Amyloid-β42 Interacts Mainly with Insoluble Prion Protein in the Alzheimer Brain*

The prion protein (PrP) is best known for its association with prion diseases. However, a controversial new role for PrP in Alzheimer disease (AD) has recently emerged. In vitro studies and mouse models of AD suggest that PrP may be involved in AD pathogenesis through a highly specific interaction with amyloid-β (Aβ42) oligomers. Immobilized recombinant human PrP (huPrP) also exhibited high affinity and specificity for Aβ42 oligomers. Here we report the novel finding that aggregated forms of huPrP and Aβ42 are co-purified from AD brain extracts. Moreover, an anti-PrP antibody and an agent that specifically binds to insoluble PrP (iPrP) co-precipitate insoluble Aβ from human AD brain. Finally, using peptide membrane arrays of 99 13-mer peptides that span the entire sequence of mature huPrP, two distinct types of Aβ binding sites on huPrP are identified in vitro. One specifically binds to Aβ42 and the other binds to both Aβ42 and Aβ40. Notably, Aβ42-specific binding sites are localized predominantly in the octapeptide repeat region, whereas sites that bind both Aβ40 and Aβ42 are mainly in the extreme N-terminal or C-terminal domains of PrP. Our study suggests that iPrP is the major PrP species that interacts with insoluble Aβ42 in vivo. Although this work indicated the interaction of Aβ42 with huPrP in the AD brain, the pathophysiological relevance of the iPrP/Aβ42 interaction remains to be established.

order. The underlying pathology in AD seems to be associated with the accumulation of soluble and insoluble aggregated species of the amyloid-␤ (A␤) peptide in the brain (1). However, the mechanisms underlying A␤ deposition and neurotoxicity remain poorly understood. The cellular prion protein (PrP C ) is a glycoprotein highly expressed in the brain, and best known for its association with prion diseases. These are unique neurodegenerative disorders with an infectious, sporadic or genetic etiology, and which are characterized by deposition of misfolded, pathological PrP (PrP Sc ) in the brain (2). Interestingly, a recent interpretation of early and newer observations suggests that PrP C may play a role in the pathogenesis of AD (3). Epidemiological studies suggest that the Met/Val polymorphism at residue 129 in PrP modulates the number of A␤ deposits (4). Also, pathological evidence indicates that PrP deposits often accompany A␤ plaques in AD (5)(6)(7). Moreover, transgenic mice expressing mutant amyloid precursor protein (APP) and overexpressing hamster PrP C present an exacerbated A␤ plaque burden (8). The circumstantial evidence of an association between PrP and A␤ was greatly strengthened by the recent finding that PrP was the protein that most strongly supported the binding of cells to soluble A␤42 oligomers in a screen of 225,000 murine clones (9). The authors also showed that although A␤42 oligomers suppressed long-term potentiation (LTP) in CA1 hippocampal neurons in mouse brain slices, LTP inhibition did not occur in brain slices from mice lacking the PrP gene (Prnp Ϫ/Ϫ ) (9). Moreover, deleting the N-terminal region of murine PrP (PrP32-106) or blocking that region with specific antibodies against PrP93-109 significantly reduced LTP inhibition (9). In a genetic AD model where the A␤ plaque burden resulted from APP and presenilin 1 (APPswe/PS1 ⌬E9 ) transgenes, the elimination of PrP (Prnp Ϫ/Ϫ ) rescued the impairment of spatial learning and memory (10). This evidence led the authors to conclude that PrP mediates A␤ neurotoxicity through direct binding to toxic A␤ species (3).
However, conflicting results raise questions about the pathophysiological relevance of the PrP/A␤ interaction. For instance, overexpression of PrP C inhibited ␤-secretase cleavage of APP and the consequent production of A␤, whereas A␤ production was increased in PrP-deficient cells and the PrP knock-out mouse brain (11). The most robust evidence against the presumed key role of PrP C in A␤-dependent LTP mediation and memory impairment comes from new observations reported independently by three research groups (12)(13)(14). Using different approaches, the three groups assessed the effect of removing Prnp, and all concluded that absence of PrP did not affect memory impairment induced by A␤42 oligomers. First, injection of A␤42 oligomers into the brain of mice resulted in longterm memory impairment regardless of the presence of PrP (12). Moreover, neither elimination nor overexpression of PrP affected LTP dysfunction caused by A␤ accumulation in a different mouse model of AD (APP KM670/671NL /PS1 L166P ) (13). Finally, viral expression of mutant APP in hippocampal slices induced LTP inhibition in control and Prnp Ϫ/Ϫ mice brains (14). Thus, despite disagreement over the role of PrP in AD (15), the specific binding of A␤42 oligomers to mouse PrP (moPrP) was confirmed in these studies.
Because there are 20 amino acid differences between the mature mouse and human PrP (huPrP), these studies did not address the question of whether huPrP interacts with A␤42. In particular, the moPrP93-109 sequence that reportedly binds A␤42 oligomers contains three different amino acids with respect to huPrP94 -110. However, another recent paper addressed this question by immobilizing recombinant huPrP to a sensor chip to determine binding to A␤42 (16). This study confirmed the binding of huPrP to A␤42 and the role of the N-terminal domain in this interaction, thus reproducing the binding results obtained with moPrP.
Now that several studies have confirmed the binding of murine and human PrP to A␤42, the next question is whether this interaction also occurs in the human brain in vivo and whether differential binding occurs in AD and control brains. In our study, using gel filtration, we found that aggregated forms of huPrP and A␤ were co-purified in AD brains. Also, A␤ was co-immunoprecipitated with huPrP in brain homogenates from AD patients. Moreover, we observed that insoluble forms of PrP and A␤ were co-captured by a single-stranded DNAbinding protein called gene 5 protein (g5p). The g5p molecule specifically captures various PrP Sc species in prion-infected brains, as well as an insoluble PrP conformer (called iPrP) in uninfected brains (17,18). Using a peptide membrane array, we identified distinct A␤42 and A␤40 binding sites on huPrP. Most A␤42-specific binding is clustered in the unfolded N-terminal domain, especially in the octapeptide repeat region, an observation that confirms previous reports (9,12,16). Two additional A␤42-specific binding sites were observed on the middle and C-terminal PrP domains, respectively. Our study demonstrates that the interaction of huPrP and A␤ is found exclusively in the AD brain, and that this interaction principally involves insoluble forms of huPrP and A␤.
Brain Tissues-Following protocols approved by the Institutional Review Board of Case Western Reserve University (Cleveland, OH), frontal cortex was collected at autopsy and stored at Ϫ80°C for this study. The postmortem interval of these brain tissues was between 1 and 24 h. Cases of clinically and pathologically diagnosed AD (n ϭ 5, ages 83.4 Ϯ 2.7 between 80 and 87 years) and normal controls (n ϭ 5, ages 73.6 Ϯ 16.7 between 51 and 92 years) were used. Gray matter was dissected out and homogenized as described below. Also, transgenic mice expressing human APP, carrying both the Swedish (K670N, M671L) and Indiana (V717F) mutations (22,23), were euthanized with pentobarbital following the university-approved animal protocol. The brains from mice aged between 2 and 18 months either expressing APP (n ϭ 5) or wild type (n ϭ 9) were dissected and immediately stored at Ϫ80°C.
Preparation of Gene 5 Protein (g5p)-The recombinant g5p was isolated from Escherichia coli, transformed with an Ff gene 5-containing plasmid, and purified using DNA-cellulose affinity plus Sephadex G75 sizing columns, as described (24). The purity was Ͼ99% as determined by quantitation of Coomassie Blue-stained bands on SDS-PAGE.
Size Exclusion Chromatography-Size exclusion chromatography, also called gel filtration, was performed as described previously (17). In brief, Superdex 200 HR beads (GE Healthcare) in a 1 ϫ 30-cm column were used to determine the oligomeric state of PrP and A␤ molecules. Chromatography was performed in an FPLC (fast protein liquid chromatography) system (GE Healthcare) at a flow rate of 250 l/min, and fractions of 250 l each were collected. Supernatant prepared by centrifugation of 20% brain homogenate at 1,000 ϫ g for 10 min at 4°C was incubated with an equal volume of 2% Sarcosyl for 30 min on ice. A 200-l sample of each was injected into the column for each size exclusion run. The molecular mass of the various PrP and A␤ species recovered in different FPLC fractions was evaluated according to a calibration curve generated with gel filtration of the molecular mass markers (Sigma) including dextran blue (2,000 kDa), thyroglobulin (669 kDa), apoferritin (443 kDa), ␤-amylase (200 kDa), alcohol dehydrogenase (150 kDa), albumin (66 kDa), and carbonic anhydrase (29 kDa). These standards were loaded independently at the concentrations recommended by Sigma in 200-l sample volumes.

Determination of Binding Sites by Peptide Membrane Arrays-
The general method for preparing multiple overlapping peptides bound to cellulose membranes has been described in detail previously (25). The PrP peptide membrane array contained 99 overlapping 13-mer peptides spanning the entire sequence of the mature full-length human PrP23-231; each successive peptide was shifted by 2 amino acids from the previous one from the N to C terminus of PrP. The arrays were synthesized in 9 lines, with 12 spots in each line except the last line with only 3 spots. The membranes were blocked with 5% skim milk in TBST (150 mM NaCl, 0.05% Tween 20, 10 mM Tris-HCl, pH 7.6) at 37°C for 2 h; the huPrP peptide membrane was incubated with A␤40 or A␤42 at the designated concentrations in TBS-T for 3 h, and then incubated with 4G8, an anti-A␤ monoclonal antibody, at 1:3,000 in 3% skim milk for 2 h at 37°C. The membrane was washed with TBST and then incubated at 37°C with 1:4,000 horseradish peroxidase (HRP)-conjugated sheep anti-mouse IgG for 1 h. After a final wash, the membrane was treated with the ECL Western blotting detection reagent (Amersham Biosciences), and the signal was detected using a Bio-Rad Fluorescent Imager. The control experiments included the membranes only probed with the secondary antibody, HRP-conjugated sheep anti-mouse IgG, omitting the 4G8 antibody or membranes probed with 4G8 antibody followed by the secondary antibody in the absence of A␤ incubation.
Specific Capture or Co-immunoprecipitation of PrP and A␤ by g5p or 3F4-The g5p molecule or anti-PrP antibody 3F4 was conjugated to 7 ϫ 10 8 tosyl-activated magnetic beads in 1 ml of PBS at 37°C for 20 h. The specific capture or immunoprecipitation of PrP or A␤ by g5p or 3F4 was performed as described (17,26) by incubating S1, S2, or P2 fractions with conjugated beads.
Negative Staining and Electron Microscopy-A␤40 or A␤42 in PBS was adsorbed onto carbon films supported on Formvar membrane-coated nickel grids. The excess buffered protein solution was removed, and negatively stained with 2% uranyl acetate. Grids were then washed by touching the buffer and the excess buffer was immediately blotted using Whatman filter paper. Grids were then air-dried and kept at room temperature. Negatively stained specimens were observed by a JEOL 1200EX electron microscope (JEOL, Tokyo, Japan) with 80 kV of electron acceleration voltage.
Western Blot Analysis-After determining the protein concentrations of each sample, equal amounts of samples were resolved on 15% Tris-HCl Criterion pre-cast gels (Bio-Rad) at 150 V for ϳ80 min. The proteins on the gels were transferred to PVDF for 2 h at 70 V. The membranes were incubated for 2 h at room temperature with primary antibodies 3F4 (1:40,000), 6D11 (1:6,000), anti-N (1:6,000), 4G8 (1:3,000), or 6E10 (1:3,000) for probing the PrP molecule, or A␤. Following incubation with HRP-conjugated sheep anti-mouse IgG at 1:3,000 or goat anti-rabbit IgG at 1:3,000, the PrP, or A␤ bands or spots were visualized on Kodak film using ECL Plus, in accordance with the manufacturer's protocol.

Co-elution of Aggregated PrP and A␤ from Brain of AD Mice-
We have previously shown that the brains of uninfected humans and animals accumulate small amounts of various insoluble PrP oligomers and aggregates in addition to the dominant monomeric PrP C species by gel filtration (17). We denoted these newly identified PrP conformers insoluble PrP species (iPrP), suggesting a possibility that they may be precursors or partners in prion diseases or other protein aggregation disorders. PrP deposits have been observed histologically around the amyloid plaques in AD brains (5-7), raising the possibility that unique isoforms of PrP physically interact with A␤ in vivo. To determine whether PrP aggregates interact with A␤ in vivo and which specific PrP conformers are involved in this interaction, we performed gel filtration with brain homogenates from the AD transgenic mouse model that expresses human APP carrying both the Swedish and Indiana mutations, resulting in high levels of A␤ accumulation upon aging (22,23). A series of gel filtration fractions were collected according to molecular mass of the proteins. Then we determined the presence of PrP and A␤ in each fraction by Western blot analysis. As expected, no monomeric or oligomeric A␤ species were detected in the brain homogenates prepared from wild type mice around fraction 30 that contains large aggregates (Fig. 1A, highlighted by rectangle) or 2-month-old APP transgenic mice (not shown) in addition to APP migrating at ϳ110 kDa and FIGURE 1. Size distribution of PrP and A␤ from brains of APP transgenic mice. Brain homogenates from 18-month-old wild type or APP transgenic mice were subjected to gel filtration prior to Western blot analysis with antibody against either A␤ (6E10, panels A and C) or PrP (6D11, panels B and D). PrP and A␤ were examined in the fractions between 21 and 81. Panels A and B represent brain samples from wild type mice, whereas panels C and D represent brain samples from APP transgenic mice. Western blotting of A␤ probed with 6E10 shows that the wild type mice do not accumulate A␤ aggregates (A), whereas APP transgenic mice produce A␤ aggregates (C) in the same fractions as PrP aggregates (D). Western blotting of PrP detected with 6D11 shows that both wild type and APP transgenic mice generate large PrP aggregates (fractions 31-37) (B and D). In addition to monomeric (Mon.) and oligomeric (Oli.) A␤ species, APP (Ͼ98 kDa), the heavy chain (HC, ϳ50 -55 kDa), and light chain (LC, ϳ22-25 kDa) of IgG were also detected in the blots probed with 6E10 (A and C). The rectangles in blots A and C are used to highlight the blot area where monomeric and oligomeric A␤ bands migrate. IgG LC might be present in the PrP bands in the blots probed with 6D11 (B and D). In comparison to wild type mice, more IgG HC and LC were detected in APP transgenic mice. Up., PrP upper band; Mi., PrP middle band; and Lo., PrP lower band. nonspecific heavy (HC) and light (LC) chains of IgG. However, these mice accumulated high molecular mass PrP moieties with molecular mass ranging from 669 to 2,000 kDa (fractions 33 to 37), indicating the presence of iPrP (Fig. 1B).
We next analyzed the brain homogenates from 18-monthold APP transgenic mice, which are known to contain amyloid pathology in the brain (22,23). Consistent with this expectation, we detected monomeric and oligomeric A␤ species in fractions 33-37 that contain aggregates with molecular masses between 669 and 2,000 kDa (Fig. 1C, highlighted by rectangle), indicating the presence of fibrillar complexes. The monomeric and oligomeric A␤ species detected by Western blotting might be derived from larger A␤ aggregates denatured during sample preparation for Western blotting. As before, there were APP and nonspecific bands including IgG HC and LC above ϳ22 kDa (Fig. 1C). The A␤ aggregates are most likely comprised of A␤42. Interestingly, PrP aggregates were eluted in the same fractions (33-37) (Fig. 1D). Because these mice do not have a prion disease, the A␤42 aggregates eluted with iPrP ( Fig. 1, C and D), suggesting that iPrP may form a complex with A␤42 aggregates in the brains of APP transgenic mice. Alternatively, iPrP and A␤42 aggregates may have the same molecular mass without physical interaction. The bands above ϳ22 kDa mainly included APP, IgG HC, and IgG LC. These bands were frequently observed in animal samples and immunoprecipitates. Because these bands are not relevant to the current study, we focused on bands below IgG LC in most of the following figures to avoid distraction from our main interest, the complex of A␤ with PrP.
Co-capture of iPrP and A␤ in Mice-To determine whether iPrP and A␤ aggregates form complexes in vivo, we next conducted capture of iPrP by g5p in brain extracts from APP transgenic mice. The iPrP species associates with g5p and are captured by g5p-conjugated magnetic beads (17). If iPrP and A␤ form complexes in vivo, A␤ should be captured together with iPrP by g5p. The g5p capture was performed with brain extracts from APP transgenic mice prior to Western blotting. Blots were probed with anti-PrP or anti-A␤ antibody. The iPrP species were easily identified in both young (2-month-old) and old (18month-old) APP transgenic mice (Fig. 2, top). Additionally, a single intense band migrating at ϳ4 kDa, corresponding to monomeric (Mon.) A␤, and a faint smear migrating above 20 kDa, corresponding to oligomeric (Oli.) A␤ species, were also co-captured from brain extracts of aged but not young mice, consistent with the expected dynamics of A␤42 accumulation (Fig. 2, bottom). Because an excess of A␤42 over A␤40 is observed in aged APP transgenic mice (22), the A␤ co-captured by g5p along with iPrP in the aged APP transgenic mice could be A␤42 instead of A␤40.
Co-elution of iPrP and A␤ in Human AD Brains-Because of our findings in the brain of the APP transgenic mouse model of AD, we decided to confirm these findings in the brains of AD patients using gel filtration. First, we analyzed PrP and A␤ in brain homogenates from non-AD controls. As shown previously (17), brains from individuals free of prion infection accumulate iPrP (Fig. 3B). As expected, these brains did not contain A␤ aggregates (Fig. 3A). In contrast, monomeric and oligomeric A␤ species were detected in brain samples of AD patients in those fractions that contained large protein aggregates including iPrP (fractions 33-37) (Fig. 3, C and D). Because A␤ aggregates were only detected in AD, but not in non-AD brains, the co-purified A␤ could be A␤42 rather than A␤40. Thus, as in the AD mouse model, PrP and A␤42 aggregates may form complexes in the human AD brain.
A␤ Co-immunoprecipitates with HuPrP-After co-elution of PrP and A␤42 aggregates from human AD brains, we next asked whether huPrP binds to A␤ in AD brains by performing a co-immunoprecipitation assay. This method has been widely used to determine whether in vivo binding occurs between the protein of interest and other protein(s). To perform co-immunoprecipitation assays, control and AD brain homogenates were incubated with the 3F4 anti-PrP monoclonal antibody conjugated to beads and the subsequently eluted proteins were analyzed on Western blots probed with the anti-A␤ monoclonal antibody 6E10. The 3F4 immunoprecipitates contained A␤ in two AD cases, but no signal appeared in two non-AD control cases (Fig. 4A). Thus, A␤ is associated with huPrP aggregates only in AD patients, suggesting that certain A␤ conformers typical of AD preferentially bind to PrP. Because AD brains are characterized by increases in A␤42, again the A␤ conformer co-immunoprecipitated with PrP is most likely A␤42. Our finding using a peptide membrane array that affinity of A␤42 for PrP was higher than that of A␤40 further supports this possibility.
It has been demonstrated that there are soluble and insoluble forms of PrP and A␤ in human brains (1,17). To more precisely define whether specific A␤ isoforms are associated with specific PrP conformers, total AD brain homogenates (S1) were further separated into detergent-soluble (S2) and detergent-insoluble (P2) fractions by ultracentrifugation. The S2 and P2 fractions were immunoprecipitated using 3F4-conjugated beads prior to Western blotting with 6E10 and anti-N. As expected, huPrP was immunoprecipitated from both S2 and P2 (Fig. 4B, upper   FIGURE 2. Co-capture of iPrP and A␤ from AD mouse brains. Co-capture of iPrP and A␤ was conducted with brain homogenates from two 2-month-old and two 18-month-old APP transgenic mice. PrP was probed with 6D11 and was detectable in both young and old mice (upper panel). Monomeric (Mon.) and oligomeric (Oli.) A␤ species were probed with 6E10 and were detectable only in older mice (lower panel).
panel). Consistent with our previous observations (17), the insoluble PrP recovered in P2 displayed a dominant unglycosylated PrP band, whereas soluble PrP recovered in S2 exhibited a dominant diglycosylated PrP band. In contrast to PrP, A␤ was mainly detected in P2, although a faint band was detected in S2, accounting for ϳ5% of the total A␤ recovered (Fig. 4B, lower  panel). As a control, beads free of antibody did not precipitate PrP or A␤ (Fig. 4B). Therefore, more than 95% A␤ was recovered in the insoluble fraction where it mainly interacted with PrP isoforms with the molecular signature of iPrP. The remaining 5% A␤ was soluble and interacted with soluble PrP C . The co-immunoprecipitation of PrP and A␤ observed in both insoluble and soluble fractions indicates that the association between PrP and A␤ is specific.
Co-capture of iPrP and A␤42 in Human Brains-The finding that iPrP and A␤42 were co-captured by g5p from the brains of APP transgenic mice suggested that the insoluble moieties of huPrP and A␤ were the major species contributing to the complexes in vivo. To confirm this specific binding in the human brain homogenates, we performed g5p capture of iPrP. As expected, g5p captured huPrP isoforms that were predominantly unglycosylated (Fig. 5A, upper panel), consistent with the glycoform ratio of iPrP (17). In the preparation captured by g5p from brain homogenates of AD patients, we also detected a band migrating at ϳ4 kDa corresponding to A␤42 on Western blots probed with 6E10 (Fig. 5A, lower panel). In addition, a smear was also detected above 4 kDa, suggesting that A␤42 oligomers may be captured by g5p. Neither PrP nor A␤ were detected in the preparations captured by the beads free of g5p (Fig. 5A). Therefore, A␤42 indeed binds insoluble PrP in the brains of AD patients.
It is well known that A␤42 is prone to form soluble oligomers and insoluble aggregates (1). We next assessed whether the A␤42 captured by g5p was soluble or insoluble by fractionation of the AD brain homogenates as detailed above. The S1, S2, and P2 fractions were incubated with g5p for capture of iPrP prior to Western blotting with 6E10. As expected, no A␤42 was

. Co-immunoprecipitation of PrP and A␤ by 3F4 from AD brains.
A, preparations were obtained by immunoprecipitation with 3F4 from P2 fractions of two non-AD controls and two AD cases. A␤ was detected by Western blotting probed with 6E10 only in AD samples. B, preparations immunoprecipitated by 3F4 from S2 and P2 fractions from AD brains were probed with an antibody against the N terminus of PrP (Anti-N, upper panel) or anti-A␤ antibody 6E10 (lower panel). Most A␤ was detected in P2, whereas only a small amount was present in S2. No signal was observed in P2 incubated with empty beads conjugated with no 3F4 antibody. The nonspecific LC of IgG was also detected in the 6E10 blots migrating at ϳ29 kDa. detected in either fraction in non-AD control samples (Fig. 5B). In contrast, A␤42 was detected in S1 and P2 in the AD brain extracts, but was virtually undetectable in S2 (Fig. 5B). Therefore, the A␤ species that bind iPrP are present mainly in the detergent-insoluble fraction.
To determine whether A␤ capture by g5p is through direct binding of g5p to A␤ or through interaction between g5p and PrP, we conducted g5p capture of synthetic A␤42 that was spiked either in PrP-free lysis buffer or in the PrP-containing wild type mouse brain homogenate. When spiked in lysis buffer, A␤42 was not captured even up to 80 ng of A␤42 (Fig.  5C). In contrast, when spiked in the PrP-containing brain homogenate, A␤42 was captured even as low as 20 ng of A␤42 (Fig. 5D). These results indicated that g5p does not bind to A␤42 directly. As a result, A␤ capture by g5p most likely results from the PrP-A␤42 complex formed in the AD brain.
Identification of A␤42 and A␤40 Binding Sites on HuPrP-The above results clearly indicate that A␤42 binds to huPrP in the AD brain. To further dissect the A␤ binding sites on huPrP, we designed and synthesized a peptide membrane array containing 99 13-mer peptides that span the complete mature sequence of huPrP. Membrane-bound peptide arrays have been used previously to identify Tau-A␤ interaction sites (27). Using this method, we determined the A␤42 binding sites on human PrP in vitro. In addition, as a control, we also determined A␤40 binding sites on huPrP, to distinguish oligomer-from monomer-specific binding domains; A␤42 has been demonstrated to have a greater tendency to form oligomers than A␤40.
Prior to performing binding assays, we characterized the physicochemical and morphological properties of the A␤42 and A␤40 preparations used in peptide membrane arrays. To do this, both samples were subjected to SDS-PAGE and Western blotting with the 6E10 antibody. Whereas A␤40 produced a single band migrating at ϳ4 kDa, the A␤42 preparation exhibited not only a band migrating at ϳ4 kDa, but also a ladder containing multiple bands migrating above 4 kDa, a sign of oligomers and higher aggregates (Fig. 6A). By densitometric analysis, we analyzed the amount of A␤42 that migrated above 4 kDa on the Western blots. The oligomeric forms of A␤42 accounted for ϳ20% of total A␤42, whereas polymeric A␤40 was virtually undetectable by Western blotting (Ͻ1%) (Fig. 6B). These samples were also examined by electron microscopy. Negatively stained samples of A␤42 exhibited a mixture of small and mature fibrils 19 nm in diameter (Fig. 6C). In contrast, no fibrils were detected in the A␤40 sample; there were some amorphous structures (Fig. 6D). Thus, the A␤42 preparation contained a mixture of aggregates of different lengths that appear consistent with the pathological accumulation of A␤42 FIGURE 5. Co-capture of iPrP and A␤ by g5p from AD brains. A, brain homogenates from AD patients were captured by magnetic beads with (ϩ) or without (Ϫ) g5p prior to Western blotting with 3F4 (upper panel) or 6E10 (lower panel). Three PrP bands were detected by 3F4 with the electrophoretic profile of iPrP (dominant unglycosylated PrP). A band migrating at ϳ4 kDa and smear bands between 7 and 110 kDa were detected by the 6E10 antibody corresponding to monomeric (Mon.) and oligomeric (Oli.) A␤ species, respectively (lower panel). PrP and A␤ were not detected in the preparations captured by the beads conjugated with no g5p. B, S1, S2, and P2 fractions from non-AD and AD brains were subjected to g5p capture prior to Western blotting with 6E10. Monomeric (Mon.) and oligomeric (Oli.) A␤ species were captured by g5p in S1 and P2, but not in S2 of AD brain samples. In contrast, no A␤ was captured in non-AD brain samples. C, capture of synthetic A␤42 spiked in PrP-free lysis buffer (LB) by g5p. Different amounts of synthetic A␤42 spiked in lysis buffer were captured by g5p prior to Western blotting with 6E10. Synthetic A␤42 (40 ng) directly loaded onto the gel was used to as a control (Ϫ). No A␤ was captured by g5p up to 80 ng of A␤42. D, different amounts of synthetic A␤42 spiked in PrP-containing brain homogenates (BH) of wild type mice were captured by g5p prior to Western blotting with 6E10. A␤ was detected in the samples captured by g5p containing 20 ng of A␤42 or higher. FIGURE 6. Physicochemical and morphological features of the synthetic A␤42 and A␤40 used for peptide membrane arrays. A, the synthetic A␤42 and A␤40 peptides dissolved in PBS were subjected to SDS-PAGE and detected by Western blotting with 6E10. In the A␤42 preparation, a dominant band migrating at ϳ4 kDa corresponding to monomeric (Mon.) A␤42 and smear bands migrating Ͼ4 kDa corresponding to A␤42 oligomers (Oli.) were detected. In contrast, only a dominant band migrating at ϳ4 kDa corresponding to A␤40 was detected in the A␤40 preparation. B, quantitative analysis of the A␤ peptide bands on three Western blots at ϳ4 or Ͼ4 kDa up to 115 kDa by densitometric analysis. Approximately 20% of A␤42 formed polymers, whereas virtually no polymeric A␤40 was detectable by Western blotting (Ͻ1%). C and D, electron micrographs of negatively stained images of A␤42 (C) exhibit mature fibrils 19 nm in diameter (arrow head) in different lengths, whereas A␤40 (D) shows less organized structure. Scale bar, 100 nm.
in the brain, whereas A␤40 showed poor aggregation under the same condition.
Compared with A␤42, A␤40 reacted with fewer huPrP peptides (17 versus 11) and exhibited weaker signals with the positives (Fig. 7, B and C, and Tables 1 and 2). Five A␤42 binding areas (huPrP129 -141, 137-151, 173-187, 193-205, and 197-209) exhibited reactivity with both A␤ peptides albeit at lower intensity with A␤40 than with A␤42. The two A␤ peptides exhibited immunoreactivity with six PrP regions that had a slight variation in amino acid sequences between A␤42 and A␤40. However, six A␤42 binding areas, localized mostly in the N-terminal unstructured domain, had no reactivity with A␤40 (huPrP45-61, 53-65, 61-73, 83-95, 119 -137, and 151-165) (Fig. 7, B and C, and Tables 1 and 2), suggesting that the N-terminal PrP binding domain is highly A␤42-specific. The two types of A␤42 binding sites on PrP may result from the unique conformation of its oligomeric state.
We compared the binding affinity of A␤42-specific (only reacting with A␤42) and nonspecific (reacting with both A␤42 and A␤40) PrP peptides. Using synthetic A␤42 at 0, 0.1, 1.0, and 10.0 g/ml on the peptide membrane arrays, we examined the A␤ binding affinities of three A␤42-specific PrP peptides, including peptides 13 (PrP47-59), 16 (PrP53-65), and 33 (PrP87-99); as well as three A␤42 nonspecific peptides, including peptides 2 (PrP25-37), 8 (PrP37-49), and 39 (PrP99 -111) from the N-terminal domain of PrP. The intensity of A␤42 binding at different concentrations to PrP peptides was quantified by densitometric analysis. The resulting binding affinity curves were a function of the affinity of huPrP peptides for A␤42 (Fig. 8). Based on the curves, the K d50 value (A␤42 concentration at the half-maximal binding) was calculated for each PrP peptide. Notably, the mean K d50 of A␤42-specific PrP pep- Ninety-nine 13-mer PrP peptides, spanning the entire mature huPrP sequence (residues 23-231), were synthesized on the cellulose membranes. Immunoreactivity of 99 PrP peptides was probed with 4G8 after incubation with BSA (A), A␤42 (B), and A␤40 (C). BSA was used as a negative control. D, relative immunoreactivity of each huPrP 13-mer from 3 independent experiments. A␤42 is shown in blue and A␤40 in red. Upper panel, diagram of the NMR-derived secondary huPrP structure (28). According to the NMR study of recombinant huPrP, PrP contains an unstructured N-terminal domain from residues 23 to 124 (a wavy line) and a folded C-terminal domain extending from residues 125-228 (a straight line). The two ␤-strands are shown as blue arrows, located from residues 128 -131 and 161-163. The three ␣-helices are shown as red ovals, located from residues 144 -154, 173-194, to 200 -218. The octapeptide repeats are indicated.
tides was significantly greater than that of A␤-nonspecific peptides (0.287 versus 0.083 g/ml, p ϭ 0.0062 Ͻ 0.01) (Fig. 8), indicating that the affinity of A␤42 for its specific PrP binding areas was lower than that of A␤42 for its nonspecific binding areas.

DISCUSSION
The remarkable hypothesis that PrP C may act as a receptor for A␤42 and play a critical role in the pathogenesis of AD has triggered great interest in the fields of Alzheimer and prion diseases while raising high hopes for finding a cure for AD (3,9). The evidence supporting this new hypothesis is based on two fundamental observations: the direct binding of A␤42 to PrP and the potential pathophysiological role of PrP in A␤42 neurotoxicity in different AD models. Several groups have independently confirmed the binding of A␤42 to PrP, leaving no question as to the high specificity of this interaction (9,12,16). On the other hand, the pathophysiological role of PrP in A␤42 pathobiology seems more controversial (15,29). Three independent studies failed to replicate such a role for PrP in various AD models using Prnp Ϫ/Ϫ mice (12)(13)(14). These conflicting results may be largely attributable to differences in experimental conditions, including AD models and outcomes for A␤42 neurotoxicity (LTP, spatial memory, object recognition). Moreover, the use of Prnp Ϫ/Ϫ animals can lead to confounding effects, including compensatory changes where other protein(s) may replace the function of PrP C . Indeed, Laurén and co-workers (9) observed that PrP C was not the only cell-surface molecule that binds to A␤42 oligomers because it accounted for only ϳ50% of A␤42 binding. Thus, animals lacking PrP might not be the best models for assessing the role of PrP in the pathophysiology of A␤42. Our assumption is supported by the recent study of Wisniewski and co-workers (30) who observed that the antibody-based interruption of the binding of PrP to A␤42 by intraperitoneal injection of the 6D11 antibody improved behavior and memory in an AD animal model, which is consistent with our own preliminary data. 3 These last two studies agree with the observation by Laurén and co-workers (9) that the binding site of A␤42 is the 6D11 antibody epitope (murine PrP 93-109). Conceivably, interference of A␤42 binding to PrP may represent not only a more appropriate method to assess the role of PrP in the pathogenesis of AD, but also a therapeutic strategy for AD.
Yet, despite these promising findings, very little evidence emerged to demonstrate the relevance of these observations to humans. The motivation for our study was precisely to provide such evidence, specifically evidence for the binding of A␤42 to PrP in the human brain. We present seven novel observations. First, large PrP and A␤ particles are recovered in the same gel filtration fractions from AD brains of patients. Second, over 95% of A␤ co-immunoprecipitated by 3F4 from AD brain samples is insoluble, whereas less than 5% of A␤ is soluble. Third, A␤ is co-captured with iPrP by g5p from AD brains, in a PrPdependent format.  N-terminal domain, whereas only one is located in the folded C-terminal domain between residues 151 and 165. The other A␤42-specific binding area is located between the unfolded Nand folded C-terminal domains (residues 119 to 137). Sixth, compared with its nonspecific binding PrP areas, the affinity of A␤42 for its specific binding area is significantly lower. Finally, the oligomeric state or conformation of A␤42 and A␤40 may determine the affinity of the two A␤ peptides for huPrP. Cumulatively, these findings indicate that A␤42 may bind to huPrP in the AD brain and that there are two types of A␤ binding sites on huPrP: the oligomer-(A␤42-specific) and monomer-specific (A␤42-nonspecific) binding sites. Furthermore, our findings carry important implications regarding the pathophysiological consequences of A␤/PrP interaction.
Our study with 13-mer peptides confirmed the high affinity of A␤42 for different binding domains on huPrP. Moreover, the A␤42 preparation we used contained a mixture of monomers, oligomers, and fibrils, of which, oligomers and fibrils may have a much higher ability to interact with huPrP than the A␤40 preparation that contained mainly monomers but not oligomers and fibrils (Figs. 7 and 8). Based on the different affinity of huPrP peptides for the two A␤ peptides with distinct oligomeric states, the huPrP domains can be grouped into two types of A␤ binding sites (Tables 1 and 2). The first group specifically binds to A␤42, which may be oligomer-specific. These binding sites are mainly distributed in the N-terminal unstructured region. For instance, four of six A␤42-specific binding sites (binding areas III through VI) are localized in the octapeptide repeat region of the unstructured N-terminal domain (Tables 1  and 2), including sites identified by Laurén and co-workers (9). In particular, the HGGGWGQPHGGGW peptide, which is repeated three times in the octapeptide domain, revealed  The binding affinity of synthetic A␤42 for three A␤42-specific (reacting with A␤42 only including peptides 13, 16, and 33) and three A␤42 nonspecific (reacting with both A␤42 and A␤40 including peptides 2, 8, and 39) PrP peptides at the N-terminal domain was measured by peptide membrane arrays. The intensity of A␤42 at 0, 0.1, 1.0, and 10.0 g/ml bound to PrP peptides were quantified by densitometric analysis. The binding affinity curves were a function of the affinity of huPrP peptides for A␤42. The K d50 value for each PrP peptide was calculated based on the curves. The mean K d50 of A␤42-specific PrP peptides is significantly greater than that of A␤42-nonspecific PrP peptides (0.287 versus 0.083 g/ml, p ϭ 0.0062 Ͻ 0.01). strong binding activity to A␤42. The octapeptide domain (residues 51-91) is comprised of four octapeptides and one nonapeptide. This domain contains at least four copper binding sites and may be involved in copper metabolism and transport to synapses as well as cellular defense mechanisms against oxidative stress (31)(32)(33)(34). It is possible that the binding of A␤42 to this domain interrupts the physiological interaction of PrP with copper, resulting in disturbed neuronal communication. In contrast to the observations of Laurén and co-workers (9), we also identified two additional A␤42-specific binding sites. One (binding area VIII) was detected in the conjunction area (PrP119 -137) between the N-terminal unfolded and the C-terminal folded domains (Tables 1 and 2). The other (binding area XII) is localized in the folded C-terminal domain. This second group binds indiscriminately to both A␤42 and A␤40, which may be monomer-specific and are mainly localized in the extremely N-terminal domain and C-terminal folded domain. The lower affinity of A␤42 for the octapeptide repeat domain suggests that the PrP physiological function may also be affected when A␤42 is significantly increased in AD.
However, these A␤ binding sites on huPrP identified in vitro may not faithfully reflect the behavior of brain-derived PrP; therefore, this observation should be established and validated by in vivo studies, especially given that the glycosylation and folding of the C-terminal region of PrP may affect the corresponding binding sites. Nevertheless, the results obtained with short synthetic peptides might still be relevant given the presence of several endogenously truncated PrP fragments in vivo including N1, N2, C1, and C2 (21,35,36). It would be interesting to determine whether these endogenous small PrP fragments participate in interactions with A␤42 in vivo. Moreover, the identified broadest PrP area to which both A␤42 and A␤40 bind including peptides 34 -39 (Tables 1 and 2 and Fig. 7) contains the binding site identified by two previous independent studies using different methods (9,16), indicating that this PrP area may be the key A␤42 binding site. In addition, identification of the A␤42 binding areas on huPrP may be significant for the development of new therapeutic strategies for AD. For instance, the intracerebral application of small PrP peptides may inhibit interaction between PrP and A␤42, thereby preventing A␤42 neurotoxicity.
PrP is a structurally dynamic protein, existing in chameleonlike conformations (37,38), and the diverse conformations of recombinant PrP may represent an intrinsic molecular spectrum of PrP C in vivo (39). Remarkably, small amounts of insoluble PrP oligomers and polymers are also observed in uninfected human and animal brains (17). These newly identified PrP conformers, which we termed iPrP, account for ϳ5-25% of total PrP, including full-length and N-terminal truncated forms. The insoluble PrP can bind to g5p, a single-stranded DNA-binding protein that specifically binds to PrP Sc or PrP Sclike conformers (17,18), suggesting that iPrP possesses a conformation different from that of PrP C but similar to that of PrP Sc . However, the physiological or pathophysiological role of iPrP is currently unclear.
Because of the specific or direct binding of A␤42 to PrP indicated by previous studies (9, 12, 16), as well as our peptide membrane array and co-immunoprecipitation of soluble PrP and A␤ here, it is conceivable that PrP and A␤42 may bind directly to each other within insoluble complexes. Therefore, it is most likely that the iPrP is the PrP conformer that interacts with the amyloid plaques in vivo. Our finding that A␤42 binds to iPrP suggests that iPrP plays a role in the fibrillization of A␤42. In fact, it has been consistently observed that PrP deposits are observed around or throughout A␤ plaques in AD brains (5)(6)(7). Interestingly, insoluble PrP Sc aggregates also seem to interact with A␤42 in vivo (40). Soto and co-workers (40) recently reported that an increase in the efficiency of A␤42 aggregation in vitro was dependent on PrP Sc dosage. Similarly, AD mice developed a strikingly higher load of cerebral amyloid plaques that appeared much faster in prion-infected than in uninfected mice. Thus, iPrP (the PrP Sc -like forms identified in uninfected human brains) may facilitate fibrillization of A␤42 in AD. Indeed, synergistic interactions between amyloidogenic proteins associated with neurodegeneration have been demonstrated to promote each others fibrillization, amyloid deposition, and formation of filamentous inclusions in transgenic mice (8,41). Because this is the case, the possibility must be considered that a significant increase in the total number of A␤ plaques observed in bigenic mice overexpressing PrP (8) might result from an increase in the formation of iPrP. Moreover, because iPrP interacts with insoluble A␤42, whereas soluble PrP C binds soluble A␤42 in vivo, it is possible that distinct PrP conformers binding to different A␤42 species thereby function either as receptors for soluble A␤42 oligomers or as modulators of insoluble A␤42 deposition. This hypothesis could be tested by intracerebrally injecting anti-PrP antibodies against either soluble or insoluble PrP species in AD animal models.
The insoluble PrP may also be involved in memory and cognitive processes. It is worth noting that the involvement of protein polymerization in long-term memory has been documented in the last few years. For instance, a neuronal isoform of the cytoplasmic polyadenylation element-binding protein, which plays a role in long-term memory by activating translationally dormant mRNA to regulate protein synthesis at the synapse, has been reported to exhibit self-perpetuating prionlike properties in sensory neurons and in yeast (42)(43)(44). These studies suggest that prion-like conformational changes may constitute a key event in the maintenance of structural synaptic changes required for long-term memory by acting as molecular "switches" (45,46). It was proposed that the impact of a putative PrP conformation, rather than the pathological PrP Sc , on memory in healthy humans is associated with physiologically occurring conformational changes (47)(48)(49). However, whether these changes in the accumulation of iPrP and their ability to seed A␤42 are responsible for memory impairment remains to be investigated.