Phosphorylation of Amyloid-β Peptide at Serine 8 Attenuates Its Clearance via Insulin-degrading and Angiotensin-converting Enzymes*

Background: Amyloid-β peptide (Aβ) is degraded by different proteases. We recently demonstrated phosphorylation of Aβ. Results: Phosphorylation of Aβ decreases its clearance by microglial BV-2 cells and selectively inhibits the cleavage by insulin-degrading and angiotensin-converting enzymes. Conclusion: Phosphorylation at Ser-8 negatively regulates Aβ degradation. Significance: Phosphorylation could play a dual role in Aβ metabolism. It decreases the clearance by microglial cells and also promotes Aβ aggregation. Accumulation of amyloid-β peptides (Aβ) in the brain is a common pathological feature of Alzheimer disease (AD). Aggregates of Aβ are neurotoxic and appear to be critically involved in the neurodegeneration during AD pathogenesis. Accumulation of Aβ could be caused by increased production, as indicated by several mutations in the amyloid precursor protein or the γ-secretase components presenilin-1 and presenilin-2 that cause familial early-onset AD. However, recent data also indicate a decreased clearance rate of Aβ in AD brains. We recently demonstrated that Aβ undergoes phosphorylation by extracellular or cell surface-localized protein kinase A, leading to increased aggregation. Here, we provide evidence that phosphorylation of monomeric Aβ at Ser-8 also decreases its clearance by microglial cells. By using mass spectrometry, we demonstrate that phosphorylation at Ser-8 inhibited the proteolytic degradation of monomeric Aβ by the insulin-degrading enzyme, a major Aβ-degrading enzyme released from microglial cells. Phosphorylation also decreased the degradation of Aβ by the angiotensin-converting enzyme. In contrast, Aβ degradation by plasmin was largely unaffected by phosphorylation. Thus, phosphorylation of Aβ could play a dual role in Aβ metabolism. It decreases its proteolytic clearance and also promotes its aggregation. The inhibition of extracellular Aβ phosphorylation, stimulation of protease expression and/or their proteolytic activity could be explored to promote Aβ degradation in AD therapy or prevention.

Alzheimer disease (AD) 3 is characterized by the progressive deposition of amyloid-␤ peptides (A␤) in the brain (1,2). A␤ derives from proteolytic processing of the amyloid precursor protein involving sequential cleavages by enzymes called ␤and ␥-secretases (3,4). A critical role of A␤ in the pathogenesis of AD is strongly supported by several mutations within the genes encoding the amyloid precursor protein itself or the two presenilins that represent the proteolytically active components of the ␥-secretase complex. All of these mutations affect the production and/or aggregation of A␤ and cause early-onset forms of familial AD (5)(6)(7). Although early-onset familial AD appears to be commonly associated with an elevated production of aggregation-prone A␤ variants, it remains unclear whether increased A␤ generation also contributes to the much more common form of late-onset AD. Recent evidence rather indicated a decreased clearance rate of A␤ in AD compared with control brains (8 -10).
IDE is localized principally in the cytosol but is also released from cells and found in extracellular fluids and conditioned media of cultured cells (24,25). However, IDE lacks canonical signal sequences that target the enzyme to the conventional secretory pathway (26). Recent data demonstrated that IDE is secreted via a nonconventional pathway in association with exosomes (27,28). This nonconventional secretion of IDE is dependent on a hexapeptide amino acid motif in its C-terminal domain (29). The pathophysiological relevance of IDE is demonstrated by the deletion of IDE in mice that showed decreased A␤ degradation and increased cerebral A␤ accumulation (30 -32). Conversely, enhancement of IDE activity in neurons effectively reduced A␤ accumulation in AD mouse models (30,33). Recent data demonstrated that extracellular A␤ could undergo phosphorylation by secreted variants of protein kinase A and that the phosphorylation of A␤ at Ser-8 strongly promoted its aggregation into oligomeric and fibrillar assemblies (34).
Here, we sought to assess the effect of phosphorylation on the clearance of A␤ by microglial BV-2 cells. Our data demonstrate that phosphorylated A␤ (pA␤) has increased stability against microglial degradation compared with non-phosphorylated A␤ (npA␤). Interestingly, phosphorylation significantly decreases its proteolytic degradation by secreted IDE.
A␤ Degradation Assays with BV-2 Cells or Recombinant Enzymes-Synthetic npA␤ and pA␤ were solubilized in 10 mM NaOH at a concentration of 1 mg/ml (230 M), sonicated for 5 min, and stored at Ϫ80°C until used. Aliquots were thawed and diluted in the appropriate buffer.
For the cellular degradation assays, BV-2 cells were cultured in Dulbecco's modified Eagle's medium with GlutaMAX (Invitrogen) supplemented with 10% fetal calf serum and 1% penicillin/streptomycin (Invitrogen) in a 24-well plate to 70% confluence. Media were replaced with serum-free media and incubated with 1 M npA␤ or pA␤. Aliquots of the media were taken at the indicated time points and mixed with 4ϫ lithium dodecyl sulfate sample buffer (Invitrogen), incubated at 70°C for 10 min, and then separated by SDS-PAGE on 16% Tricine gels. A␤ was detected by Western blotting using primary monoclonal antibody 82E1 and HRP-conjugated rabbit antimouse secondary antibody.
For A␤ degradation with recombinant enzymes, synthetic npA␤ and pA␤ were diluted in assay buffer (50 mM Tris and 100 mM NaCl, pH 7.5) to a final concentration of 25 ng/l (A␤). Reactions were started by the addition of recombinant enzymes in assay buffer (IDE, 0.3 ng/l; NEP, 1 ng/l; ACE, 0.5 ng/l; and plasmin, 10 ng/l) and then allowed to proceed for the indicated time points at 37°C on a block heater at 650 rpm. Sample aliquots (10 l) were taken, snap-frozen in liquid nitrogen, and stored for further use at Ϫ80°C. Samples were then either analyzed by SDS-PAGE and Western blot analysis or by MALDI-TOF-MS.
Mass Spectrometry-The molecular masses of intact peptides and proteolytic products of npA␤ and pA␤ were determined using MALDI-TOF-MS. Samples were purified through reversed-phase ZipTip C 18 tips following the manufacturer's instructions and analyzed by a MALDI-TOF-MS system (BIFLEX III, Bruker Daltonics GmbH). Samples were mixed with ␣-cyano-4-hydroxycinnamic acid and spotted onto a ground-steel MALDI target plate (Bruker Daltonics GmbH). Spectra were recorded in the linear mode at a laser frequency of 20 Hz within a mass range of 1000 -6000 Da. Each spectrum is the result of an average outcome of at least 300 laser shots collected in 30-shot steps. FlexAnalysis 1.0 software (Bruker Daltonics GmbH) was used for visual estimation, smoothening and base-line substraction of the mass spectra.
siRNA-mediated Knockdown of IDE-BV-2 cells were transfected with 22.5 nM siRNA (target sequence GCCTGTTGTCA-GAACTCAA) using HiPerFect transfection reagent (Qiagen) according to the supplier's instructions. Knockdown of IDE was analyzed after 24 h in cell supernatants and cell lysates by Western immunoblotting. For A␤ degradation experiments, cells were then incubated for 12 h in serum-free medium. Synthetic npA␤ and pA␤ variants were added to conditioned media, and aliquots were taken at the indicated time points. A␤ was then detected by Western immunoblotting.

RESULTS
Recent data demonstrated phosphorylation of extracellular A␤ at Ser-8 by secreted or cell surface-localized variants of PKA ( Fig. 1A) (34,35). To test whether this phosphorylation affects the clearance of A␤ by microglia, mouse microglial BV-2 cells were incubated with synthetic pA␤ or npA␤, and stability was assessed by detection of the peptides after different time periods of incubation by Western blotting (Fig. 1B). Consistent with previous data (28,36), extracellular monomeric A␤ was efficiently cleared in the conditioned media of BV-2 cells. About 80% of npA␤ was cleared within the first hour of incubation. In contrast, only ϳ5% of pA␤ was cleared after the first hour. Even after 6 h, ϳ50% of pA␤ was detected in the cell supernatant, a time point at which only residual amounts (Ͻ10%) of npA␤ were left (Fig. 1C). The calculated half-life times of npA␤ and pA␤ in these experiments were 45 min and 6 h, respectively. These data demonstrate that phosphorylation of A␤ strongly decreases the clearance of extracellular A␤ by microglial cells.
As shown previously (24,28,36), microglial BV-2 cells efficiently degrade extracellular monomeric A␤ by secreted IDE. To specifically analyze the effect of A␤ phosphorylation on IDE-mediated cleavage, we next performed in vitro experiments using recombinant IDE. Synthetic pA␤ or npA␤ was incubated with IDE, and time-dependent cleavage was examined by SDS-PAGE and subsequent silver staining or by MALDI-TOF-MS (Fig. 2, A-C). Although the amount of npA␤ gradually decreased during the incubation, pA␤ appeared to be very stable under these conditions (Fig. 2, A and B). MALDI-TOF-MS analysis demonstrated that, at the start of incubation, only full-length A␤ variants were detected (Fig. 2C). Already after 5 and 15 min, an additional peak was detected in the samples with npA␤, demonstrating cleavage of A␤ by IDE. The m/z ratio of this peak corresponds to A␤(1-14), indicating cleavage between amino acids 14 and 15. This peak increases over time and was the predominant peak detected after 30 and 60 min of incubation. However, some additional peptides were detected that represent minor degradation products of less efficient cleavages between amino acids 13 Table 1). Importantly, very few (if any) cleavage products of pA␤ were detected even after 60 min of incubation with recombinant IDE. These results demonstrate that phosphorylation of A␤ strongly decreases its degradation by IDE. Because we observed clearance by microglial cells also of the pA␤ variant, although strongly decreased, we tested whether phosphorylation completely blocks IDE-mediated degradation. pA␤ and npA␤ were incubated with higher amounts of recombinant IDE, and degradation was analyzed by SDS-PAGE and silver staining or by MALDI-TOF-MS. Notably, under these conditions, we observed almost complete degradation of both A␤ variants (Fig. 3, A and B). The resulting pattern of degradation products of pA␤ and npA␤ was very similar, except for the increased masses due to the phosphoryl group ( Fig. 3B and Table 1). These data indicate that phosphorylation of A␤ does not completely block but rather decreases the efficiency of cleavage by IDE. In addition, the cleavage specificity of IDE is also not affected by phosphorylation of A␤. The detection of phosphorylated degradation products also indicates that IDE-mediated cleavage of A␤ does not require dephosphorylation.
In addition, we also carried out similar experiments with human cerebrospinal fluid. At similar concentrations of IDE, the degradation of A␤ was lower in cerebrospinal fluid samples compared with the IDE assay buffer. These data indicate that cerebrospinal fluid contains proteins or other factors that negatively affect the degradation of A␤. However, the inhibitory effect of phosphorylation on A␤ degradation was still evident (supplemental Fig. 1).
Recently, we demonstrated that IDE is the main protease contributing to A␤ degradation in BV-2 cells (28). Consistent with these results, siRNA-mediated knockdown of IDE strongly reduced the expression of cellular as well as secreted IDE in control cells. After knockdown, IDE levels were strongly decreased in the conditioned media and cell lysates by ϳ75% (Fig. 4, A and B). Importantly, siRNA-mediated knockdown of IDE in these cells decreased the degradation of A␤ also by ϳ75% (Fig. 4, C and D). Consistent with the previous results, the degradation of pA␤ was also lower compared with npA␤ upon siRNA-mediated knockdown of IDE. Together, these data strongly indicate that secreted IDE is the major protease in the degradation of extracellular A␤ in this cell system.
Because other proteases also can contribute to the degradation of A␤ in other cell systems or in vivo, we further tested the effect of phosphorylation on A␤ degradation by known A␤-cleaving enzymes, including ACE, NEP, and plasmin. Although NEP showed comparably low A␤-degrading activity under the experimental conditions and even at higher enzyme concentrations (supplemental Fig. 2), plasmin and ACE efficiently degraded the synthetic A␤ variants. The main cleavage products generated by plasmin had m/z ratios of ϳ3265 and ϳ1958, representing A␤(1-28), and A␤(1-16), respectively (Fig. 5, A-C, and Table 1). However, phosphorylation had little (if any) inhibitory effect on plasmin-mediated degradation of A␤, which was also indicated by the similar appearance of phosphorylated degradation products (Fig. 5C). In contrast, the degradation of pA␤ by ACE was strongly decreased compared with that of npA␤ (Fig. 6, A-C). Interestingly, cleavage of A␤ by ACE occurred directly N-or C-terminal of Ser-8, either between Asp-7 and Ser-8 or between Ser-8 and Gly-9, as indicted by the m/z ratios of C-terminal cleavage products of ϳ3451 (A␤ (8 -40)) and ϳ3370 (A␤(9 -40)).

DISCUSSION
This study revealed a novel mechanism in the regulation of A␤ metabolism. The phosphorylation at Ser-8 inhibits the degradation by IDE, a major A␤-degrading enzyme secreted by microglial BV-2 cells.
Increased concentrations of A␤ favor its aggregation and deposition in the form of ␤-amyloid plaques in the human brain. This is well supported by rare mutations in amyloid precursor protein and the presenilin proteins that cause autosomal dominant early-onset AD and commonly increase the production of A␤ and/or its aggregation (37). However, mechanisms that alter the metabolism of A␤ in the pathogenesis of the much more common late-onset form of AD are largely unclear. Interestingly, recent data indicated a decreased rate of A␤ clearance in the AD brain rather than increased production (8,9). The half-life time of A␤ is ϳ8 -12 h in human cerebrospinal fluid and 3-4 h in the interstitial fluid of mouse brain, indicating efficient clearance mechanisms that counteract the production of A␤ (9,10,38). The clearance of A␤ from the brain involves internalization via pinocytosis or receptor-mediated endocytosis/phagocytosis (13-15, 39) and subsequent degradation in the endosomal/lysosomal compartments (40), transcytosis and drainage via the blood-brain barrier to the vasculature (12,41), and proteolytic degradation of extracellular A␤ by cell surface-localized and secreted proteases (21,28,36).
Microglial cells could contribute to the clearance of A␤, as they are the main phagocytes in the brain and also express several proteases at the cell surface that could degrade extracellular and internalized A␤ (27). Recent evidence demonstrated that microglia secrete substantial amounts of IDE, thereby allowing efficient degradation of monomeric A␤ variants also at some distance from the microglial cell (19,42). IDE is a zinc metalloprotease with broad substrate specificity that is involved in the degradation of several peptides, including insulin, glucagon, transforming growth factor, and other peptide hormones (43). The different peptide substrates have little (if any) sequence homology, but many of the substrates share a propensity to form a ␤-sheet-rich conformation under certain conditions (e.g. A␤, insulin, glucagon, amylin, atrial natriuretic factor, and calcitonin) (44 -47). Accordingly, substrate selection of IDE might be determined mainly by the size and secondary and tertiary conformation of the peptides (46). Co-crystallization indicated that, after being engulfed into the catalytic center of IDE, A␤ could undergo conformational changes that might be important for the subsequent hydrolysis reaction (48). We recently showed that the phosphorylation of A␤ affects its conformation (34). Thus, the conformational change induced by phosphorylation might decrease the efficiency of hydrolysis by IDE. The introduction of a negatively charged phosphoryl group could also directly impair enzyme-substrate interaction. Similar effects of phosphorylation have been shown previously for caspase-mediated processing of

MALDI-TOF-MS analysis of cleavage products of npA␤ and pA␤
Shown are the various peptide fragments generated upon cleavage of npA␤ and pA␤ variants by different proteases and their calculated mass (Da) and observed mass (Da) by MALDI-TOF. 1-40* represents the double-ionized full-length peptide (m/2z instead of normal m/z). The indicated masses of the full-length peptide  and the double-ionized full-length peptide (1-40*) of both peptide variants are the mean values (n ϭ 10). The difference between the observed mass and the calculated mass is ϳ1-5 Da. The proteolytic products with a mass below 1000 kDa and C-terminal fragments of A␤ generated after the proteolytic cleavage were barely detectable even at spectrum recordings at low mass range (between 100 and 1000 kDa). ND, peaks that were not detected by MALDI-TOF-MS.   A and B, synthetic npA␤ and pA␤ were incubated with a higher concentration of recombinant IDE (0.9 ng/l) for 120 min at 37°C. Samples obtained before (0) and after (120) the incubation period were analyzed by SDS-PAGE and silver staining (A) and by MALDI-TOF-MS (B). Degradation products from both peptide variants are clearly evident after 120 min of incubation. Notably, the resulting pattern of degradation products of p〈␤ is very similar to that of npA␤ except for the increased masses due to the phosphoryl group.
several other AD-related proteins, including presenilin-1, presenilin-2, and Tau (49 -51). Thus, it would be interesting to co-crystallize IDE with pA␤ or npA␤ to further elucidate the structural basis of the inhibitory effect of phosphorylation on A␤ degradation. In addition to IDE, several other proteases involved in A␤ degradation that contribute to the regulation of A␤ levels in the human brain have been identified (52). In particular, NEP, ACE, and plasmin are considered to be physiologically and pathologically relevant in sporadic AD. NEP is a membrane-bound zinc metallopeptidase localized at the cell surface and in cytoplasmic vesicles preferentially hydrolyzing extracellular oligopeptides on the amino side of hydrophobic residues (53) and has been shown to be capable of degrading A␤ both in vivo (54,55) and in vitro (55)(56)(57). NEP is also reported to be expressed on neuronal pre-and postsynaptic membranes (52). However, only very little concentrations of NEP are found in extracellular fluids (58). In the current experimental paradigm, NEP showed low A␤-degrading activity even at higher enzyme concentrations. Furthermore, the expression of NEP is very low, and its activity is hardly detectable in microglial BV-2 cells (28). In addition, we observed a very strong correlation between IDE expression after RNAi-mediated knockdown and residual A␤-degrading activity by BV-2 cells (Fig. 4). Thus, a major contribution of NEP to the degradation of extracellular A␤ by this cell type is unlikely. However, it will be interesting to further investigate the effect of A␤ phosphorylation on NEP-dependent degradation.
Plasmin, a serine protease released after cleavage of its zymogen plasminogen, can also modulate the clearance of A␤ (59). In the nervous system, plasmin/plasminogen is expressed in neurons, whereas plasminogen activator is synthesized by neurons and microglial cells (60). The plasmin system is involved in many neural functions, such as neuronal plasticity, learning, and memory (61). The plasmin-dependent degradation of A␤ decreases its neurotoxicity in rat cortical cultures (62,63). Our mass spectrometric analysis revealed main cleavage sites between Lys-16 and Leu-17 and between Arg-28 and Gly-29. However, the phosphorylation of A␤ had a negligible (if any) inhibitory effect on plasmin-mediated degradation.
ACE, also known as dipeptidyl carboxypeptidase (EC 3.4.15.1), is a membrane-bound zinc metalloprotease (64) and is widely expressed in peripheral tissues and the brain. The involvement of ACE in the pathogenesis of AD was suggested by the genetic association of polymorphisms in the ACE-1 gene (DCP1) with an increased risk of AD (65)(66)(67). However, recent large genome-wide association studies did not identify ACE as a significant genetic risk factor (68 -70). However, recombinant ACE is capable of cleaving both A␤40 and A␤42 in vitro and thereby decreases A␤ aggregation and toxicity (71)(72)(73). We also observed efficient cleav- age of A␤ by ACE. This cleavage occurs directly before or after Ser-8. Interestingly, the phosphorylation at Ser-8 strongly inhibited ACE-mediated degradation of A␤.
In this study, we have demonstrated that phosphorylation at Ser-8 could selectively affect the degradation of A␤ by different proteases. Although the degradation of A␤ by plasmin was independent of phosphorylation, the cleavage of A␤ by IDE and ACE was strongly decreased upon phosphorylation at Ser-8. These data could have important implications for the accumulation of A␤ observed during the pathogenesis of AD. The decreased degradation of phosphorylated A␤ by IDE and ACE would eventually result in increased concentrations of this peptide in the brain. As shown recently, ϳ20 -30% of the total extractable pool of A␤ was phosphorylated at Ser-8 and found in plaques and oligomeric aggre-gates (34). Thus, phosphorylation not only decreases the clearance of A␤ but also increases its propensity to aggregate (35). This modification would promote the formation of neurotoxic variants in a dual way. As the phosphorylation of A␤ at Ser-26 has also been described (74,75), it will be interesting to also assess the effect of phosphorylation at this site on A␤ metabolism. The targeting of extracellular A␤ phosphorylation by inhibition of extracellular kinases or stimulation of A␤ dephosphorylation could be explored to facilitate A␤ clearance in the brain for prevention or therapy of AD.
Acknowledgment-We thank Dr. Sebastian Franken (Institute of Biochemistry and Molecular Biology, University of Bonn) for helpful suggestions and technical assistance in MALDI-TOF measurements. FIGURE 5. Plasmin degrades A␤ and pA␤. A-C, synthetic npA␤ and pA␤ were incubated with purified plasmin at 37°C for various time intervals (0, 5, 15, 30, and 60 min). Aliquots of the reaction mixture were analyzed by SDS-PAGE and silver staining (A and B) or by MALDI-TOF-MS (C). The main cleavage products representing A␤(1-28) and A␤(1-16) fragments were generated in both peptide variants. The corresponding masses of the cleavage products are given in Table 1. Values represent means Ϯ S.D. (n ϭ 3).