A d-enantiomeric peptide interferes with heteroassociation of amyloid-β oligomers and prion protein

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that affects millions of people worldwide. One AD hallmark is the aggregation of β-amyloid (Aβ) into soluble oligomers and insoluble fibrils. Several studies have reported that oligomers rather than fibrils are the most toxic species in AD progression. Aβ oligomers bind with high affinity to membrane-associated prion protein (PrP), leading to toxic signaling across the cell membrane, which makes the Aβ–PrP interaction an attractive therapeutic target. Here, probing this interaction in more detail, we found that both full-length, soluble human (hu) PrP(23–230) and huPrP(23–144), lacking the globular C-terminal domain, bind to Aβ oligomers to form large complexes above the megadalton size range. Following purification by sucrose density–gradient ultracentrifugation, the Aβ and huPrP contents in these heteroassemblies were quantified by reversed-phase HPLC. The Aβ:PrP molar ratio in these assemblies exhibited some limited variation depending on the molar ratio of the initial mixture. Specifically, a molar ratio of about four Aβ to one huPrP in the presence of an excess of huPrP(23–230) or huPrP(23–144) suggested that four Aβ units are required to form one huPrP-binding site. Of note, an Aβ-binding all-d-enantiomeric peptide, RD2D3, competed with huPrP for Aβ oligomers and interfered with Aβ–PrP heteroassembly in a concentration-dependent manner. Our results highlight the importance of multivalent epitopes on Aβ oligomers for Aβ–PrP interactions and have yielded an all-d-peptide–based, therapeutically promising agent that competes with PrP for these interactions.

the accumulation of extracellular neuritic plaques consisting mainly of fibrillar ␤-amyloid (A␤) peptide (1). Initially, these plaques were thought to be the toxic species in AD, but several lines of evidence now indicate that the levels of soluble oligomeric forms of A␤ (A␤ oligo ) correlate best with the neurotoxic effects observed during AD (2,3).
Many A␤ receptors have been described to date (4). One of them is the cellular prion protein (PrP C ), which binds A␤ oligo with nanomolar affinity (5)(6)(7)(8)(9)(10). PrP C is a glycosylphosphatidylinositol (GPI)-anchored glycoprotein highly expressed in the brain. PrP C itself can misfold into the scrapie isoform PrP Sc sporadically or after infection, leading to neuronal damage and disease such as the transmissible spongiform encephalopathies (11). The interaction of A␤ oligo with PrP C bound to the metabotropic glutamate receptor 5 leads to toxic signaling across the cell membrane by activating intracellular Fyn kinase (12,13). Fyn kinase phosphorylates N-methyl-D-aspartate (NMDA) receptors (14,15) and alters NMDA receptor localization, leading to destabilization of dendritic spines (12). Furthermore, Fyn kinase hyperphosphorylates the tau protein, which assembles into neurofibrillary tangles, a further hallmark of AD (16). Hyperphosphorylation of tau depends on the A␤-PrP interaction (17). Therefore, understanding the A␤-PrP pathway will open new therapeutic strategies by targeting the A␤-PrP interaction (18).
The binding regions of A␤ oligo have been mapped to residues 23-27 and ϳ95-110 in the N-terminal part of PrP (5, 19 -22) (see Fig. 1A). Soluble N-terminal PrP fragments inhibit the assembly of A␤ into amyloid fibrils and block neurotoxic effects of soluble oligomers (20,23), presumably by competing with membrane-anchored PrP C for A␤ oligo . This competition might also explain the suggested neuroprotective function of the naturally produced soluble N1 fragment (amino acids 23-110/111) of PrP (24), which contains both A␤ oligo -binding regions. The binding regions on A␤, however, have not been identified so far and might constitute a specific conformational epitope of A␤ oligo (21). All these data show that the A␤-PrP interaction is a promising point of intervention to prevent the toxic signaling in AD.
In the past years, we have identified a number of D-enantiomeric peptides as promising drug candidates for direct elimination of A␤   oligo (25)(26)(27)(28)(29)(30). The advantage of D-peptides over L-peptides is their higher protease resistance, resulting in slower degradation and longer half-life (31,32). For A␤(1-42)directed D-peptides, high stability in media simulating the route of orally administered drugs (33) and enhanced proteolytic stability in murine plasma and organ homogenates were shown (34). The lead compound of these D-peptides, D3, had been selected by mirror-image phage display (26, 35). D3 and its tandem version, D3D3, convert toxic A␤ species into nontoxic species (25,28). Treatment with D3 reduces the number of amyloid plaques (26) and improves cognition in transgenic AD mice (28). One derivative of D3 called RD2 shows enhanced binding to A␤ (36, 37), and both RD2 and D3 have demonstrated desirable pharmacokinetic properties (29,38). A further promising derivative is the D-peptide RD2D3, a head-to-tail tandem combination of RD2 and D3 (30,34,39). RD2D3 binds A␤(1-42) with a K D of 486 Ϯ 69 nM (39). All of these therapeutically promising D-peptides contain a high fraction of basic residues, which is reminiscent of the proposed binding sites for A␤ on PrP (5, 19 -21). Therefore, the A␤-binding D-peptides might be suitable compounds for interference with the A␤-PrP interaction by competing with PrP for A␤   oligo .
Recently, we introduced the "quantitative determination of interference with A␤ aggregate size distribution" (QIAD) assay, which allows the determination of a compound's efficacy to eliminate A␤   oligo (25). This assay enables the separation of A␤ assemblies by density-gradient ultracentrifugation (DGC) and the quantification of these assemblies by UV-detected reversed-phase (RP)-HPLC. For the present study, we have refined the QIAD assay to achieve simultaneous quantification of A␤ , recombinant anchorless human PrP (huPrP) constructs, and D-peptides in a single RP-HPLC run to (i) characterize the A␤-huPrP interaction in detail and (ii) evaluate the influence of the tandem D-peptide RD2D3 on this interaction. We investigated four different huPrP protein constructs, namely huPrP , huPrP , huPrP(90 -230), and huPrP(121-230) (see Fig. 1A), and added them in different concentrations to preformed A␤  oligo . In the case of huPrP  and huPrP , this resulted in high-molecular-weight A␤(1-42) oligo -huPrP complexes. The separation of these complexes from A␤  or huPrP monomers and A␤(1-42) oligo by sucrose DGC and subsequent RP-HPLC analytics (see Fig. 1B) allowed the determination of molar ratios between A␤  and huPrP within the complexes. We show that these ratios are dependent on the concentration of huPrP added. We imaged A␤(1-42) oligo -huPrP  complexes by atomic force microscopy (AFM) and observed a correlation between the applied huPrP  concentration and the size and morphology of the heteroassemblies. We analyzed the influence of the D-peptide RD2D3 on the A␤(1-42) oligo -huPrP  interaction by determining its effect on the A␤ :huPrP ratio within the assemblies. We show that RD2D3 competes with the A␤(1-42) oligo -huPrP  interaction and might thus be a potential therapeutic agent.
Before investigating the A␤(1-42) oligo -huPrP interaction, we checked the binding partners separately in their purified states to confirm that they remain soluble at the chosen buffer conditions, which were a compromise between neutral pH and conditions required for stability of A␤ oligomers and detergent-free solubility of huPrP constructs favoring absence of phosphate and low salt. This is of note as all huPrP protein constructs (40 -43) as well as A␤ (44,45) are able to convert into fibrils under certain conditions, and such a conversion may hamper the analysis of A␤(1-42) oligo -huPrP complexes. We performed CD spectroscopy analysis of huPrP(23-144), huPrP , and huPrP(90 -230); solution NMR spectroscopy of huPrP  and huPrP ; and AFM measurements of huPrP .
The solubility and overall conformational properties of huPrP  and huPrP  were confirmed in more detail by solution NMR spectroscopy. The solution structure of huPrP had originally been determined in acetate buffer at pH 4.5 and 20°C (47) and found to comprise a highly disordered N-terminal region (residues 23-124) followed by a globular C-terminal domain (residues 125-228) featuring three long ␣helices and a relatively small two-stranded antiparallel ␤-sheet.

Interference with A␤-PrP complex formation
Under these buffer conditions, we indeed obtained well dispersed solution NMR spectra of excellent quality for huPrP   (Fig. S3A) with sharp narrow line shapes and chemical shifts similar to those reported in the literature (47), thereby demonstrating that the protein is soluble and natively folded. huPrP(23-144) also exhibits high-quality solution NMR spectra under these buffer conditions (Fig. S3A). As expected for huPrP , only backbone amide resonances in the random-coil region ( 1 H chemical shifts between about 8.0 and 8.6 ppm (48)) were observed (Fig. S3A), suggesting that not only  (5, 19 -22) are marked in blue, and the corresponding sequence is shown in the huPrP(23-230) construct with basic amino acid residues highlighted in red. OR marks the octarepeat region from residues 51 to 91. huPrP , huPrP(90 -230), and huPrP(121-230) contain a disulfide bond between Cys 179 and Cys 214 in the structured C-terminal part of the protein. B, 80 M A␤(1-42) was incubated for 2 h to obtain A␤(1-42) oligo before different quantities of the respective prion protein were added to the sample. After 30-min coincubation, the sample was separated by sucrose DGC and fractionated. Each fraction was analyzed by SDS-PAGE as well as by quantitative RP-HPLC. show predominantly ␣-helical CD spectra, whereas the N-terminal huPrP(23-144) (A, red) is present in random-coil conformation (MRE, mean residue ellipticity; deg, degrees). Shown are 1-m 2 AFM images of 200 nM huPrP(23-144) (B) or 800 nM A␤(1-42) oligo (C). Scale bars, 200 nm. huPrP(23-144) forms a thin film on the mica surface, not higher than 1-2 nm (B). The generated A␤(1-42) oligo species are seen as spherical particles with heights ranging from 1 to 6 nm (C).

Interference with A␤-PrP complex formation
the N-terminal region from residues 23 to 124 is highly disordered but that also residues 125-144 are disordered in huPrP . To build a bridge between the quality control of the huPrP samples done at pH 4.5 and the solution conditions used for the interaction studies done at pH 7, we investigated the pH dependence of the overall conformational properties of huPrP(23-144) by solution NMR spectroscopy in a series of pH steps from 4.5 to 7.0. Although the shift in protonation equilibrium of the seven histidine side chains upon increasing the pH was associated with readily identifiable chemical shift changes for several backbone amide resonances, the quality and overall appearance of the solution NMR spectra of huPrP , including the limited resonance dispersion indicative of a disordered conformation, remained very similar over this pH range (Fig. S3B). Over the course of several days to weeks, the NMR samples did not show obvious signs of any significant formation of visible precipitate, any deterioration, or signal loss of the solution NMR spectra. To test for any sample degradation or aggregation in a more quantitative fashion, we monitored the intensity of 58 sufficiently well resolved amide resonances of a sample of 89 M [U-15 N]huPrP(23-144) in 50 mM Tris-HCl, pH 7.2, in 10% D 2 O in a series of 1 H, 15 N heteronuclear single quantum coherence spectra recorded at 600 MHz at 5.0°C, but no change in resonance intensity exceeding even a fraction of a percent was observed over the monitoring period of more than 48 h (Fig. S3D). Taken together, these NMR spectroscopic results show that huPrP(23-144) is readily soluble up to concentrations of about 0.3 mM, is highly disordered in solution under mildly acidic to neutral buffer conditions, and remains soluble and disordered for at least several days at the conditions tested.
A␤  oligo was prepared by incubating monomeric A␤  in buffer at pH 7.4 for 2 h at 22°C under agitation. AFM of the A␤(1-42) oligo samples showed small spherical particles with heights of 1-6 nm, rarely up to 10 nm (Figs. 2C and S4) and confirmed that the chosen incubation conditions produce high amounts of A␤(1-42) oligo without formation of A␤(1-42) fibrils or larger aggregates.
Our assay for studying the A␤(1-42) oligo -huPrP interaction is based on the QIAD protocol (25). In the present work, this assay includes the separation of a sample containing A␤  assemblies and/or huPrP by sucrose DGC followed by qualitative and quantitative analyses of the interaction partners by SDS-PAGE and RP-HPLC (Fig. 1B). For calibration of the sucrose gradient, standard proteins ranging from 43 to 669 kDa were used (Fig. S5). Thyroglobulin, the reference protein with the highest molecular mass (669 kDa), was found in fractions 7-10, indicating that proteins, complexes, or aggregates that can be found in higher (and thus denser) fractions (fractions 11-14) must have molecular masses in the megadalton range or larger.
Initially, A␤(1-42) oligo and huPrP were separately analyzed by sucrose DGC. Either 80 M of preincubated A␤(1-42) or a 10 or 20 M concentration of the respective huPrP protein construct was applied on a sucrose gradient and centrifuged for 3 h. Silver-stained Tris/Tricine SDS-PAGE gels as well as RP-HPLC quantification of all gradient fractions revealed the distribution of the proteins among the fractions and hence among different assembly states (Figs. 3 and S6A). The preincubated A␤  sample showed a broad distribution of A␤(1-42) within the sucrose gradient (Fig. 3, A and E), containing mainly A␤(1-42) oligo (fractions 3-7) as well as residual monomeric A␤(1-42) (fractions 1 and 2) as we have established previously (25). The highest concentrations of huPrP , huPrP(23-144), huPrP(90 -230), and huPrP(121-230) were found in fractions 1-3, confirming that huPrP occurs predominantly in a soluble monomeric state. In addition, minor amounts of huPrP were present in fractions 4 -6 and 11-14, the latter representing high-molecular-weight aggregates. AFM and CD spectroscopy (Fig. 2) suggest that these were nonfibrillar, amorphous aggregates.

A␤(1-42) oligo forms high-molecular-weight heteroassemblies with huPrP(23-230) as well as with huPrP(23-144)
After confirmation of the soluble, nonfibrillar state of all huPrP constructs as well as the size distribution of preincu- . This clearly confirms direct interaction between huPrP(23-230) and A␤(1-42) oligo . huPrP  interaction was preferential with A␤(1-42) oligo as the concentration of A␤(1-42) monomers in fractions 1 and 2 decreased only slightly with increasing huPrP(23-230) concentration. Moreover, instead of simply forming one-to-one complexes, which would be found not far away from fractions 3-7, huPrP(23-230) and A␤(1-42) oligo form large supramolecular heteroassemblies, which are located in fractions 11-14 and hence have molecular masses in or above the megadalton range (according to the calibration of the density gradient; Fig. S5).
We verified that the A␤:PrP ratio of ϳ4 is constant when A␤(1-42) oligo is saturated with huPrP by adding different final A␤(1-42) oligo concentrations of 20, 40, 60, and 80 M to a saturating concentration (40 M) of huPrP . At all A␤(1-42) oligo concentrations, the A␤:PrP ratio in the heteroassociates was ϳ4 with deviations within the experimental error ( Table 1).

The morphology of A␤(1-42) oligo -huPrP(23-144) heteroassemblies depends on the huPrP concentration
For structural characterization of the A␤(1-42) oligo -huPrP complexes, we focused on the N-terminal huPrP(23-144) construct as the PrP C terminus is not required for complex formation. Heteroassemblies formed in the presence of different huPrP(23-144) concentrations ranging from 1 to 40 M were

huPrP ratios within A␤(1-42) oligo -huPrP heteroassemblies
The ratios were calculated for every experiment as the quotient of the total amount of A␤(1-42) molecules and the total amount of huPrP in fractions 11-14 after sucrose DGC. For the calculation of A␤ :huPrP(23-230) ratios the huPrP(23-230) concentrations in fractions 11-14 found with huPrP(23-230) alone (Fig. 3, B and F) were not considered as they were negligibly small. For concentrations labeled with "ND," molar ratios were not determined for huPrP . Comparing the same (huPrP) concentration, similar ratios were obtained for both huPrP  and huPrP . Full saturation of the PrP-binding capacity of A␤

Interference with A␤-PrP complex formation
analyzed by AFM. Unbound A␤  or huPrP(23-144) was removed by centrifugation and repeated washing steps followed by imaging of the heteroassemblies on mica in air using intermittent contact mode. Imaging was particularly challenging because the assemblies were several micrometers high with deep holes and high stickiness, which led to rapid contamination of cantilever tips. To display the different features of the samples in more detail, images were taken of the outer surfaces of the heteroassemblies (Fig. 8) several hundred nanometers above the mica support. Heteroassemblies prepared at an increased huPrP(23-144) concentration of 5 M (Fig. 8C) were up to 500 nm high and had a more compact appearance, suggesting a tighter interaction between the subassemblies. The surface of the globular subassemblies seemed to be smoother in this case. When the huPrP(23-144) concentration for heteroassociation was further raised to 40 M (Fig. 8D), the resulting heteroassemblies were up to 1 m high, exhibiting globular subassemblies with smooth surface appearance and unresolved substructure. In all cases, A␤(1-42) oligo -huPrP(23-144) heteroassemblies (Figs. 8, B-D) were much larger than A␤(1-42) oligo (Fig. 8A), demonstrating that the heteroassemblies contain multiple copies of A␤(1-42) oligo .
We have previously shown that an average A␤(1-42) oligo consists of about 23 monomeric units (25). Combining this

Interference with A␤-PrP complex formation
finding with the A␤:PrP ratios determined here, the heteroassemblies contain approximately six huPrP(23-230) or huPrP(23-144) molecules on average per A␤(1-42) oligo in the presence of an excess of huPrP. Therefore, a simplified model of A␤(1-42) oligo -huPrP assemblies can be drawn (Fig. 8, E and F). This model also considers the potential of huPrP to cross-link A␤(1-42) oligo via its two basic N-terminal binding sites (residues 23-27 and ϳ95-110), which of course cannot be formed by huPrP(90 -230) or huPrP(121-230). Such cross-links may play a crucial role in A␤(1-42) oligo -huPrP assembly due to the multivalent presentation of epitopes on A␤(1-42) oligo .

The D-enantiomeric peptide RD2D3-FITC competes with huPrP for binding to A␤(1-42) oligo
Soluble N-terminal PrP fragments, including the naturally produced neuroprotective N1 fragment, block neurotoxic effects of soluble oligomers (20,23,24), presumably by competing with membrane-anchored PrP C for A␤ oligo . In line with this, we found that huPrP(23-144) rescues the viability of PC-12 cells from A␤(1-42) oligo -induced toxicity in a concentrationdependent manner according to the 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide (MTT) cell viability test (Fig. S9). Compounds that compete with membrane-anchored PrP C for A␤ oligo in a similar fashion might therefore be of therapeutic interest. Similarly to huPrP, our well characterized A␤-binding D-peptides form heteroassemblies with A␤(1-42) oligo (28). These specific D-peptides contain a high ratio of basic amino acid residues and are in that respect similar to the A␤(1-42) oligo -binding sites within the PrP N terminus (5, 19 -22). Similar to the soluble N-terminal PrP fragment huPrP , the D-peptide RD2D3 shows rescue of PC-12 cell viability from A␤(1-42) oligo -induced toxicity in the MTT test ( Fig. S9 and Ref. 30). Therefore, the D-peptides might act by a similar mechanism as N-terminal huPrP fragments, i.e. competition with membrane-anchored PrP C for A␤ oligo . To investigate this hypothesis, we analyzed the effect of the D-peptide RD2D3, labeled with a fluorescent dye (FITC), on the A␤(1-42) oligo -huPrP(23-144) interaction.

Discussion
In 2009, Laurén et al. (5) reported that oligomeric A␤ binds to membrane-anchored PrP C , leading to toxic signaling across the cell membrane. Although subsequent studies questioned the role of PrP C in toxic signaling (7, 49 -51), further evidence was gained in support of the original findings (5,12,13,20). According to the current view of PrP C -A␤ oligo -induced signaling, metabotropic glutamate receptor 5 interacts with PrP C and activates the intracellular Fyn kinase when A␤ oligomers are bound to membrane-anchored PrP C (12,13). This activation leads to hyperphosphorylation of tau protein as well as to phosphorylation of NMDA receptors, two mechanisms that finally lead to neuronal damage (12,14,15,17). The elucidation of these mechanisms has opened new strategies to prevent toxic signaling in AD by targeting one of these proteins or mediators.
The A␤(1-42) oligo -PrP interaction is at the core of the PrPmediated toxic signaling cascade. Here, we have characterized the A␤(1-42) oligo -PrPinteractionbyapplyingasetofsolublehuPrPconstructs and by taking advantage of the QIAD assay (25), which enables determination of the size distribution of A␤ assembly species and their complexes based on separation by DGC. We found that A␤(1-42) oligo and huPrP readily associate to form heteroassemblies above the megadalton range (Figs. 4-6 and 8). The heteroassemblies were imaged by AFM as m-sized clusters of nm-sized globular substructures (Fig. 8).
Heteroassociation is greatly impaired for the huPrP variants devoid of the N terminus, huPrP(90 -230) and huPrP(121-230) (Figs. 7 and S6), in agreement with the notion that both A␤-binding sites in the huPrP N terminus (residues 23-27 and ϳ95-110 (5, 19 -22)) are required for high-affinity interaction. This is in line with recent reports showing that the effect of soluble, anchorless PrP(90 -231) with respect to prevention of A␤-mediated cytotoxicity was substantially weaker compared with full-length huPrP or N-terminal huPrP (52). In addition, Nieznanski et al. (23) showed that about 10-fold higher concentrations of huPrP(90 -231) than of huPrP(23-231) or huPrP(23-144) were required to achieve comparable inhibitory effects on A␤(1-42) fibril formation. Similarly, a complete loss of binding capacity to A␤(1-42) oligo after deletion of the N-terminal region 23-89 was observed (19). The A␤-binding sites in huPrP constitute positively charged patches, suggesting that an electrostatic component may promote the interaction. In this context, it is worth noting that the presence of negatively charged patches on A␤(1-42) oligo has been inferred from engineering of A␤(1-42) oligo -binding proteins (53).
Further distinctive features of the A␤(1-42) oligo -huPrP heteroassociation comprise (i) disordered conformation of the binding sites in free PrP, (ii) multivalency (an average A␤(1-42) oligo can interact with six huPrP molecules), and (iii) a stoichiometry that is not fixed but constrained to a relatively narrow window (the A␤:PrP ratio is in the range of 4:1-12:1). We

Interference with A␤-PrP complex formation
searched the literature for protein-protein interactions with similar characteristics and found two notable cases, the interaction of nucleophosmin with nucleophosmin-binding proteins (54) and heteroprotein coacervation of ␤-lactoglobulin and lactoferrin (55,56). The interaction of nucleophosmin with binding proteins containing arginine-rich linear motifs is involved in nucleolus formation by liquid-liquid phase separation. The interaction features an electrostatic component, intrinsic disorder in the free binding motifs, as well as multivalency: acidic oligomers of nucleophosmin interact with proteins that contain at least two basic linear motifs (54). Heteroprotein coacervation of ␤-lactoglobulin and lactoferrin features a constrained stoichiometry with some variation depending on the molar ratio of the initial mixture (55,56).
The molar ratios in the A␤(1-42) oligo -huPrP heteroassemblies suggest that an average A␤(1-42) oligo can directly interact with up to six huPrP molecules. This multivalent interaction, established here for soluble huPrP constructs, may also have consequences for GPI-anchored PrP C . For example, clustering of PrP C can promote the activation of Fyn kinase (57,58). Moreover, the multivalency of A␤(1-42) oligo may contribute to the formation of ternary complexes with other membrane receptors (59). N1, a secreted, soluble N-terminal fragment resulting from ␣-cleavage of huPrP, comprises both A␤(1-42) oligo -binding sites and is therefore primed for heteroassociation with A␤(1-42) oligo . Intriguingly, N1 is neuroprotective, inhibits A␤(1-42) oligo -mediated neurotoxicity (20), and forms coaggregates with A␤ that have been detected in post-mortem brain tissue (60).
As the A␤-PrP interaction might be a possible therapeutic target in treating Alzheimer's disease, there is great effort to identify either A␤or PrP-binding compounds that inhibit the A␤-PrP C interaction. For example, dextran sulfate sodium (61) and Chicago Sky Blue 6B (62) inhibit binding of A␤(1-42) oligo to PrP. Similarly, anti-PrP antibodies blocked oligomeric A␤ binding to PrP and prevented A␤ oligomer-induced neurotoxicity (5,(63)(64)(65). The QIAD assay in its version introduced here, permitting simultaneous quantification of A␤(1-42), huPrP, and compound, allows identification and characterization of a compound's interference with the A␤-PrP interaction. We found that the A␤:PrP ratio in the heteroassociates (Fig. 10) is a suitable indicator of a compound's competition with PrP for A␤(1-42) oligo .
We have previously identified a number of D-enantiomeric peptides as promising drugs that eliminate A␤(1-42) oligo and improve cognition in transgenic AD mice (25)(26)(27)66). Here, we have observed that the D-peptide RD2D3 interferes with the binding of huPrP  to A␤(1-42) oligo . As a rescue of cell viability of A␤-treated cells (Fig. S9 and Ref. 30) and enhancement of cognition (39) were shown for RD2D3, our studies suggest that interference with the A␤-PrP interaction might be one potential mechanism of action of this class of D-peptides.

Interference with A␤-PrP complex formation
natant was separated by size exclusion chromatography on a HiLoad 26/60 Superdex 200 preparative grade column. Analytical samples of every second elution fraction were precipitated with 20% (w/v) TCA to remove the guanidinium HCl and investigated by SDS-PAGE. huPrP(23-230)-or huPrP(90 -230)containing fractions were pooled and purified by RP-HPLC. A semipreparative C 8 column (Zorbax 300 SB-C8 semipreparative, 9.4 ϫ 250 mm, Agilent) allowed the separation of huPrPs from impurities using a 20 -30% (v/v) gradient of acetonitrile ϩ 0.1% (v/v) TFA within 15 min followed by a 10-min step of isocratic 30% (v/v) acetonitrile ϩ 0.1% (v/v) TFA. The purifications were performed at 80°C at a flow rate of 4 ml/min. The elution fractions containing huPrP were pooled, dried by lyophilization, finally dissolved in Milli-Q water and adjusted to concentrations ranging from 96 to 140 M. Stocks of 100 -200 l were frozen in liquid nitrogen and stored at Ϫ80°C. We chose water for the preparation of the huPrP stock solutions as huPrP is highly soluble in water.
huPrP  was cloned in a pET302/NT-His vector and transformed in E. coli BL21(DE3). This huPrP construct also contains the natural polymorphism Met-129. Cells were grown in LB medium at 37°C and 160 rpm shaking and incubated overnight after induction at these conditions. For the preparation of isotope-labeled [U-15 N]huPrP  or [U-13 C, 15 N]huPrP(23-144), M9 minimal medium containing the desired isotopes was used. Resuspension and disruption of the cells were performed as described above. The insoluble inclusion bodies were dissolved in 10 ml of 6 M guanidinium HCl, 100 mM NaCl, 30 mM Tris-HCl, pH 7.4, and centrifuged (see above). The supernatant was separated by IMAC with two serially connected 5-ml Protino nickel-nitrilotriacetic acid columns (Macherey-Nagel). A washing step with 30 mM Tris-HCl, pH 7.4, allowed the removal of the denaturing agent guanidinium HCl. The elution occurred with a linear gradient from 0 to 500 mM imidazole, 30 mM Tris-HCl, pH 7.4. The huPrP(23-144)-containing fractions (verified by SDS-PAGE) were pooled, and the hexahistidine tag was cleaved by tobacco etch virus protease. RP-HPLC purification, lyophilization, and storage of the protein were performed as described above.
The expression plasmid for huPrP(121-230) was obtained from Dr. Werner Kremer, University of Regensburg. As described previously (47), it was cloned in pRSET A vector with an N-terminal histidine tail and thrombin cleavage site. The plasmid was transformed in Rosetta 2 (DE3). Before induction, E. coli was grown in 2YT medium (3.5% Tryptone, 2% yeast extract, 0.5% NaCl) at 37°C and 160 rpm shaking. At an OD 600 of 0.6, recombinant protein expression was induced by adding 1 mM isopropyl 1-thio-␤-D-galactopyranoside, and the growth temperature was lowered to 22°C for overnight expression. After harvesting and washing the cells twice with 5 mM EDTA, 25 mM Tris-HCl, pH 8.0, they were resuspended in 2 mM EDTA, 1% Triton X-100, 0.1 mg/ml lysozyme, 50 mM Tris-HCl, pH 8.0, and incubated for 30 min at 37°C. After addition of 0.1 mg/ml DNase and 15 mM MgCl 2 and incubation for 30 min at 37°C, they were sonicated on ice five times for 1 min (Bandelin Sonopuls, sonotrode VS 70 T, 60% amplitude).
The insoluble inclusion bodies were harvested by centrifugation (see above); washed with 5 mM EDTA, 12.5 mM Tris-HCl, pH 8.0; and dissolved in 8 M guanidinium HCl, 12.5 mM Tris-HCl, pH 8.0, at 4°C. After 1-h centrifugation (see above), the supernatant was separated by IMAC with two serially connected 5-ml Protino nickel-nitrilotriacetic acid columns. The elution of the hexahistidine-tagged PrP(121-230) occurred with a linear gradient of 500 ml from 0 to 500 mM imidazole in 6 M guanidinium HCl, 12.5 mM Tris-HCl, pH 8.0.
huPrP(121-230)-containing fractions were pooled and purified by RP-HPLC. A semipreparative C 8 column (Zorbax 300 SB-C8, 9.4 ϫ 250 mm) allowed the separation of huPrP(121-230) from impurities using a 20 -48% (v/v) gradient of acetonitrile ϩ 0.1% (v/v) TFA within 20 min followed by a 10-min step of isocratic 48% (v/v) acetonitrile ϩ 0.1% (v/v) TFA. The purification was performed at 80°C at a flow rate of 4 ml/min. The elution fractions containing huPrP(121-230) were pooled and dried by lyophilization. Thrombin cleavage of the fusion protein was performed with 2.5 mg/ml fusion protein in 50 mM MES, pH 6.0, with a final concentration of 8 units of thrombin (Serva 36402.02)/mg of protein for 7 days, when nearly 98% of the protein was digested. RP-HPLC purification, lyophilization and storage of the protein were performed as described above.

Preparation of A␤(1-42) stocks
1 mg of synthetic A␤(1-42) (Bachem AG) was incubated with 700 l of hexafluoro-2-propanol (HFIP) overnight. The solution was divided in 36-g aliquots in LoBind reaction tubes (Eppendorf AG) and lyophilized in a rotational vacuum concentrator system connected to a cold trap (both Martin Christ Gefriertrocknungsanlagen GmbH). The lyophilizates were stored at room temperature.  Table 1). The strongest interference of RD2D3-FITC with the A␤(1-42) oligo -huPrP  interaction was observed at the higher RD2D3-FITC concentration (40 M) when RD2D3-FITC was preincubated with A␤ before addition of huPrP . Experiments were done in replicates of n ϭ 3 for all orders of application of RD2D3-FITC or huPrP  to the sample. Error bars represent S.D.

Interference with A␤-PrP complex formation
Standard proteins for DGC calibration 40 g of the standard proteins ovalbumin, conalbumin, aldolase, apoferritin, and thyroglobulin in 30 mM Tris-HCl, pH 7.4, from a gel filtration high-molecular-weight calibration kit (GE Healthcare) were analyzed by sucrose DGC (see below) to calibrate the gradient.

Preparation of samples containing A␤(1-42) and huPrP (any construct)
Preincubation of A␤
Mixture of huPrP  and RD2D3-FITC (simultaneous)-A␤   Addition of huPrP  during A␤ incubation (first huPrP and then RD2D3-FITC)-A␤ preincubation was done as described before but in the presence of 0.5 molar eq of huPrP .

DGC and RP-HPLC analysis of the fractions
This method is an adjusted assay based on the QIAD assay described in Brener et al. (25). In our case, not only A␤  but also the prion protein (either huPrP , huPrP , or huPrP(90 -230)) and the D-peptide RD2D3-FITC are quantified. This assay contains the following steps.
DGC-Each 100-l sample (see "Preparation of samples containing A␤  and huPrP (any construct)" or "Preparation of samples containing A␤ , huPrP , and RD2D3-FITC") was applied on a discontinuous 30 mM Tris-HCl, pH 7.4, buffered sucrose gradient containing the following volumes and sucrose concentrations (from bottom to top): 300 l of 60% (w/w), 200 l of 50% (w/w), 200 l of 25% (w/w), 400 l of 20% (w/w), 400 l of 15% (w/w), 150 l of 10% (w/w), and 150 l of 5% (w/w). Each gradient was stepwise layered in a 11 ϫ 34-mm polyallomer centrifuge tube. Gradients were centrifuged in a TL 100 ultracentrifuge using a TLS-55 rotor (Beckman) for 3 h at 4°C and 259,000 ϫ g. The centrifuged gradients were manually fractionated from top to bottom into 13 142-l fractions. The remaining volume (arithmetically 54 l) was mixed with 80 l of 30 mM Tris-HCl, pH 7.4, forming fraction 14. For all calculations, a dilution factor of 2.48 was included for fraction 14.
RP-HPLC analysis-Each fraction was analyzed by Tris/Tricine SDS-PAGE (see below) and RP-HPLC. For the quantification of A␤ , huPrP (all constructs), and RD2D3-FITC, 20 l of each fraction was applied on a Zorbax 300 SB-C8 Stable Bond Analytical column, 4.6 ϫ 250 mm (Agilent) and measured with an Agilent 1260 Infinity system. Each compound was separated by a 10 -40% (v/v) acetonitrile gradient ϩ 0.1% (v/v) TFA within 25 min at 80°C and a flow rate of 1 ml/min. These harsh conditions are necessary to ensure the dissociation of the formed complexes, especially in density gradient fractions 11-14. For detection of the substances, the UV absorbance at 214 nm was used. Known concentrations of A␤ , huPrP (all constructs), as well as RD2D3-FITC and their corresponding plot of peak area versus protein concentration enabled the calibration and finally the calculation of the protein concentrations present in the fractions. The program package ChemStation by Agilent allowed data recording and peak area integration. All histograms were illustrated with OriginPro 9.0.
Determination of A␤:huPrP ratios-All generated complexes containing A␤(1-42) and huPrP (and RD2D3-FITC) were verified in gradient fractions 11-14. For the calculation of A␤: huPrP ratios, A␤(1-42) and huPrP amounts in fractions 11-14 were summed. Averaging over fractions 11-14 was necessary as the appearance of the complexes can shift a little within the different fractions due to manual fractionation of the gradients. Then A␤(1-42) amounts were divided by huPrP amounts to get a ratio for each experiment. The mean Ϯ S.D. of the ratios was calculated over all performed experiments.

Verification of the disulfide bond in huPrPs between Cys-179 and Cys-214 by RP-HPLC
To analyze whether purified huPrP(23-230), huPrP(90 -230), and huPrP(121-230) under study contain a disulfide bond between Cys-179 and Cys-214 in the fully oxidized state, purified samples of 5 M protein were reduced overnight with 25 Interference with A␤-PrP complex formation mM tris(2-carboxyethyl)phosphine hydrochloride in 6 M guanidinium HCl, 100 mM Tris-HCl, pH 7.4, and analyzed by RP-HPLC as described above. Samples treated only with 6 M guanidinium HCl, 50 mM Tris-HCl, pH 7.4, were used as controls. The reductive opening of the disulfide bond results in a characteristic elongation of the retention time for the reduced state when compared with the oxidized states.

SDS-PAGE and silver staining
Qualitative analysis of the DGC fractions was done by Tris/ Tricine SDS-PAGE followed by silver staining. To this end, 15 l of each fraction was diluted 1:1 in sample buffer (12% glycerol, 4% SDS, 50 mM Tris-HCl, pH 7.4, 2% ␤-mercaptoethanol), applied onto 20% Tris/Tricine gels, and subjected to gel electrophoresis at 40 mA/gel. The preparation of the Tris/Tricine gels is based on the protocol by Schägger and von Jagow (69). Gels containing samples with RD2D3-FITC were analyzed by fluorescence detection (excitation, 470 nm; emission, 530 nm) before silver staining. Silver staining of the gels based on the protocol by Heukeshoven and Dernick (70) allowed visualization of protein bands.

Dynamic light scattering
DLS was performed on a submicron particle sizer, Nicomp 380 (Particle Sizing Systems Nicomp, Santa Barbara, CA). Data were analyzed with the Nicomp algorithm using the volume-weighted Nicomp distribution analysis. The DLS sample of A␤(1-42) oligo -huPrP(23-144) complexes derived from 80 M A␤(1-42) and 40 M PrP(23-144) was prepared by pooling fractions 11-14 after sucrose DGC. For data analysis, a measured refractive index in the sample of 1.409 corresponding to 44.8% sucrose and a viscosity of 9.2 centipoise was taken into account (71).
Added A␤(1-42) oligo , either alone or after mixing and further incubation with huPrP(23-144) or RD2D3, was prepared as described under "Preparation of samples containing A␤  and huPrP (any construct)." The prepared stock solutions contained either 80 M preincubated A␤   After further incubation for 24 h in a 95% humidified atmosphere with 5% CO 2 at 37°C, cell viability was measured using Cell Proliferation Kit I (MTT) (Roche Applied Science) according to the manufacturer's instruction. The absorbance of the formazan product was determined by measuring at 570 nm after subtracting the absorption at 660 nm in a Polarstar Optima plate reader (BMG LABTECH, Offenburg, Germany). All results were normalized to cells that were treated with buffer only. The arithmetic mean of all 15 measurements per approach was calculated.

AFM
AFM was done using a Nanowizard 3 system (JPK Instruments AG). All samples were prepared as described under "Preparation of samples containing A␤  and huPrP (any construct)". 50 l of A␤(1-42) oligo or 25 l of huPrP  was incubated on freshly cleaved mica for 3 or 30 min, respectively. A␤(1-42) oligo -huPrP(23-144) heteroassemblies had to be cleared from unbound A␤  or huPrP(23-144) and were therefore centrifuged at 16,100 ϫ g at 4°C for 30 min and washed twice with 100 l of 30 mM Tris-HCl, pH 7.4, respectively. The complexes were then resuspended in 100 l of 30 mM Tris-HCl, pH 7.4. Then 50 l of the complexes was incubated for 30 min on freshly cleaved mica. All samples were washed three times with Milli-Q water and dried in a gentle stream of N 2 .
The samples were measured using intermittent contact mode with a resolution of 512 or 1024 pixels and line rates of 0.5-1 Hz in ambient conditions with a silicon cantilever with nominal spring constant of 26 newtons/m and average tip radius of 7 nm (Olympus OMCL-AC160TS). Due to the differing composition of the megadalton-sized aggregates concerning adhesion, stiffness, and perforation, the imaging parameters (amplitude, set point, and gain) had to be adapted slightly, and the cantilever had to be changed often.
The height images of A␤(1-42) oligo and huPrP(23-144) were flattened with JPK Data Processing software 5.0.69. The statistics of particle dimensions of A␤(1-42) oligo were done with Gwyddion 2.44 grain analysis. After plane leveling, grains were marked with a threshold of 13%. The maximum height of the individual grain was corrected with subtraction of the grain minimum.
The lateral size is affected by tip convolution effects (⌬) in AFM images. Considering the nominal radius of r tip ϭ 7 nm of the AFM tip, we corrected the size of the lateral dimension according to Interference with A␤-PrP complex formation Equation 1 for objects below the tip round end as shown in Canet-Ferrer et al. (72). h describes the height of the object.

CD spectroscopy
6 M huPrP(23-230), huPrP(23-144), or huPrP(90 -230) in 10 mM Tris-HCl, pH 7.4, was analyzed by CD spectroscopy. Each sample was transferred into a cuvette (110-QS, 1 mm, Hellma Analytics), and spectra from 186 to 280 nm were recorded at a scan speed of 50 nm/min in a Jasco J-815 spectropolarimeter. Spectra of 10 mM Tris-HCl, pH 7.4, were used as reference and subtracted from the protein spectra. Ten spectra of each huPrP sample were recorded and accumulated to improve the signal-to-noise ratio.  (73) at different temperatures ranging from 5.0 to 20.0°C on a Bruker 600-MHz, Varian 800-MHz, or Varian 900-MHz NMR spectrometer equipped with cryogenically cooled triple-or quadruple-resonance probes with z-axis pulsed-field gradient capabilities. The sample temperature was calibrated using methanol-d 4 (99.8%) (74). The 1 H 2 O resonance was suppressed by gradient coherence selection with quadrature detection in the indirect 15 N dimension achieved by the echo-antiecho method (75,76). A WALTZ-16 sequence (77) with a field strength of at least 1.1 kHz was used for 15 N decoupling during acquisition. At least 927 (128) complex data points were acquired with a spectral width of 16 ppm (26.0 ppm) in the 1 H ( 15 N) dimension. All NMR spectra were processed using NMRPipe and NMRDraw (78) and analyzed with NMRViewJ (79). 1 H chemical shifts were referenced with respect to external 4,4-dimethyl-4-silapentane-1-sulfonic acid in D 2 O, and 15 N chemical shifts were referenced indirectly (80).