Soluble Aβ aggregates can inhibit prion propagation

Mammalian prions cause lethal neurodegenerative diseases such as Creutzfeldt–Jakob disease (CJD) and consist of multi-chain assemblies of misfolded cellular prion protein (PrPC). Ligands that bind to PrPC can inhibit prion propagation and neurotoxicity. Extensive prior work established that certain soluble assemblies of the Alzheimer's disease (AD)-associated amyloid β-protein (Aβ) can tightly bind to PrPC, and that this interaction may be relevant to their toxicity in AD. Here, we investigated whether such soluble Aβ assemblies might, conversely, have an inhibitory effect on prion propagation. Using cellular models of prion infection and propagation and distinct Aβ preparations, we found that the form of Aβ assemblies which most avidly bound to PrP in vitro also inhibited prion infection and propagation. By contrast, forms of Aβ which exhibit little or no binding to PrP were unable to attenuate prion propagation. These data suggest that soluble aggregates of Aβ can compete with prions for binding to PrPC and emphasize the bidirectional nature of the interplay between Aβ and PrPC in Alzheimer's and prion diseases. Such inhibitory effects of Aβ on prion propagation may contribute to the apparent fall-off in the incidence of sporadic CJD at advanced age where cerebral Aβ deposition is common.


Introduction
Prion diseases are fatal neurodegenerative disorders associated with propagation of multi-chain assemblies of misfolded cellular prion protein (PrP C ) [1,2]. Prions propagate by recruitment of a-helical-rich PrP C into b-sheet-rich infectious rod-like structures [3,4]. In addition to serving as the precursor of infectious prions, expression of PrP C is also required for the neurotoxicity in prion infection [5][6][7][8].
Numerous studies suggest PrP C may play a role in Alzheimer's disease (AD), and there is evidence that PrP can modulate the production, aggregation and toxicity of the amyloid b-protein (Ab) [9][10][11]. In 2009, Lauren et al. [11] reported that a preparation of aggregated synthetic Ab1-42 known as Ab-derived diffusible ligands (ADDLs), which contained a mixture of globular oligomers and protofibrils, bound to PrP C . Using a series of deletion constructs and anti-PrP antibodies, it was shown that PrP residues 95 -110 were required for ADDL binding. In accord with this finding, the authors demonstrated that knock-out of the mouse PrP gene (Prnp) or pre-treatment of hippocampal slices with an antibody directed to PrP 93-109 protected against ADDL-induced synaptotoxicity. These provocative findings were followed by multiple in vivo and in vitro studies, most of which supported a role for PrP C in aspects of Abmediated toxicity [10,[12][13][14][15][16][17][18][19][20][21]. However, others have reported deleterious effects of Ab that do not require PrP C expression [22 -25]. Recently, we reported that only certain Ab assemblies exert toxicity in a PrP-dependent fashion which may explain some of these apparently discrepant findings [26].
However, all published studies that have examined binding of Ab to PrP C agree that there is high affinity and specific binding for soluble aggregates [11,13,[21][22][23][26][27][28][29], and high-resolution analysis suggests that binding of Ab occurs at two sites: centred around residues approximately 23 -33 and approximately 88 -113 [29]. Although initially surprising, the finding that PrP C can serve as an acceptor for soluble aggregates of Ab [11] is consistent with the hypothesis that the unstructured N-terminus (encompassing residues approx. 23 -128) of PrP C acts as a molecular sensor which can interact with a broad range of ligands [30], including other b-sheet-rich oligomeric proteins [28]. Moreover, the same binding sites for soluble aggregates of Ab have previously been shown to also be important for binding of prions to PrP C [31][32][33][34][35][36].
While many studies have investigated the interaction between Ab and PrP C , and how it might contribute to AD pathogenesis, there has been little research on whether Ab binding to PrP C can affect prion propagation. Here, we used the well-established cell-based prion bioassay (the scrapie cell assay using PK1/2 neuroblastoma-derived cells) [37] and a chronically prion-infected cell line (iPK1/2 cells) [38,39] to address these critical issues. We found that soluble Ab aggregates (ADDLs), but not Ab monomers or fibrils, could prevent infection of PK1/2 cells when ADDLs were co-administered with the prion inoculum. Strikingly, when added to iPK1/2 cells already chronically infected with prions, ADDLs had a marked cell curing effect. This protective effect appears to be mediated by ADDL binding to PrP C . While diverse studies have linked PrP to AD [9,11,40 -43], our data raise the possibility that soluble Ab aggregates may actually protect against prion disease. Thus, whether Ab binding to PrP C has pathogenic or protective effects may depend on the relative concentrations of relevant Ab and PrP assemblies.

Reagents
All chemicals and reagents were purchased from Sigma-Aldrich unless otherwise noted. Synthetic Ab1-42 and Ab1-40 were synthesized and purified using reversed-phase HPLC by Dr James I. Elliott at the ERI Amyloid Laboratory (Oxford, CT, USA). Peptide mass and purity (greater than 99%) were confirmed by reversed-phase HPLC and electrospray/ion trap mass spectrometry. All tissue culture reagents were obtained from Invitrogen.

Ab preparations
Ab is prone to aggregate and can form an array of different assemblies. In this study, we used conditions to yield preparations highly enriched in: (i) monomers, (ii) pre-fibrillar aggregates, known as ADDLs and (iii) amyloid fibrils. Monomeric Ab was prepared by dissolving dry Ab1-40 peptide at 2 mg ml 21 in 6 M guanidine hydrochloride (GuHCl) and then subjecting this preparation to asymmetric flow fieldflow fractionation (AFFFF). The AFFFF channel was eluted in 50 mM ammonium bicarbonate pH 8.5, and fractions containing monomeric Ab, as judged by molar mass (approx. 4000 g mol 21 ), were collected and immediately frozen at 2808C. ADDLs were prepared essentially as described previously [19], approximately 25 mg of Ab1-42 peptide was dissolved in anhydrous DMSO, gently rocked for approximately 5 min and then diluted to 0.5 mg ml 21 in phenol red-free Ham's F12 medium without L-glutamine (Caisson Labs) and incubated quiescently at room temperature (RT). At approximately 6 h intervals, aliquots were removed, briefly centrifuged at 16 100g and analysed using AFFFF. Typically, at 24 -36 h, less than 20% of the injected mass eluted as monomer, as judged by the area under the curve of both monomer and oligomer peaks. Thereafter, the material was aliquoted and stored frozen at 2808C. To form fibrils, Ab was solubilized and incubated as for ADDLs, but the incubation continued for 30 days. For cell culture experiments, Ab preparations were buffer exchanged into Opti-MEM using a centrifugal concentrator (Amicon, Ultra 0.5 ml, 5 K cut-off ).

Asymmetric flow field-flow fractionation and multiangle light scattering
Experiments were conducted using a 24.6 cm long channel fitted with a 350 mm spacer and a 5 kDa MWCO polyethersulfone membrane. Aliquots of Ab preparations (190 ml) were injected onto an Eclipse DualTec AFFFF (Wyatt Technology, Santa Barbara, CA, USA) and eluted with 50 mM ammonium acetate pH 8.5. The sample was injected at 0.2 ml min 21 , followed by a 1 min focusing period, and then eluted with a 1.5 ml min 21 cross-flow for 45 min. Light scattering was performed using a Wyatt Dawn Heleos II multi-angle light scattering module to calculate the molar mass.

Electron microscopy
Negative stain electron microscopy (EM) was performed as described previously [26]. Peptide solutions (5 ml) were loaded onto negatively charged glow-discharged copper grids coated with a continuous carbon film. Samples were left to adhere for 120 s and excess solution blotted with grade 4 Whatman paper. Thereafter, grids were stained with 2% uranyl acetate for 40 s, blotted and air-dried. Images were acquired on an FEI Tecnai T10 electron microscope operating at 100 kV and recorded on a 1 k Â 1 k charge-coupled device camera (Gatan) at a typical magnification of 34 000 with a pixel size of 5 Å . Prion rods were purified as described previously [4,44] concentrated to 100Â (relative to starting 10% brain homogenate) and mixed with 10 mM ADDLs and incubated at 218C for 1 h. Prion rods were pelleted by centrifugation at 16 100g and 258C for 30 min. The pellet was washed once with Opti-MEM and centrifuged a final time. The pellet was resuspended in Opti-MEM to one half the volume of the starting prion/ADDL solution and stained for EM as described above. Images were analysed for evidence of ADDLs binding to prion rods. First, the number of protofibrillar and spherical Ab species in an area containing a rod cluster were counted, then the number of Ab species in an equivalent sized area that did not include prion rods were counted. This was repeated for three rod clusters and three rsob.royalsocietypublishing.org Open Biol. 7: 170158 rod-free areas, and images were analysed by two different users.

Automated scrapie cell assay
An automated version of the standard scrapie cell assay (SCA) using PK1/2 cells [45] was used as described previously [37,46]. Briefly, PK1/2 cells were grown in Opti-MEM, containing 10% fetal calf serum; 100 U ml 21 penicillin and 100 mg ml 21 streptomycin at 378C, 5% CO 2 . Twenty-four hours before infection with Rocky Mountain Laboratory (RML), PK1/2 cells were seeded in 96-well plates at 18 000 cells per well and grown in Opti-MEM. ADDLs or bovine serum albumin (BSA) were incubated with RML prioninfected brain homogenate (I-BH) (designated I8700; [44]) for 1 h at RT and then added to cells and incubated for 72 h. Thereafter, cells were split 1 : 8 into fresh cell culture media containing fetal calf serum and grown to confluence. Two further passages were conducted, removing initial inoculum, before transferring a sample of the cells to enzyme-linked immunospot (ELISPOT) plates for measurement of the number of prion-infected cells (identified by detection of proteinase K-resistant PrP, PrP Sc ), the 'spot count' [45]. The viability of the cells was monitored using Trypan Blue. Prion titre in the experimental samples was determined by reference to a calibration curve in each experiment derived from a serial dilution of an RML brain homogenate of known prion titre (10 8.3 intracerebral LD 50 units g 21 brain) determined by prior mouse bioassay [37,44].
The ability of ADDLs to retard prion propagation was calculated relative to the number of infected cells 'spot count' of cells incubated with the equivalent RML concentration alone ( positive control) and the 'spot count' (background noise) of cells incubated without RML present.

Curing assay of chronically Rocky Mountain Laboratory prion-infected cells
Chronically RML prion-infected PK1/2 (designated iPK1/2) cells were used to assay curing activity. As described previously, these cells are able to maintain a robust prion infection long term in culture [38,39,47]. Briefly, iPK1/2 cells were produced by incubating cells with 1 Â 10 23 RML-I-BH for 72 h [44]. Thereafter, cells were passaged every 2-3 days for 2 weeks to remove any remaining inoculum. A portion of infected cells was analysed for RML prion infectivity by ELISPOT and the remainder stored in liquid nitrogen. For experiments, cells were thawed and cultured as described above. In order to maintain a consistent level of prion infection, cells were never passaged more than 15 times. iPK1/2 cells were seeded at 6000 cells per well, in 384well plates. The cells were grown in Opti-MEM for 3 days at 378C and 5% CO 2 + Ab. Additionally, positive (2 mM 5000 Da dextran sulfate) and negative (cells only) controls were included on each plate. Infected cells produce PrP Sc . On day 4, cells were analysed for both viability and PrP Sc content. Cell viability was assessed using the CellTiter-Glo Luminescent assay (Promega) and PrP Sc levels were measured by dot blot. For PrP Sc analysis, the media was removed from the cells, lysis buffer added (Tris buffer pH: 7.5 with NaCl, Triton X-100 and sodium deoxycholate) and the cells lysed on ice for 20 min. The lysates were mixed and bound to nitrocellulose membrane, using a 96-well biodot microfiltration apparatus (BioRad). The samples were treated with proteinase K (5 mg ml 21 ) for 1 h at 378C and then denatured using 3 M guanidine thiocyanate. PrP Sc was detected using the anti-PrP antibody ICSM18 (D-Gen Ltd, London) and goat anti-mouse IgG-IRDye 800CW (LI-COR Biosciences, Santa Clara, CA, USA). Spots were visualized using an Odyssey infrared imaging system (LI-COR Biosciences) and the relative intensity of the infrared signal was determined using the systems software.

Immunofluorescence
PK1/2 cells were seeded at 18 000 cells per coverslip in 24well plates. The cells were grown in serum-free media for 3 days at 378C and 5% CO 2 + Ab. On day 4, coverslips were washed three times with PBS and then fixed in 4% PFA at RT for 15 min. After fixation, coverslips were washed twice with PBS and then blocked with 5% BSA/PBS (1 h at RT) and stained. Cells were incubated for 1 h at RT with the anti-PrP antibody ICSM18 (1.25 mg ml 21 ) and/or the anti-Ab rabbit antiserum #2454 at 1 : 2000 dilution (Cell Signalling, Danvers, MA, USA). Thereafter, cells were washed with PBS and then incubated with Alexa Fluor 488 tagged donkey anti-mouse IgG (H þ L) (#A-21 202) at 2 mg ml 21 and/or Alexa Fluor 546-tagged donkey anti-rabbit IgG (H þ L) (#A10040) at 3.3 mg ml 21 (Invitrogen Life Technologies). Nuclei were stained with 4,6-diamidino-2phenylindole, 1 mg ml 21 for 1 h at RT. Cells were washed with PBS and then mounted using fluorescence mounting medium (DAKO). Images were captured using a Zeiss LSM710 confocal laser scanning microscope and co-localization quantified using VOLOCITY 3D imaging software (Perkin Elmer).

Ab-derived diffusible ligands inhibit prion propagation and cure prion infection
ADDLs are a polydisperse solution of soluble Ab aggregates which include globular oligomers, protofibrils and monomer [11,48] and bind to the PrP C specifically and with high affinity [11,13,21,26,27,29]. Two regions of PrP C (one centred around residues 23 -33 and the other around 88-113) are particularly important for Ab binding [11,13,27,29], and these are the same sites thought to be important for PrP Sc binding to PrP c [31 -36]. Thus, we sought to determine if ADDLs could compete with prions for binding to PrP C and attenuate prion propagation. As ADDLs are known to bind with high affinity to PrP C , whereas Ab monomers show little or no binding and fibrils exhibit only weak binding [26], we generated ADDLs from Ab42 and relatively homogeneous preparations of Ab40 monomers and Ab42 fibrils, and characterized each using AFFFF and EM (figure 1). AFFFF is a flow-based method in which separation takes place in a channel where sample retention is caused by the action of a cross-flow that is generated by a second independent stream that runs across the channel at right angles to the primary channel flow [49]. Unlike more commonly used size exclusion chromatography, rsob.royalsocietypublishing.org Open Biol. 7: 170158 in AFFFF, small particles elute earlier than larger particles. AFFFF of ADDLs confirmed the presence of a small amount of monomer and a range of Ab assemblies with molar masses from 300 000 to 3 000 000 g mol 21 (figure 1b). EM also indicated that ADDLs contained a mixture of structures, including imperfect spheres of approximately 5-10 nm rsob.royalsocietypublishing.org Open Biol. 7: 170158 diameter and abundant protofibrils (flexible fibrils) of approximately 5-10 nm diameter and less than 100 nm in length. By contrast, our monomer and fibril preparations were relatively homogeneous. Monomer preparations had a molar mass of 4000-5000 g mol 21 (figure 1a) and contained no structures detectable by EM, whereas fibrils had a molar mass of greater than 10 9 g mol 21 and formed complex latticeworks of long fibrils with diameters of approximately 10 nm ( figure 1c).
To test if ADDLs could attenuate prion infectivity, we used an automated high-throughput prion bioassay referred to as the automated crapie cell assay (ASCA) [37,45,50,51]. PK1/2 cells were incubated with an RML prion-I-BH [44]. Cells were grown to confluence, split 1 : 8 and grown to confluency again. The cycle of growth and passage was repeated a further two times to remove initial infecting inoculum, and confluent cells from the third passage were used to measure the proportion of infected cells [37].  [37]. Importantly, addition of ADDLs to the inoculum caused a dosedependent decrease in the extent of prion propagation, an effect that was directly related to the prion titre in the starting inoculum (figure 2b). For instance, when I-BH was used at a dilution of 3 Â 10 26 (blue curve), maximal inhibition of infectivity was achieved with a dose of 5 mM ADDLs, whereas when more concentrated I-BH (1 Â 10 24 , green curve) was used, higher concentrations of ADDLs were required to significantly attenuate prion propagation (figure 2b). In comparison, addition of BSA had no effect on prion propagation (black curve, figure 2c).
Next, we investigated if ADDLs could cure cells with an established chronic prion infection. Chronically infected PK1/2 cells (iPK1/2 cells) accumulate PrP Sc yet remain viable (see Material and methods) and have been successfully used in drug screening to identify anti-prion compounds [38,39,47]. Dextran sulfate is effective at curing prion infection in this assay and has been used a positive control in drug screening [39,52] and was used as a comparator to assess the inhibitory activity of ADDLs across experiments. iPK1/ 2 cells were incubated with increasing concentrations of dextran sulfate for 72 h (the maximum time before the cells require passage) and the levels of PrP Sc were measured by dot blot. As expected, dextran sulfate caused a dose-dependent decrease in PrP Sc with an apparent IC 50

Ab-derived diffusible ligands, but not Ab monomers or fibrils inhibit prion propagation
Previous studies found that Ab monomers do not bind to PrP C and that Ab fibrils bind to PrP C less well than pre-fibrillar intermediates [11,26,27], hence we sought to determine if there was a relationship between the ability of different Ab structures to bind to PrP C and their ability to reduce prion propagation ( figure 4). First, we examined if Ab40 monomer could influence prion propagation in the ASCA. As expected, ADDLs caused a dose-dependent decrease in prion propagation, whereas Ab monomers had no effect (figure 4a). To determine if this prion-curing activity was similarly specific for pre-fibrillar Ab species, we compared the effects of ADDLs versus Ab40 monomers and Ab42 fibrils using the chronic prion-infected cell assay. As before ( figure 4b,c), ADDLs caused a dose-dependent decrease in prion infection (blue solid circles), whereas Ab monomers and fibrils had no effect (figure 4b). Given that reduction in detectable PrP Sc could occur due to cell loss, and that Ab is known to be toxic to certain cells, we were careful to measure cell viability in all of the cultures treated with Ab. The number of metabolically active cells (as assessed by the CellTiter-Glo Luminescent Cell Viability assay) did not change over the concentration range at which ADDLs inhibited prion infectivity (less than or equal to 2 Â 10 25 M; figure 4c). Therefore, the reduction in PrP Sc levels mediated by ADDLs is not a consequence of cell compromise, but rather a specific effect comparable to that seen with other prion-curing agents [39,52,53].

Ab-derived diffusible ligands bind to the surface of PrP C -expressing cells, but not to purified prion rods
As both PrP C and PrP Sc share the same primary structure, including the amino acids that comprise the Ab binding sites [27,29], we investigated whether the prion inhibition we observed was due to Ab acting on PrP C or prions, or both. If ADDLs inhibited infectivity by binding to prions, then it should be possible to detect ADDLs bound to PrP Sc .
To address this issue, we incubated highly purified infectious prion rods [4,44] with ADDLs under the same conditions used in the ASCA, and then searched for binding of ADDLs to prions using negative stain EM ( figure 5). Both prion rods and ADDLs were readily detected, but we saw no evidence of co-localization. This rather rudimentary assay provides the first evidence that Ab does not bind to PrP Sc .  rsob.royalsocietypublishing.org Open Biol. 7: 170158 As ADDLs did not seem to interact with prion rods, we looked for evidence of ADDL binding to PrP C on the surface of PK1/2 cells. Immunostaining of non-permeabilized cells detected Ab on the surface of PK1/2 cells and partial colocalization with PrP C (figure 6a-c). Interestingly, we also observed that treating PK1/2 cells with ADDLs increased cell surface levels of PrP C (figure 6b,d). In accord with earlier reports, ADDLs appear capable of binding to PrP C [11,15] and retaining PrP C at the plasma membrane [54]. Thus, it seems likely that ADDLs inhibit prion levels and propagation by competing with prions for binding to PrP C , and may also retard internalization of PrP C .
Here, we show that ADDLs inhibit prion infectivity in a dose-dependent manner and reduced the levels of proteinase K-resistant PrP in chronically prion-infected cells. As both PrP C and PrP Sc have the same primary structure, including the sites involved in ADDL binding, the ability of ADDLs to attenuate prion propagation could result from interactions involving either PrP C or PrP Sc . When ADDLs were mixed with highly purified prions, we found no evidence of  Infectious RML prion rods were purified from RML-infected mouse brain [4,44] and incubated with 10 mM ADDLs for 1 h before imaging by EM. The number of Ab oligomers was counted in equivalent areas containing rods or without rods in three EM images of different RML rod-Ab oligomer clusters by two different users. No enrichment of Ab oligomers on the rods surface was observed (non-significant, p ¼ 0.456, unpaired t-test). Scale bars, 50 nm. *p 0.05; ****p 0.0001. rsob.royalsocietypublishing.org Open Biol. 7: 170158 binding, whereas when ADDLs were added to PrP C -expressing cells ADDLs partially co-localized with cell surface PrP C . Moreover, as we and others have shown previously [11,13,21,26,27], and we confirmed for preparations used in this study (data not shown), recombinant monomeric PrP (rPrP) readily binds ADDLs. Taken together, these data suggest that ADDLs can attenuate prion infectivity by directly binding to PrP C and acting as a competitive inhibitor (electronic supplementary material, figure S1). Such a mechanism would allow the cells' natural prion clearance rate [57] to outpace any residual propagation, resulting in the low to absent levels of PrP Sc observed when ADDLs were used in our experiments. Consistent with this mechanism, we also found that Ab species (monomer and fibrils) which show little or no affinity for monomeric PrP lack the ability to attenuate prion propagation. The binding response between rPrP and ADDLs indicates an apparent dissociation constant of approximately 100 nM,  rsob.royalsocietypublishing.org Open Biol. 7: 170158 whereas the IC 50 of ADDLs in the chronic prion-infected cell assay was approximately 15 mM. The difference between binding to rPrP and the ability to inhibit prion propagation probably results because (i) ADDLs are known to bind non-PrP membrane components [11], (ii) our assays use mitotic cells which have a doubling time of approximately 24 h and (iii) ADDLs are competing with prions for binding to PrP C . In terms of ADDLs, the concentration that might be needed to attenuate prion formation in vivo, it is worth considering that the amount of ADDLs used in our experiments are expressed as monomer equivalents, yet we know that the component of ADDLs that binds to PrP has a relatively high molecular weight and only contributes a fraction of the total Ab present [26]. Therefore, the actual K D for the binding component of ADDLs must be significantly lower, and may be in the picomolar range. As to how much ADDLs would be required to inhibit prion propagation in vivo, that will depend on the amount of infectious prions. Our findings are in apparent conflict with a prior study that reported prion inoculation of Tg2576 APP transgenic mice accelerated both Ab deposition and prion disease [58]. A possible explanation for the divergence in results seen with Tg2576 mice and those we detected in PK1 cells relates to the forms of Ab tested in our study and those produced by Tg2576 mice. We and others have previously shown that only certain forms of Ab bind to PrP [11,13] and that only particular effects of Ab are mediated by PrP [26]. Similarly, it is known that certain APP transgenic mice exhibit cognitive phenotypes that depend on the expression of PrP, whereas others do not [16,25]. In terms of the acceleration of prion disease in Tg2576 mice, it is interesting to note that deleting PrP C expression in Tg2576 results in only a partial rescue of cognitive performance as opposed to the complete recovery seen in other APP transgenic lines [59]. Further, Tg2576 mice have been shown to produce little or no Ab species capable of binding to PrP [59]. Given that Tg2576 mice show minimal PrP-dependent deficits and produce little Ab that binds PrP, it is perhaps not surprising that Tg2576 mice are unable to attenuate prion infectivity and propagation.
Clearly, high concentrations of ADDLs should completely inhibit prion propagation, but they are also expected to cause neuronal dysfunction. Thus, high levels of soluble Ab assemblies may provide relative protection from human prion disease, but cause AD. The lack of co-localization of disease-associated PrP and Ab deposits seen in a recent study is of interest in this regard [60]. These observations support the notion that soluble aggregates of Ab and PrP may compete for binding to PrP C in vivo and that the balance between the levels of these aggregates is a critical determinant of whether and what form of neurodegenerative disease will result. The most common human prion disease, sporadic Creutzfeldt -Jakob disease, which has a relatively uniform incidence worldwide and apparently random population distribution, is thought to represent the spontaneous production of prions as a rare stochastic event [61,62]. In this regard, it has always been intriguing why its apparent incidence falls at advanced age (greater than 80 years) [63,64]. While this may in part be due to lower diagnosis rates in the elderly, it is conceivable that this could also be related to the common occurrence of Ab deposition in this age group.
Data accessibility. All relevant data are presented in figures in the manuscript and electronic supplementary material.