these different pools of IDE are regulated. A potential association with exosomes is discussed (30,41), and further work will be needed to investigate the involvement of BRI2 in these processes in more detail.
A␤ levels are elevated in the brains of IDE-deficient mice (34,42), and in mouse models of amyloid deposition, increased levels of IDE, through overexpression, reduce amyloid plaque load (43). Because 95% of IDE is contained in the cytosol and only a FIGURE 8. Increased IDE levels in cerebral spinal fluid of wtBriPP mice. A, IDE expression in total brain homogenates of 1.8 -2-month-old wtBriPP mice and non-tg littermates was analyzed. GAPDH serves as a loading control. B, densitometric analysis of IDE band intensities normalized to GAPDH reveals no significant differences between tg and non-tg mice. C, Western blot analysis of IDE levels in cerebral spinal fluid collected from mice analyzed in A. D, densitometric analysis of IDE band intensities reveals 4.5-fold higher levels IDE in the cerebral spinal fluid of tg mice compared with control mice (*, p ϭ 0.0196; all females, n ϭ 3 tg, n ϭ 3 non-tg).

TABLE 1 MALDI-TOF MS analysis of the cleavage products of A␤(1-40) peptide by recombinant insulin degrading enzyme
Expected proteolytic products and the molecular mass of the respective peptide fragments upon cleavage by IDE. The calculated mass (Da) and observed mass (Da) are indicated.  OCTOBER 28, 2011 • VOLUME 286 • NUMBER 43 small fraction is secreted (39), small changes within this pool prove to be difficult to detect in the mouse brain. We did not find significant differences in IDE levels in total brain homogenates of wtBriPP tg mice compared with control mice. To specifically analyze secreted IDE in the brain, we collected cerebral spinal fluid and found that IDE was significantly increased in the cerebral spinal fluid of wtBriPP tg mice compared with non-tg littermates. These results suggest that BRI2 can mediate an increase specifically in secreted IDE in both cells and mice. A second line of evidence that indicates that enhanced IDEmediated degradation may be responsible for the plaque load reduction in mice comes from the fact that levels of the major secreted A␤ forms, A␤40, A␤42, and A␤38, are equally reduced in wtBriPP/APPPS1 mice compared with single transgenic APPPS1 mice early, prior to amyloid deposition. The reduced plaque load in older mice may thus be caused by lower A␤ levels, rather than by an inhibition of A␤ aggregation. Taken together, our results suggest that BRI2-mediated changes in the secreted form of IDE in the brain are responsible for the observed decrease in amyloid deposition. It remains possible that in vivo other proposed mechanisms (11)(12)(13)(14) may work in concert with this previously undescribed function of BRI2.

BRI2 Promotes A␤ Degradation by IDE
At present, the physiological role of the BRI2 protein is not known. It is cleaved by several proteases, whereby extracellular parts are shed and an intracellular domain is generated (19), similar to proteins involved in signal transduction by regulated intramembrane proteolysis (44). Having these properties, BRI2 resembles receptor proteins involved in signal transduction. Like many receptors, BRI2 can form homodimers (45). Thus BRI2 signaling may regulate levels of extracellular IDE. Several pathways can regulate IDE levels in different experimental systems. The PI3K pathway (46), insulin receptor signaling (47), the transcription factor peroxisome proliferator-activated receptor -␥ (48,49), palmitic acid, docosahexaenoic acid (50), as well as up-regulation of exosome secretion dependent on protein isoprenylation (30) have been implicated. Future work will investigate the activation of these pathways by BRI2.
The regulation of A␤ levels is of great interest for potential therapies for Alzheimer disease. Apart from interference with A␤ generation, a promising alternative may be the enhancement of A␤ degradation by targeting A␤-degrading enzymes (6). Recently published results indicate the contribution of decreased central nervous system ␤-amyloid clearance in late onset AD (7). Thus, a better understanding of naturally occurring A␤ regulation may lead to new approaches to therapeutically address sporadic AD.