The production of amyloid-β (Aβ) is a key factor driving pathogenesis in Alzheimer's disease (AD). Increasing concentrations of soluble Aβ oligomers within the brain lead to synapse degeneration and the progressive dementia characteristic of AD. Since Aβ exists in both disease-relevant (toxic) and non-toxic forms, the factors that affected the release of toxic Aβ were studied in a cell model. 7PA2 cells expressing the human amyloid precursor protein released Aβ oligomers that caused synapse damage when incubated with cultured neurones. These Aβ oligomers had similar potency to soluble Aβ oligomers derived from the brains of Alzheimer's patients. Although the conditioned media from 7PA2 cells treated with the cellular prion protein (PrPC) contained Aβ, it did not cause synapse damage. The loss of toxicity was associated with a reduction in Aβ oligomers and an increase in Aβ monomers. The suppression of toxic Aβ release was dependent on the glycosylphosphatidylinositol (GPI) anchor attached to PrPC, and treatment of cells with specific GPIs alone reduced the production of toxic Aβ. The efficacy of GPIs was structure-dependent and the presence of sialic acid was critical. The conditioned medium from GPI-treated cells protected neurones against Aβ oligomer-induced synapse damage; neuroprotection was mediated by Aβ monomers. These studies support the hypothesis that the ratio of Aβ monomers to Aβ oligomers is a critical factor that regulates synapse damage.
Alzheimer's disease (AD) is a complex neurological disorder characterised by a progressive dementia. The amyloid hypothesis maintains that the pivotal event in AD is the production of amyloid-β (Aβ) peptides following the metabolism of the amyloid precursor protein (APP) . Increasing concentrations of Aβ in the brain correlate with disease progression . Critically, not all forms of Aβ have equal biological significance; toxicity is dependent on the state of Aβ whether that is the length of peptide, state of aggregation, homogeneity of aggregates or specific Aβ conformations. The key to understanding the amyloid hypothesis is the realisation that there exist disease-relevant forms of Aβ, while other forms are less toxic or biologically inert  and may play a role in normal synapse function . Thus, in this study, we sought to identify factors involved in the production and release of toxic forms of Aβ. The pathogenesis of AD is intimately linked with the loss of synapse function [5,6]. Many studies demonstrated close correlations between the loss of synaptic proteins such as synaptophysin, indicative of synapse degeneration and the degree of dementia in AD [7,8]. In this study, the amounts of synaptophysin and cysteine string protein (CSP) were measured to quantify synapse density in cultured neurones incubated with Aβ. The loss of synaptic proteins from cultured neurones incubated with Aβ provides a useful in vitro model in which to investigate AD-related synapse damage.
7PA2 cells [Chinese hamster ovary (CHO) cells stably transfected with human APP751 ] release the soluble Aβ oligomers that are considered to be key mediators of synapse damage in AD [10,11]. The properties of soluble Aβ released from 7PA2 cells are similar to those derived from the brains of AD patients [12–14]. Since the production of Aβ is affected by the presence of the cellular prion protein (PrPC) , the biochemistry of PrPC-induced suppression of Aβ production in 7PA2 cells was examined. PrPC is linked to membranes via a glycosylphosphatidylinositol (GPI) anchor  and is rapidly incorporated into living cells . Here, we show that the treatment of 7PA2 cells with PrPC reduced the release of toxic Aβ as measured by their ability to cause synapse degeneration in cultured neurones. This effect of PrPC was dependent on the composition of its GPI anchor; the GPI anchor attached to PrPC is unusual in that it contains sialic acid  and PrPC with a GPI anchor lacking sialic acid (desialylated PrPC) did not alter the release of toxic Aβ. Further studies demonstrate that the suppression of toxic Aβ oligomers was achieved with specific sialylated GPIs alone. The treatment of 7PA2 cells with either PrPC or sialylated GPIs reduced the release of Aβ oligomers, but increased the release of Aβ monomers.
Culture of 7PA2 cells
CHO cells stably transfected with a cDNA encoding APP751 (7PA2 cells) were maintained as described previously . For experiments, 7PA2 cells were grown in six-well plates until 80% confluent. Culture media were replaced with neurobasal medium containing B27 components (Invitrogen) ± test compounds and the cells were cultured for a further 3 days, and the conditioned medium (CM) from these cells (7PA2-CM) was collected. Cells were washed three times with ice-cold PBS and homogenised in an extraction buffer [150 mM NaCl, 10 mM Tris–HCl (pH 7.4), 10 mM EDTA and 0.2% SDS] containing mixed protease inhibitors [4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride, Aprotinin, Leupeptin, Bestatin, Pepstatin A and E-46] and a phosphatase inhibitor cocktail [PP1, PP2A, microcystin LR, cantharidin and p-bromotetramisole (Sigma)] at 106 cells/ml. Cellular debris was removed by centrifugation (20 min as 16 000×g), and the supernatant was collected. Both 7PA2-CM and cell extracts were centrifuged at 100 000×g for 4 h at 4°C and passed through a 50 kDa filter (Sartorius). CM from CHO cells (CHO-CM) were used as controls. 7PA2-CM, containing Aβ monomers, were prepared by filtration through a 10 kDa filter (Sartorius). To deplete 7PA2-CM of Aβ, they were incubated with 1 µg/ml monoclonal antibody (mAb) 4G8 (reactive with amino acids 17–24 of Aβ) or an isotype control (mock depletions) and incubated on rollers for 2 h. Protein G microbeads were added (10 µl/ml; Sigma) for 30 min and protein G-bound antibody–antigen complexes were removed by centrifugation (16 000×g for 5 min). For immunoblot analysis, 7PA2-CM/monomer or oligomer preparations were concentrated from 2000 to 100 µl using a 3 kDa filter (Sartorius). Ten microlitres of sample were mixed with an equal volume of 0.5% NP-40, 5 mM CHAPS and 50 mM Tris (pH 7.4) and separated by electrophoresis. Proteins were transferred onto a Hybond-P polyvinylidene fluoride membrane by semi-dry blotting and blocked using 10% milk powder. Aβ was detected by incubation with mAb 6E10 (Covance), biotinylated anti-mouse IgG, extravidin peroxidase and enhanced chemiluminescence.
To determine cell viability, thiazolyl blue tetrazolium bromide was added to cells at a final concentration of 50 µM for 3 h at 37°C. The supernatant was removed, the formazan product was solubilised in 200 µl of dimethyl sulfoxide, transferred to an immunoassay plate and absorbance read at 595 nm. Neuronal survival was calculated with reference to untreated cells (100% survival).
Isolation of detergent-resistant membranes (DRMs) (lipid rafts)
These membranes were isolated by their insolubility in non-ionic detergents as described previously . Briefly, cells were homogenised in an ice-cold buffer containing 1% Triton X-100, 10 mM Tris–HCl (pH 7.2), 150 mM NaCl, 10 mM EDTA and mixed protease/phosphatase inhibitors, and nuclei and large fragments were removed by centrifugation (300×g for 5 min at 4°C). The post-nuclear supernatant was incubated on ice (4°C) for 1 h and centrifuged (16 000×g for 30 min at 4°C). The supernatant was reserved as the detergent soluble membrane while the insoluble pellet was homogenised in extraction buffer supplemented with protease and phosphatase inhibitors (as above) at 106 cells/ml, centrifuged (10 min at 16 000×g), and the soluble material was reserved as the DRM fraction.
Primary neuronal cultures
Cortical neurones were prepared from the brains of mouse embryos (day 15.5) after mechanical dissociation and cell sieving as described previously . Neurones were plated at 2 × 105 cells/well in 48-well plates in Hams F12 containing 5% foetal calf serum for 2 h. Cultures were shaken (600 r.p.m. for 5 min) and non-adherent cells were removed by two washes in PBS. Cells were subsequently grown in neurobasal medium containing B27 components (Invitrogen) and nerve growth factor (5 ng/ml; Sigma) for 10 days. Immunohistochemistry showed that greater than 90% of cells were neurofilament positive. Neurones were incubated with Aβ preparations or prostaglandin E2, and synapse damage was assessed after 24 h. Neurones were washed three times in PBS and homogenised in extraction buffer containing mixed protease and phosphatase inhibitors (as above) at 106 cells/ml. All experiments were performed in accordance with European regulations (European Community Council Directive, 1986, 56/609/EEC) and approved by the local authority veterinary service/ethical committee.
Samples were mixed with Laemmli buffer containing β-mercaptoethanol, heated to 95°C for 5 min and proteins were separated by electrophoresis on 15% polyacrylamide gels. Proteins were transferred onto a Hybond-P polyvinylidene fluoride membrane by semi-dry blotting. Membranes were blocked using 10% milk powder; synapsin-1 was detected with goat polyclonal (Santa Crux Biotech), synaptophysin with MAB368 (Abcam), CSP with rabbit polyclonal anti-CSP (sc-33154; Santa Cruz), vesicle-associated membrane protein (VAMP)-1 with mAb 4H302 (Abcam), rabbit polyclonal antibodies to caveolin (Upstate), platelet activating factor (PAF) receptor with rabbit polyclonal anti-PAF receptor (Cayman Chemicals), APP with rabbit polyclonal anti-APP (Sigma) and PrPC by mAb 4F2 (Jaques Grassi; ). These were visualised using a combination of biotinylated anti-mouse/goat/rat/rabbit IgG (Sigma), extravidin peroxidase and enhanced chemiluminescence.
Maxisorb immunoplates (Nunc) were coated with an anti-synaptophysin mouse mAb (MAB368 — Millipore) and bound synaptophysin was detected using rabbit polyclonal anti-synaptophysin (Abcam) followed by a biotinylated anti-rabbit IgG, extravidin alkaline phosphatase and 1 mg/ml 4-nitrophenol phosphate solution (Sigma). The absorbance was measured on a microplate reader at 405 nm. Samples were expressed as ‘units synaptophysin’ where 100 units were defined as the amount of synaptophysin in 106 control neurones.
Cysteine string protein ELISA
Maxisorb immunoplates were coated with an anti-CSP mAb (Santa Cruz) and blocked with 5% milk powder. Samples were added and bound CSP was detected using rabbit polyclonal anti-CSP (Santa Cruz) followed by a biotinylated anti-rabbit IgG, extravidin alkaline phosphatase and 1 mg/ml 4-nitrophenol phosphate solution. The absorbance was measured at 405 nm. Samples were expressed as ‘units CSP’ where 100 units were defined as the amount of CSP in 106 control neurones.
Isolation of PrPC
PrPC and the control Thy-1 protein were isolated from murine GT1 neuronal cell membranes using a combination of immunoaffinity columns, size-exclusion chromatography (Superdex) and reverse-phase chromatography on C18 columns (Waters) as described previously . N-linked glycans were removed from PrPC by digestion with 2 units/ml endoglycosidase F (PNGase; Sigma), monoacylated PrPC by digestion with 100 units/ml bee venom phospholipase A2 (PLA2; Sigma) and desialylated PrPC by digestion with 0.2 units/ml neuraminidase (Clostridium perfringens — Sigma) for 2 h at 37°C . Digested PrPC preparations were purified using reverse-phase chromatography (C18 columns); PrP (prion protein)-positive fractions were pooled, desalted and lyophilised. For tissue culture studies, PrP-containing fractions were solubilised in culture medium by sonication prior to further use.
The amount of PrP in samples was measured by ELISA as described previously . Briefly, Maxisorb immunoplates were coated with mAb ICSM18 (gift by Dr M. Tayebi, Royal Veterinary College) and blocked with 5% milk powder. Samples were applied and detected with biotinylated mAb ICSM35 (gift by Dr M. Tayebi), followed by extravidin alkaline phosphatase and 1 mg/ml 4-nitrophenyl phosphate solution. The absorbance was measured at 405 nm and the concentrations of PrP in samples were calculated by reference to serial dilutions of recombinant murine PrP (Prionics).
Isolation of GPIs
Purified proteins were digested with 100 µg/ml proteinase K, at 37°C for 24 h, resulting in GPIs attached to the terminal amino acid. The released GPIs were extracted with water-saturated butanol, washed with water five times and loaded onto C18 columns. GPIs were eluted under a gradient of propanol and water. The presence of GPIs was detected by ELISA as described previously . Maxisorb immunoplates were coated with 0.5 µg/ml concanavalin A (binds mannose) and blocked with 5% milk powder in PBS-Tween. Samples were added and any bound GPI was detected by the addition of the phosphatidylinositol-reactive mAb 5AB3-11, followed by a biotinylated anti-mouse Ig (Sigma), extravidin alkaline phosphatase and 1 mg/ml 4-nitrophenyl phosphate solution. The absorbance was measured on a microplate reader at 405 nm. For some experiments, GPIs derived from PrPC were digested with 100 units/ml bee venom PLA2 (monoacylated GPI) or 0.2 units/ml neuraminidase (desialylated GPI) for 2 h at 37°C. The modified GPIs were isolated on C18 columns as above. GPIs were dissolved in ethanol at 2 µM (stock solutions) and diluted in tissue culture medium for bioassays.
Analysis of GPIs
The presence of phosphatidylinositol in GPI anchors was identified using mAb (5AB3-11) and specific glycans were detected with biotinylated lectins. Isolated GPIs were bound to nitrocellulose membranes by dot blot and blocked with 5% milk powder. Samples were incubated with mAb 5AB3-11, biotinylated Sambucus nigra agglutinin (SNA) (detects terminal sialic acid residues bound α-2,6 or α-2,3 to galactose), biotinylated concanavalin A (detects mannose) or biotinylated Ricinus communis Agglutinin I (RCA-I) (detects terminal galactose; Vector Labs). Bound lectins were visualised using extravidin peroxidase and enhanced chemiluminescence. The mAb was visualised by incubation with a horseradish peroxidase conjugated anti-murine-Ig and chemiluminescence. GPIs were separated by high-performance thin layer chromatography (HPTLC) on silica gel 60 plates using a mixture of chloroform/methanol/water (10/10/3, v/v/v). Plates were soaked in 0.1% polyisobutyl methacrylate in hexane, dried and blocked with 5% milk powder. GPIs were detected with a mAb that binds to phosphatidylinositol as described recently .
Soluble extracts were prepared from brain tissue from patients with a clinical, and pathologically confirmed, diagnosis of AD as described previously . Briefly, brain tissue, stored at −80°C, was thawed, cut into pieces of ∼100 mg and added to 2 ml tubes containing lysing matrix D beads (Q-Bio). Ice-cold 20 mM Tris (pH 7.4) containing 150 mM NaCl was added, so that there was the equivalent of 100 mg brain tissue/ml. The tubes were shaken for 10 min (Disruptor Genie, Scientific Instruments). This process was performed three times before tubes were centrifuged at 16 000×g for 10 min to remove particulate matter. Soluble material was prepared by passage through a 50 kDa filter (Sartorius) (16 000×g for 30 min to remove proteolytic enzymes, membrane-bound and plaque Aβ). The soluble material was then desalted (3 kDa filter; Sartorius) to eliminate bioactive small molecules and drugs and the retained material was collected (preparation contains peptides with molecular masss between 3 and 50 kDa) and stored at −80°C.
Maxisorb immunoplates were coated with mouse mAb anti-APP [Clone 1G6; BioLegend (epitope 573–596)] and blocked with 5% milk powder in PBS-Tween, and samples were applied for 1 h. Bound APP was detected using rabbit polyclonal antibodies against the N-terminal of APP (amino acids 40–60 — Sigma) followed by anti-rabbit IgG conjugated to alkaline phosphatase followed by 1 mg/ml 4-nitrophenol phosphate solution and optical density was read in a spectrophotometer at 405 nm. Results were calculated by comparison to serial dilutions of cell extracts from control cells.
To detach Aβ42 from membrane components that block specific epitopes, samples (300 µl) were mixed with 700 µl of propan-2-ol and sonicated. Proteins were precipitated by adding 250 µl of 100% (w/v) trichloroacetic acid, incubating on ice for 30 min and centrifugation (16 000×g for 10 min at 4°C). The pellet was washed twice with ice-cold acetone, dried, suspended in a buffer containing 150 mM NaCl, 10 mM Tris–HCl (pH 7.4), 10 mM EDTA and 0.2% SDS and sonicated.
Nunc Maxisorb immunoplates were coated with mAb 4G8 (epitope 17–24; Covance) in carbonate buffer overnight. Plates were blocked with 5% milk powder in PBS-Tween and samples were applied. The detection antibody was an Aβ42 selective rabbit mAb BA3-9 (Covance) followed by biotinylated anti-rabbit IgG and extravidin alkaline phosphatase (Sigma). Total Aβ was visualised by the addition of 1 ng/ml 4-nitrophenol phosphate solution and optical density was read in a spectrophotometer at 405 nm.
Maxisorb immunoplates were coated with mAb 4G8 (epitope 17–24) and blocked with 5% milk powder in PBS-Tween. Samples were applied and Aβ40 was detected with rabbit polyclonal PC-149 (Merck) followed by biotinylated anti-rabbit IgG and extravidin alkaline phosphatase. Total Aβ was visualised by the addition of 1 ng/ml 4-nitrophenol phosphate and optical density was read in a spectrophotometer at 405 nm.
Comparison of treatment effects was carried out using Student's paired t-tests. Error values are standard deviation (SD), and significance was determined where P < 0.01. Bivariate analysis using Pearson's coefficient (IBM SPSS statistics 20) was used to examine correlations between data sets.
7PA2 cells release toxic Aβ
In support of the hypothesis that Aβ caused synapse damage, the CM from 7PA2 cells caused dose-dependent reductions in synapsin-1, VAMP-1, CSP and synaptophysin from cultured neurones indicative of synapse damage (Figure 1A). The addition of 100 µg/ml 7PA2-CM did not reduce neuronal viability as measured by thiazolyl blue tetrazolium ((101 ± 6% cell survival compared with 100 ± 5%, n = 9, P = 0.6), indicating that synapse degeneration occurred in the absence of any significant neuronal death. Whereas 7PA2-CM caused dose-dependent reductions in both synaptophysin (Figure 1B) and CSP (Figure 1C), the addition of Aβ-depleted CM did not, indicating that Aβ was the synaptotoxic element. Immunoprecipitation studies demonstrated that the toxic entity in these CM was Aβ. Immunodepletion with mAb G48 reduced the concentrations of both Aβ42 (0.04 ± 0.04 nM Aβ42 compared with 2.21 ± 0.3 nM, n = 9, P < 0.01) and Aβ40 (0.57 ± 0.28 nM Aβ40 compared with 7.42 ± 0.6 nM Aβ40, n = 9, P = 0.38). Mock depletions with a control IgG mAb did not significantly affect the concentration of either Aβ40 (6.95 ± 0.81 nm Aβ40 compared with 7.42 ± 0.6 nM Aβ40, n = 9, P = 0.2) or Aβ42 (2.03 ± 0.21 nM Aβ42 compared with 2.21 ± 0.3 nM Aβ42, n = 9, P = 0.21). Not all forms of Aβ have equal biological significance, as toxicity is dependent on the state of Aβ. To determine whether 7PA2-CM contained Aβ similar to the toxic Aβ in the brains of Alzheimer's patients it was compared with that of soluble brain extracts. 7PA2-CM and brain extracts containing similar concentrations of Aβ42 caused similar dose-dependent reductions in synaptophysin (Figure 1D) and CSP (Figure 1E).
Soluble Aβ caused synapse damage in neurones.
PrPC reduced the release of toxic Aβ from 7PA2 cells
Since PrPC had been reported to inhibit the production of Aβ42 , the effects of PrPC on 7PA2 cells were studied. The viability of 7PA2 cells, as measured by thiazolyl blue tetrazolium, was not significantly affected by 10 nM PrPC (102 ± 6% cell survival compared with 100 ± 5%, n = 6, P = 0.37). CM from 7PA2 cells treated with 10 nM PrPC did not cause synapse damage, as measured by the loss of synaptophysin (Figure 2A) from cultured neurones. PrPC (10 nM) was used in these assays as this is the PrPC concentration in neurones . The effects of PrPC were dose-dependent. CM from cells with recombinant PrP (PrP protein lacking any post-translational modifications) had similar effects to CM from control cells, indicating that the inhibitory effect of PrPC was due to a post-translational modification. The effects of N-linked glycans PrPC upon the production of toxic Aβ were examined; CM from 7PA2 cells treated with PrPC from which N-linked glycans had been removed did not cause synapse damage. There were no significant differences between PrPC and PrPC lacking N-linked glycans upon the suppression of toxic Aβ production as measured by loss of synaptophysin (Figure 2B).
PrPC reduced the release of toxic Aβ.
PrPC-mediated suppression of toxic Aβ release is dependent on its GPI
PrPC is connected to cell membranes by a GPI anchor . The role of the GPI in PrPC-mediated suppression of toxic Aβ was examined using monoacylated PrPC, a form of PrPC differing from PrPC only in the lack of an acyl chain in its GPI anchor . Following the addition of 10 nM PrPC preparations, similar amounts of PrPC and monoacylated PrPC were found in 7PA2 cells (9.3 ± 0.6 nM compared with 9.5 ± 0.5 nM, P = 0.74, n = 9). In addition, the viability of 7PA2 cells as measured by thiazolyl blue tetrazolium was not significantly affected by 10 nM monoacylated PrPC (101 ± 6% cell survival compared with 100 ± 5%, n = 6, P = 0.75). CM from cells treated with 10 nM monoacylated PrPC caused a reduction in synaptophysin similar to CM from untreated cells, demonstrating that the inhibitory effect of PrPC on toxic Aβ production was GPI-dependent (Figure 2C). The CM from 7PA2 cells treated with Thy-1, another GPI-anchored protein, caused synapse damage, as measured by the loss of synaptophysin, similar to CM from control cells (Figure 2C), indicating that not all GPI-anchored proteins had this suppressive effect. Since the toxicity of 7PA2-CM is Aβ-dependent, it was surprising to see that treatment with 10 nM PrPC had only a small effect on the concentrations of Aβ42 (2.05 ± 0.22 nM compared with 2.45 ± 0.38 nM, n = 9, P = 0.02) or Aβ40 (6.94 ± 1.04 nM compared with 7.28 ± 0.72 nM, n = 9, P = 0.24) in CM.
PrPC reduced the release of Aβ oligomers
Reports that Aβ monomers are not toxic [25,26] suggested that the toxicity of 7PA2-CM was relative to the concentrations of Aβ oligomers rather than the total Aβ. Immunoblots showed that 7PA2-CM contained monomers, dimers and trimers (Figure 3A, Lane 1) while 7PA2-CM retained by a 10 kDa filter was depleted of monomers but contained dimers and trimers (Figure 3A, Lane 2). Treatment with PrPC, but not monoacylated PrPC, caused a dose-dependent reduction in Aβ42 oligomers (Figure 3B). Treatment with 10 nM PrPC also reduced the concentrations of Aβ40 oligomers from 0.41 ± 0.16 to 0.2 ± 0.14 nM (n = 9, P = 0.03), whereas 10 nM monoacylated PrPC had no significant effect (0.47 ± 0.11 nM compared with 0.41 ± 0.16, n = 9, P = 0.4). There was a significant inverse correlation between the concentrations of Aβ42 oligomers in CM from cells treated with PrPC (1.25–10 nM) and amounts of synaptophysin in neurones incubated with these CM (Pearson's coefficient = −0.91, P < 0.01; Figure 3C).
PrPC reduced the release of Aβ oligomers.
Immunoblots showed that 7PA2-CM that passed through a 10 kDa filter contained only Aβ monomers (Figure 4A, Lane 2). Treatment with PrPC, but not monoacylated PrPC, caused a dose-dependent increase in the concentrations of Aβ42 monomers in CM (Figure 4B). Treatment with 10 nM PrPC did not significantly alter the concentrations of Aβ40 monomers (7.1 ± 0.62 nM Aβ40 compared with 6.6 ± 0.58 nM, n = 9, P = 0.07). A significant inverse correlation between the concentrations of Aβ42 oligomers and Aβ42 monomers in CM from 7PA2 cells incubated with PrPC (1.25–10 nM, Pearson's coefficient = −0.75, P < 0.01) was observed (Figure 4C).
PrPC increased the release of Aβ monomers.
GPIs reduced the release of toxic Aβ
Since these results indicated that the GPI attached to PrPC had a role in suppressing Aβ42 oligomer production, we hypothesised that GPIs alone could alter Aβ production. The analysis of GPIs isolated from PrPC and Thy-1 by HPTLC (Figure 5A) and reverse-phase chromatography on C18 columns (Figure 5B) showed that they had different properties. Whereas CM from 7PA2 cells treated with 10 nM GPIs derived from PrPC did not cause synapse damage [did not affect the synaptophysin (Figure 5C) or CSP (Figure 5D) content of neurones], the CM from 7PA2 cells treated with 10 nM GPIs isolated from Thy-1 caused extensive loss of synaptophysin and CSP. The effects of PrPC-derived GPIs on the production of toxic Aβ were dose-dependent (Figure 5E). These results showed that the GPI-induced suppression of toxic Aβ production was structure-dependent.
GPIs reduced the release of toxic Aβ.
CM from cells treated with PrPC-derived GPIs, but not Thy-1-derived GPIs, showed a dose-dependent reduction in Aβ42 oligomers (Figure 6A) and increase in Aβ42 monomers (Figure 6B). There was a significant inverse correlation between concentrations of Aβ42 oligomers and Aβ42 monomers in CM from 7PA2 cells incubated with PrPC-derived GPIs (1.25–10 nM; Pearson's coefficient = 0.87, P < 0.01; Figure 6C). The loss of Aβ oligomers and corresponding increase in Aβ monomers suggested that the GPIs may interact directly with Aβ oligomers causing them to dissociate into monomers. When Aβ preparations (containing 10 nM Aβ42) were incubated with 10 nM GPIs at 37°C for 3 days and analysed by electrophoresis and immunoblot, there was no difference in the proportion of oligomers and monomers (Figure 6D). In addition, monomeric forms of Aβ42 were not found in Aβ oligomer preparations incubated with GPIs, indicating that GPIs did not cause the dissociation of Aβ oligomers.
GPIs increased the release of Aβ monomers.
Sialylated GPIs reduced the release of toxic Aβ42
The composition of GPIs derived from PrPC is unusual in that they contain sialic acid ( and (Figure 7A)), a rare modification of mammalian GPIs. To examine the structure–function relationship in more detail, the GPIs derived from PrPC were digested to create desialylated GPIs and monoacylated GPIs. GPIs, desialylated GPIs and monoacylated GPIs were visualised by HPTLC (Figure 7B) and isolated by reverse-phase chromatography on C18 columns (Figure 7C). In dot blots, all GPIs reacted with mAb 5AB3-11 (reactive with phosphatidylinositol) and concanavalin A (mannose). SNA (detects terminal sialic acid residues bound α-2,6 or α-2,3 to galactose) bound to GPIs and monoacylated GPIs but not to desialylated GPIs (Figure 7D). Conversely, RCA-1 (detects terminal galactose) only bound to desialylated GPIs.
Analysis of GPIs.
The CM from 7PA2 cells treated with 10 nM GPI preparations was added to cultured neurones. CM from 7PA2 cells treated with control medium, 10 nM monoacylated GPIs or 10 nM desialylated GPIs reduced the amounts of synaptophysin in neurones, whereas CM from cells treated with GPIs did not (Figure 8A). The effects of PrPC-derived GPIs were dose-dependent (Figure 8B). The treatment of 7PA2 cells with 10 nM GPIs, but not monoacylated GPIs or desialylated GPIs (10 nM), reduced the concentrations of Aβ42 oligomers (Figure 8C) and increased the concentrations of Aβ42 monomers in CM (Figure 8D).
GPI-mediated suppression of toxic Aβ release is dependent on sialic acid.
GPIs increased the release of neuroprotective Aβ
To determine if Aβ monomers have a protective role, the CM from GPI-treated 7PA2 cells were mixed with brain extract containing 2 nM Aβ42 and incubated with neurones. The brain extract caused a reduction in synaptophysin (Figure 9A) and CSP (Figure 9B). The presence of CM from GPI-treated 7PA2 cells protected neurones against the brain extract (Aβ)-induced synapse damage. It was important to determine whether the protective effect of CM from treated 7PA2 cells was mediated by Aβ. Therefore, brain extracts were incubated with CM from GPI-treated cells after Aβ had been removed by immunoprecipitation and added to neurones. The removal of Aβ removed the protective effects of CM from GPI-treated cells (Figure 9A,B). Next, Aβ monomers were isolated from the CM of GPI-treated cells. Aβ monomers blocked the brain extract-induced reductions in synaptophysin (Figure 9C) and CSP (Figure 9D). The protective effects of monomer preparations derived from GPI-treated 7PA2 cells were compared with that of monomers prepared from brain extracts. Monomer preparations derived from 7PA2-CM and brain extract had similar activity when based on their Aβ42 content (Figure 9E). The protective effect of these monomer preparations was stimulus-specific; there were no significant differences in either synaptophysin (31 ± 7 units compared with 34 ± 10, n = 9, P = 0.37) or CSP (54 ± 6 units compared with 50 ± 13, n = 9, P = 0.4) content of neurones incubated with 10 nM prostaglandin E2 ± monomer preparations containing 10 nM Aβ42.
GPI-treated 7PA2 cells release neuroprotective Aβ monomers.
PrPC and GPIs reduced cell-associated Aβ
The possibility that PrPC or GPIs affected the release rather than the production of Aβ was examined by measuring Aβ in cell extracts from treated 7PA2 cells. The concentrations of Aβ42 in 7PA2 cells was significantly reduced by treatment with 10 nM PrPC (6.7 ± 1.7 nM Aβ42 compared with 12.15 ± 1.93 nM, n = 12, P < 0.01) or with 10 nM GPIs derived from PrPC (5.67 ± 1.34 nM Aβ42 compared with 12.15 ± 1.93 nM, n = 12, P < 0.01). The concentrations of Aβ42 were not significantly affected by treatment with 10 nM monoacylated PrPC (12.55 ± 2.1 nM Aβ42 compared with 12.15 ± 1.93 nM, n = 12, P = 0.74) or 10 nM desialylated GPIs (12.76 ± 2.16 nM Aβ42 compared with 12.15 ± 1.93 nM, n = 12, P = 0.42).
GPIs altered the distribution of APP
Further studies examined the effects of GPIs upon APP. Treatment with 10 nM GPI did not significantly alter concentrations of APP within cells (97 ± 3.4 units compared with 100 ± 2.8 units, n = 6, P = 0.14). The processing of APP to toxic Aβ peptides is thought to occur within membrane micro-domains called lipid rafts, and in control cells, ∼25% of cellular APP was found within DRMs (lipid rafts). Treatment with 10 nM GPI, but not with 10 nM desialylated GPI, reduced the amounts of APP found within DRMs in a dose-dependent manner (Figure 10A). The treatment of 7PA2 cells with 10 nM GPI did not affect the amounts of other raft-associated proteins including caveolin and the PAF receptor in DRMs (Figure 10B). In cells treated with GPIs (1.25–10 nM), there was a significant correlation between the amounts of APP found in lipid rafts and the concentrations of Aβ42 oligomers found in CM (Figure 10C). There was also a significant inverse correlation between the amounts of APP found in lipid rafts and the concentrations of Aβ42 monomers found in CM (Figure 10D).
GPIs reduced APP within lipid rafts.
There are two key findings from the present study; firstly that the release of toxic Aβ from 7PA2 cells is controlled by a pathway sensitive to the presence of PrPC. More specifically that it was the sialic acid contained within the GPI anchor attached to PrPC that affected Aβ production. Secondly, we report that the GPI-induced changes involved both a reduction in toxic Aβ oligomers and an increase in neuroprotective Aβ monomers.
The present study concentrated on the biologically active forms of Aβ released by 7PA2 cells by measuring their effects upon synapses (based on reports that synapse damage, as measured by the loss of synaptic proteins, is a good correlate of dementia in AD [7,8,27,28]). The finding that synapse damage caused by CM from 7PA2 cells was comparable to that caused by soluble Alzheimer's brain extracts (with regard to their Aβ42 content), implied that 7PA2 cells release toxic Aβ oligomers that are similar to those found within the brains of AD patients. The major observation was that the CM collected from 7PA2 cells treated with PrPC did not damage synapses. However, in contrast with the initial report that expression of PrPC reduced Aβ production in transfected human neuroblastoma cells , we found that PrPC had only minor effects on the concentrations of Aβ40 and Aβ42. It should be noted that the two systems are very different with regard to cell types and the amounts of PrPC expressed.
Observations that recombinant PrP, which does not contain post-translational modifications such as GPIs and N-linked glycans, did not affect the production of toxic Aβ demonstrated that the protein alone was not responsible for suppression of Aβ production and led us to examine the effects of post-translational modifications of PrPC. We concluded that N-linked glycans were not necessary for the suppression of toxic Aβ production as their removal from PrPC also inhibited the production of toxic Aβ. In contrast, monoacylated PrPC, a form of PrPC that differed only in the composition of its GPI anchor, did not suppress toxic Aβ release, indicating a role for the GPI anchor in regulating Aβ production. The observation that isolated GPIs were capable of suppressing toxic Aβ release and altering the ratio of Aβ monomers to Aβ oligomers showed that the GPI alone affected Aβ production and that the GPI was not simply a contributory factor to protein interactions mediated by PrPC. The structure of the GPI was important as GPIs isolated from Thy-1, or GPIs derived from PrPC that had been modified (monoacylated or desialylated GPIs) did not affect the release of toxic Aβ.
The observation that PrPC reduced the toxicity of CM from 7PA2 cells without causing major changes in the concentrations of Aβ40 and Aβ42 indicated that it affected the forms of Aβ produced. Thus, the presence of PrPC has two major effects; it reduced the concentrations of Aβ42 oligomers, responsible for synapse damage, and it increased the release of Aβ42 monomers. Immunodepletion and filtration studies showed that the protective effect was mediated by Aβ monomers, an observation consistent with a prior report . These results are consistent with the hypothesis that it is the ratio of Aβ monomers to Aβ oligomers that is critical factor in determining the toxicity of CM and explained the observation that CM from cells treated with GPIs (which increased concentrations of Aβ42 monomers) blocked synapse damage caused by soluble Aβ oligomers. The protective effects of Aβ monomers were stimulus-specific; they did not affect synapse damage induced by prostaglandin E2 indicating a selective rather than a universal action.
The GPIs did not have a direct effect on Aβ42 oligomers (did not cause the dissociation of oligomers into monomers); rather, the effects of GPIs were upon the production of oligomers and monomers. Currently, we can only speculate how specific sialylated GPIs might alter the production of toxic Aβ. Observations that GPIs help solubilise cholesterol and that GPI-anchored proteins triggered the formation of lipid rafts [30,31] implicate GPI-anchored proteins as regulators of lipid raft structure and function. The processing of APP by β- and γ-secretases to form Aβ is affected by the composition of cell membranes, more specifically, the form and function of lipid rafts [32,33]. Notably, PrPC-mediated inhibition of β-secretase was dependent on lipid rafts , and in this study, monoacylated PrPC and monoacylated GPIs, which were not found within lipid rafts , did not affect Aβ production, observations that support the idea that GPIs affected lipid raft function.
Not all GPIs suppressed the production of toxic Aβ; GPIs derived from Thy-1, monoacylated PrPC or desialylated PrPC had an inhibitory effect. The glycan structure of the GPI anchor mediates protein association with specific rafts  and affects the function of those rafts . The composition of lipid rafts surrounding GPIs is dependent on multiple interactions between the glycans and membrane lipids [36,37], and the removal of sialic acid from the GPI of PrPC changed the properties of the surrounding lipid raft; it allowed increased concentrations of gangliosides and cholesterol . Although APP metabolism to toxic Aβ is thought to occur within lipid rafts , cells contain multiple, heterogeneous lipid rafts each with different composition and functions . We hypothesised that GPIs derived from PrPC are targeted to lipid rafts involved in the metabolism of APP to toxic Aβ. Lipid rafts are enriched with signalling molecules and act as domains in which the GPI anchors attached to PrPC interact with cell signalling pathways . The GPIs attached to PrPC activate cPLA2 , an enzyme that affects APP processing . As this enzyme is essential for the maintenance of the endoplasmic reticulum–trans Golgi network , a pathway reported to regulate APP metabolism , then inhibition of cPLA2 may affect the intracellular trafficking of APP and hence its metabolism to toxic Aβ.
APP and many of the enzymes involved in the generation of Aβ are found in lipid rafts [44–46], and in control 7PA2 cells, ∼25% of APP was found within lipid rafts. The observation that GPIs reduced the amounts of APP within lipid rafts is consistent with reports that GPIs sequester cholesterol and consequently affect lipid raft composition and function [31,47]. The finding that there was a significant positive correlation between the concentrations of Aβ42 oligomers and the amounts of APP in lipid rafts of Aβ42 monomers suggests that membrane targeting of APP is a key factor in production of Aβ oligomers. The protein cargos of the cell membrane traffic via different pathways to those within lipid rafts  and APP in lipid rafts may be targeted to different cell compartments (and consequently interacts with a different range of enzymes) than APP found in the cell membrane.
In summary, the present study demonstrated that the release of toxic Aβ by 7PA2 cells is sensitive to the presence of PrPC and more specifically, its sialylated GPI anchor. Critically, sialylated GPIs derived from PrPC increased the release of neuroprotective Aβ monomers and reduced the release of toxic Aβ oligomers.
amyloid precursor protein
Chinese hamster ovary
cysteine string protein
high-performance thin layer chromatography
platelet activating factor
Ricinus communis Agglutinin I
Sambucus nigra agglutinin
All authors were involved in the planning, implementation and analysis of results. CB was responsible for writing and preparation of manuscript.
This work was supported by a grant from the European Commission FP6 ‘Neuroprion’—Network of Excellence and Royal Veterinary College Bioveterinary Science Research project funding.
We thank Professor E. Koo for the gift of 7PA2 cells.
The Authors declare that there are no competing interests associated with the manuscript.