Heterologous Amyloid Seeding: Revisiting the Role of Acetylcholinesterase in Alzheimer's Disease

Neurodegenerative diseases associated with abnormal protein folding and ordered aggregation require an initial trigger which may be infectious, inherited, post-inflammatory or idiopathic. Proteolytic cleavage to generate vulnerable precursors, such as amyloid-β peptide (Aβ) production via β and γ secretases in Alzheimer's Disease (AD), is one such trigger, but the proteolytic removal of these fragments is also aetiologically important. The levels of Aβ in the central nervous system are regulated by several catabolic proteases, including insulysin (IDE) and neprilysin (NEP). The known association of human acetylcholinesterase (hAChE) with pathological aggregates in AD together with its ability to increase Aβ fibrilization prompted us to search for proteolytic triggers that could enhance this process. The hAChE C-terminal domain (T40, AChE575-614) is an exposed amphiphilic α-helix involved in enzyme oligomerisation, but it also contains a conformational switch region (CSR) with high propensity for conversion to non-native (hidden) β-strand, a property associated with amyloidogenicity. A synthetic peptide (AChE586-599) encompassing the CSR region shares homology with Aβ and forms β-sheet amyloid fibrils. We investigated the influence of IDE and NEP proteolysis on the formation and degradation of relevant hAChE β-sheet species. By combining reverse-phase HPLC and mass spectrometry, we established that the enzyme digestion profiles on T40 versus AChE586-599, or versus Aβ, differed. Moreover, IDE digestion of T40 triggered the conformational switch from α- to β-structures, resulting in surfactant CSR species that self-assembled into amyloid fibril precursors (oligomers). Crucially, these CSR species significantly increased Aβ fibril formation both by seeding the energetically unfavorable formation of amyloid nuclei and by enhancing the rate of amyloid elongation. Hence, these results may offer an explanation for observations that implicate hAChE in the extent of Aβ deposition in the brain. Furthermore, this process of heterologous amyloid seeding by a proteolytic fragment from another protein may represent a previously underestimated pathological trigger, implying that the abundance of the major amyloidogenic species (Aβ in AD, for example) may not be the only important factor in neurodegeneration.


INTRODUCTION
Several human neurodegenerative syndromes, such as Alzheimer's, Parkinson's, Huntington's and Prion diseases, are thought to possess an underlying common pathological mechanism in which protein misfolding leads to protein aggregation and polymerization. The polymerization process results in the formation of amyloid fibrils with a cross-b sheet fold that deposits and accumulates in the brain. Amyloid fibrilization is a multistep process characterized by an energetically unfavorable formation of nuclei (lag phase) followed by cooperative amyloid elongation. In the case of Alzheimer's disease (AD), the extracellular deposition of amyloid-b-peptide (Ab) in senile plaques and intracellular deposition of hyperphosphorylated Tau in neurofibrillary tangles are characteristic of the pathology [1]. Ab is a 40 to 43 amino acid peptide resulting from paired endoproteolysis of the b2amyloid precursor protein (APP) [2,1]. Recent attention has focussed on Ab catabolism to understand the mechanisms leading to its excessive accumulation during AD. A number of unrelated proteases were identified, among which insulysin (IDE) and neprilysin (NEP) are undoubtedly involved in Ab clearance [3,4]. IDE is a thiol zinc metalloprotease found in the brain and located mainly in cytosol [5,6,7]. IDE cleaves a broad range of peptides and has been proposed to be an amyloid scavenger recognizing structural b-rich folds found in amyloid forming peptides (e.g. insulin and Ab) [4]. A genetic linkage was found between the chromosome 10q locus encoding IDE and onset of AD [8]. In IDE deficient mice, cerebral levels of Ab are increased [5,7], conversely mice overexpressing IDE and APP exhibited decreased Ab levels, reduced plaque burden and protection from premature death [6]. NEP is also a zinc metalloprotease found in the brain; it cleaves on the amino side of hydrophobic residues in a variety of peptides (e.g. substance P and enkephalin) [9]. NEP localization at the plasma membrane makes it a candidate for degradation of extracellular Ab [9]. NEP deficiency and over-expression studies in mice gave comparable results to those for IDE [3,6]. Whereas IDE was found to degrade only soluble monomeric Ab, NEP can hydrolyze both monomeric and oligomeric Ab [10,11].
Although Ab is a key player in the pathology associated with senile plaques, other proteins such as cholinesterases have been implicated [12,13,14,15]. The evidence is as follows; both human acetylcholinesterase (hAChE) and butyrylcholinesterase (hBuChE) are associated with senile plaques and both the pattern of hAChE oligomerisation and its enzymatic activity are altered in brain areas affected by AD [12,13,16,14,17]. BuChE inhibitors were shown to reduce level of APP and also to improve cognitive function in patients with moderate AD [18,19]. Whereas hAChE activity diminishes in the cortex of AD patients, hBuChE activity remains unchanged or increases [20,13,14] and it has been proposed that hBuChE acts as a substitute for hAChE when hAChE is impaired [21]. However, the role of hBuChE in normal or AD brains remains unclear. Various studies suggest that hAChE promotes Ab fibrilization and deposition in pathological aggregates [22,23] but the mechanism remains unknown. Moreover, double transgenic mice expressing human APP (hAPP) and hAChE developed earlier disease than single transgenic hAPP mice, accompanied by increased plaque deposition and pathology [23]. Neurodegenerative diseases associated with abnormal protein folding and aggregation are nucleation-dependent, which involves a slow and unfavorable nucleation phase (known as the lag phase) during which monomers associate to form ordered oligomeric nuclei, an elongation phase during which the nuclei exceed a threshold size, become stable and monomers can be added favorably to them, and finally a plateau phase in which the monomer concentration falls below the threshold aborting further fibril extension [24]. Because of this nucleation-dependency, neurodegenerative diseases may require an initial trigger. This trigger may involve different pathological pathways including the effects of molecules acting as pathological chaperones, as well as seeding events (defined as involvement of exogenous nuclei to bypass the slow nucleation event), both homologous and heterologous. In the case of heterologous seeding, the seed originates independently of the molecular species that will make up the bulk of the accumulating misfolded material. Along with hAChE, other molecules have been shown to enhance the nucleation phase of Ab fibrilization. For example, glycosaminoglycans and membrane glycolipids can mediate Ab aggregation [25,26]. Thus, one may postulate that the subtle effect of these 'secondary' molecules might be heterologous seeding of Ab, which could represent one of the trigger for more severe Ab pathology during AD.
The non-amyloidogenic and a-helical C-terminal oligomerisation domain of hAChE (T40, AChE 575-614 ) [27,28,29] contains a region that shares homology with Ab. Computational identification of non-native (hidden) b-strand propensity in protein sequences had predicted the minimal amyloidogenic fragments for Ab and a-synuclein [30]. When applied to T40, a short and unique predicted conformational switch region (CSR, from W 585 to K 599 ) with high propensity for conversion to non-native (hidden) b-strand was identified ( Figure 1) [31,30], with a strong propensity for conversion to b-strand for the sequence Y 594 MVHWK 599 and A 586 EFHR 590 more weakly. A peptide synthesized to include this CSR region (AChE 586-599 ) adopts a b-sheet conformation and selfassembles into amyloid fibrils, a structure associated with AD plaques [31,30]. AChE 586-599 was previously identified as a region of high hidden b-propensity by the independent computational study that predicted the Ab and a-synuclein amyloidogenic fragments. The mechanisms that could trigger such a conformational switch in this region are unknown but are worth seeking since they could represent the connection between AChE and increased Ab fibril formation during AD pathogenesis. Proteolytic processes could liberate fibrilogenic peptides (CSR-like) from the non-amyloidogenic and a-helical C-terminus of AChE that is exposed in the monomer and implicated in tetramer formation [32]. The plausibility of such a mechanism is reinforced by the observation that hydrophilic monomers of bovine brain AChE are not reactive with an antibody raised against the extreme Cterminus of the T40 domain, consistent with C-terminal truncation events [33]. Moreover, the normal reactivity of the AChE tetrameric form with this antibody is lost after limited proteolysis of the tetrameric species, suggesting that the T40 domain remains vulnerable to proteolysis even in assembled tetramers [33]. Candidate proteases include IDE and NEP, which are known to be present and active in the extracellular space of the brain and are already clearly implicated in processing of amyloidogenic peptides in the central nervous system.
In this study, we examined the activity of IDE and NEP on formation of relevant b-sheet molecular species from the nonamyloidogenic and a-helical T40 fragment of AChE and degradation of pre-assembled b-sheet oligomers. IDE cleaved both nonamyloidogenic T40 and amyloid forming AChE 586-599 , whereas NEP only cleaved the AChE 586-599 substrate. Digestion of the nonamyloidogenic and a-helical T40 by IDE triggered the formation of b-structures that formed amyloid precursors (oligomers) and generated surface-active CSR species (detergent-like), which seeded Ab fibrilization by reducing the lag phase and enhancing the rate of amyloid elongation. The heterologous seeding of Ab by IDEmediated amyloidogenic hAChE fragments may offer an explanation for the implication of hAChE in the extent of Ab deposition in the brain [34]. Ab heterologous seeding by proteolytic fragments from another abundant CNS protein may also represent a previously undescribed pathological trigger, in which the abundance of Ab may not be the only important factor in AD.

T40-degrading activity of IDE
We examined the effect of proteolytic processes on the nonamyloidogenic and a-helical C-terminus of AChE (T40) that is exposed in the monomer and implicated in tetramer formation [32]. The ability of IDE to digest T40 is depicted in the western blot probed with a rabbit anti-T40 antiserum (KD69 antiserum)(Figure 2A). In the absence of IDE, the 5 kDa T40 migrates as monomers and dimers. In the presence of IDE, monomeric T40 was progressively digested with complete disappearance by 80 min incubation, whereas a proportion of the dimeric form remained undigested. The specific activity of IDE on T40 was demonstrated by a dose dependent inhibition with insulin (a natural IDE substrate) or 1,10-phenanthroline (a zinc metalloprotease inhibitor)( Figure 2B). Complete inhibition was achieved at equimolar levels of insulin (16 mM), which did not occur with an irrelevant protein (IgG). The 1,10-phenanthroline was prepared in methanol, which addition to the IDE reaction did not affect degradation of T40.
To determine a full map for IDE cleavage of T40, a temporal series of products were analyzed by mass spectrometry (MS) and reported based upon T40 numbering (Asp 1 to Leu 40 ). After 2 min incubation, four major products (a-d) were visible and corresponded to initial cleavages between Phe 14 -His 15 , Ser 19 -Tyr 20 , and Trp 24 -Lys 25 ( Figures 2C and 2D). As digest time increased, more products appeared with peaks a and b dominating up to 15 min before peak e became the prominent product at 30 min, coinciding with near complete digestion of T40. The progression and regression of peaks a-d suggests that after digestion at the primary positions, T40 fragments are serially digested to generate other peptides. A complete digestion map of T40 by IDE after 30 min incubation is available as supporting information ( Figure S1). Peaks b to h contained significant amounts of CSR species ( Figure 2E) along with other identified peptides. Peaks b, e and f contained species encompassing the sequence YMVHW with high propensity for conversion to b-sheet as major IDE cleavage products. Peak e, the major peak at 30 min, also contained a peptide almost identical to the previously studied AChE 586-599 peptide (KAEFHRWSSYMVH versus AEFHRWSSYMVHWK). Analysis of peaks a-h in all the incubation times revealed that CSR species appeared very early during digestion (from 2 min) with the number and abundance increasing with time ( Figure 2F). Species containing the highb propensity sequence YMVHW appeared to be the most abundant CSR species identified during T40/IDE digest ( Figure 2F, bottom of the table). Potential precursors to CSR species were detected, which increased and disappeared, coincident with the appearance of smaller peptides (e.g. WKAEFHRWSSYMVHWKNQFD versus HRWSSYMVHWK). Almost all CSR species, except RQWKAEFHRWSSY and KAEFHRWSSYMVH, were still present after a longer exposure to IDE (2 hours), indicating that the CSR species are resistant to and can survive further digestion by IDE ( Figure 2F).

IDE digests monomeric and oligomeric species of AChE 586-599
IDE has been proposed to be an amyloid scavenger recognizing structural b-rich folds found in amyloid forming peptides and has been shown to degrade soluble monomeric Ab [4]. Since AChE 586-599 is the only hAChE peptide reported to form amyloid fibrils, we investigated the capability of IDE to digest this peptide. IDE degraded AChE 586-599 and the initial cleavage sites of IDE on this peptide were between Ser 8 -Tyr 9 , and His 12 -Trp 13 ( Figure 3A). The identity of cleavage products is reported based upon AChE 586-599 synthetic peptide numbering (Ala 1 to Lys 14 )( Figure 3B). An initial IDE cleavage site was identical on both AChE 586-599 (Ser 8 -Tyr 9 ) and T40 (Ser 19 -Tyr 20 ), suggesting that one of the dominant binding motifs for IDE can be found within the AChE 586-599 portion of T40. Also, a significant proportion of generated products contained the YMVH motif of high propensity for b-sheet conversion.
Since IDE digests only monomeric forms of Ab [35,36], we examined the ability of IDE to degrade oligomeric AChE 586-599 . AChE 586-599 oligomers were recognized by monoclonal antibody (Mab) 105A, demonstrating b-sheet conformation, and migrated mainly as a smear of SDS-stable oligomers (4-36 kDa) ( Figure 3C, lane '-IDE'). In contrast to the lack of effect on Ab oligomers, IDE degraded AChE 586-599 oligomers ( Figure 3C), in an insulinsensitive manner (data not shown). Initially (5-80 min), IDE preferentially digested small and large oligomeric species (bottom and top of the smear). From 160 to 480 min, the 5 and 8 kDa species were partially degraded although persisted after 480 min incubation, suggesting resistance to IDE. These two oligomeric forms may correspond to trimers and pentamers according to their observed molecular weights (AChE 586-599 being 1.86 kDa).

NEP digests preferentially monomeric and oligomeric AChE 586-599
Along with IDE, NEP is an important Ab-degrading enzyme in the brain. The levels of non-amyloidogenic monomeric and dimeric forms of T40 remained unchanged after exposure to NEP ( Figure 4A). The control degradation of substance P confirmed the presence of NEP activity under these assay conditions. NEP digests both monomeric and oligomeric forms of Ab, hence it was appropriate to test its ability to degrade monomeric and oligomeric AChE 586-599 [11]. NEP degraded AChE 586-599 with complete digestion by 4 hours (Figures 4B and 4C). Although NEP targeted AChE 586-599 more broadly than IDE, some of the cleavages were conserved between the two enzymes (Ser 8 -Tyr 9 and His 12 -Trp 13 , AChE 586-599 numbering). Some peptides were generated by both NEP and IDE (e.g. AEFHRWSS), however NEP also allowed some larger peptide species to remain intact (e.g. AEFHRWSSYMVH)(see Figures 3B and 4C).
NEP degradation of AChE 586-599 oligomers was different and less efficient than IDE with a greater range of untargeted oligomers. Indeed, NEP only degraded the oligomer species at 10 and 16 kDa ( Figure 4D, arrows), which remained intact during 240 min incubation in the absence of NEP ( Figure 4D, lane '-NEP'). NEP-resistant 14-15 kDa species became apparent and might correspond to octamers according to their observed molecular weights. As previously observed for IDE, the 5 and 8 kDa species were resistant to digestion.
To assess the potential for interrelationships between IDE and NEP degradation, peaks e-f (see Figure 2C) from a 30 min IDE/ T40 digest were subjected to NEP digestion for 2 hours. Peaks e-f were selected for the largest variety of CSR species among non-CSR related peptides. NEP was capable of degrading to completion the T40 products generated by IDE ( Figure 4E).

T40/IDE digestion triggers conformational changes
Non-amyloidogenic T40 is a-helical either as a synthetic peptide or within hAChE [27,28,29], whereas AChE 586-599 adopts a bsheet conformation and self-assembles into amyloid fibrils [31,30]. We have shown that some of the cleavage products generated from T40 by IDE contained CSR species encompassing motifs predicted to have a propensity for conversion to b-sheet (e.g. AEFHR and YMVHW). Therefore, we investigated the conformation of the T40/IDE digestion products. Circular dichroism (CD) spectra were determined during digestion of T40 by IDE at various times ( Figure 5A). Prior to digestion, T40 displayed an ahelical spectrum with double minima at 209 and 222 nm. During IDE digestion, the ellipticity at 222 nm progressively decreased, suggesting a reduction in a-helices, accompanied by a shift of the minimum at 209 nm toward lower wavelengths that can be accounted by an increase of unordered structures (characterized by a single minimum below 200 nm). The negative ellipticity in the 210-220 nm indicated the presence of b-structures. After digestion, the spectrum intensity was reduced in the entire far-UV region, probably as the result of the formation of aggregates (see below). The total digest of T40 by IDE, as shown in Figures 2C  and 2E, resulted in a mixture of peptides, not all of which necessarily adopted a b-sheet conformation. This could explain the complex nature of the CD spectrum observed during digestion, which did not show a fully b2sheet conformation. Therefore, to extract secondary structure components from the complex spectrum, we used spectral deconvolution utilizing an algorithm validated against the largest reported reference protein dataset [37]. Quantification of the secondary structures revealed that up to 10 min digestion, b-structure content remained negligible (5.9%), increasing substantially to 45.5% at 30 min, concomitant with a decrease in a-structure (from 81% to 20.5%) and an increase in unordered structure (from 14.9% to 37.4%)( Figure 5B). These temporal secondary structure changes correlated with the progressive accumulation of CSR species or their precursors peaking at 30 min ( Figure 2E). When applied to three independent T40/ IDE digests of 30 min, analysis of the CD spectra by deconvolution showed large and unambiguous changes with a decrease in ahelical content from 80.8360.38% to 14.56610.45% (p,0.008), an increase in b-structure from 6.9060.89% to 53.70615.34% (p,0.03), and an increase in unordered structure from 15.3760.42% to 32.9367.22%.
T40/IDE generated CSR species are surface-active At acidic pH, AChE 586-599 remains monomeric and is not surfaceactive. In contrast, at neutral pH this peptide self-assembles into amyloid fibrils and reduces the surface tension of an air-water interface (measured by differential absorbance) in a similar manner to Ab [38,39]. The T40/IDE products a-k ( Figure 5C inset) were analyzed for surface activity. Only peaks a, b, d, e, f and g showed an increase of DOD at neutral pH (p,0.05), which demonstrated their pH dependant surfactant properties ( Figure 5C). This is comparable to a reduction of surface tension from ,72 N.m 22 at low pH to ,50 N.m 22 at neutral pH. All of these peaks contained CSR species. Products in peaks e-f, that contained the largest variety of CSR species including YMVHW (with high b-propensity) as a major species, had the biggest effect on surface tension upon neutralization. Peaks c and h were not surface-active at neutral pH, which may be explained by the very low levels of CSR species represented. For most of the surfaceactive products (peaks a, b, e, f and g), the effect was greater than seen with 50 mM AChE 586-599 . Peaks i, j and k did not contain CSR species and did not exhibit surfactant properties upon neutralization. Thus, some of the peptides generated by IDE digestion of T40 share with synthetic AChE 586-599 the unusual characteristic that their surface tension effects are strongly pH dependent, which appears to be linked with the presence of CSR species. Moreover, these effects upon neutralization were not solely due to the presence of hydrophobic or aromatic amino acids in their sequences since some non surface-active peaks (e.g. peak c) contained as many or more hydrophobic and aromatic residues as some surface-active peaks (e.g peak a).

T40/IDE digestion promotes Ab fibrilization and amyloid protofibril formation
Having established that T40/IDE digestion produced CSR species, some of which possessed surfactant activity and changed conformation from a-helical to b-structures, we examined their ability as heterologous seeds to promote Ab fibrilization. The quantity of Ab fibrils was determined by changes in thioflavin T (ThT) fluorescence emission. For Ab heterologous seeding, we used peptide seeds instead of monomers, experimental conditions that were identical to Diamant et al and were chosen for direct comparison with this previous study reporting the effect of hAChE on Ab fibrilogenesis [40]. However, to preclude artifacts due to ThT binding to peptide oligomers/seeds themselves formed during IDE digests of T40, the values for peptide seed-ThT (no Ab) were subtracted from all ThT assays (with Ab). Thus, while it is possible that some experimental variation may be due to variable seed formation, these variations were accounted for in the statistical analysis of the experiment, which was based on three independent T40 digest experiments with replicates within each experiment. Ab showed an increase in fluorescence after ,42 hour nucleation process (lag phase) (Figures 6A and B). The addition of equimolar ratio of the non-amyloidogenic T40 (15 mM) to Ab did not cause any changes in the lag phase or the apparent rate or the plateau height of Ab fibril formation. This result confirmed that T40 did not affect fibrilogenesis even when it was present at a 1:1 ratio with the fibrilizing substrate and validated the use of T40 as a negative control. In contrast, equimolar ratio of AChE 586-599 seeds (15 mM) to Ab reproducibly reduced the Ab lag phase by 11.6 fold (p,0.004), which confirmed the use of AChE 586-599 as a positive control. Even at 1.5 mM (which represented 3.3% by mass of the total peptide), AChE 586-599 peptide reduced the Ab lag phase from 42 hours to 1360.9 hours (p,0.006)(data not shown). The total T40/IDE digest (,10 mM starting T40) also reproducibly reduced the Ab lag phase by 1.8 fold (p,0.006). Moreover, the total T40/IDE digest also increased the apparent rate of fibril formation (from 48615 for Ab alone to 8276288 fluorescence units min 21 )(p,0.043) and the plateau height (from 64936930 for Ab alone to 1233361085 fluorescence units)(p,0.003). Each individual peak (a-h, see Figure 2C) from the T40/IDE digest also reduced the Ab lag phase (p,0.005, p,0.05 for peak h) and some peaks increased the plateau height (p,0.012 for peaks e-f and p,0.0012 for peak d) (Figures 6C and D). Peaks g-h containing the fewest CSR species were the least efficient in lag phase reduction and plateau height increase. For the largest peptide fragments in any of the individual digest peaks, the maximum mass ratio of seed to soluble Ab was 5.1% and for most peaks the mass ratio was lower. Thus, the measured kinetics show that reduced lag phase, increased rate and elevated plateau are all significant consequences of seeding with IDE digests of T40 (and that undigested T40 has no significant effect on any of these parameters). The mechanisms involved in the Ab increase of both rate of fibrilization and plateau level after seeding with IDE digests of T40 are unknown. However, these results are entirely consistent with the reduction in lag phase and increase of both rate and plateau height observed when hAChE is present during Ab fibrilogenesis [40].
The aggregation status of the total T40/IDE digest was examined by negative staining electron microscopy, which revealed predominantly spherical structures (diameter 4-14 nm) (Figure 7). Also observed were annular protofibrils (outer and inner diameters of 11 and 3 nm respectively), and ''rods'' (9 nm wide, 24-29 nm long) with some appearing as ''beaded chains'' composed of spherical subunits. All these observations are consistent with the presence of amyloid precursors (oligomers) [41,42,43].

DISCUSSION
A number of proteases are involved in Ab clearance in the brain, including two metalloproteases; IDE and NEP, with IDE digesting monomeric Ab and NEP, oligomeric Ab [10,3,4,11]. hAChE promotes Ab fibrilization and deposition in senile plaques but the hAChE domain involved remains uncertain [44,22]. Therefore it is important to understand the mechanisms for the formation of hAChE amyloid species that could increase Ab fibril formation during AD pathogenesis. We have examined the ability of IDE and NEP to generate amyloid-forming species from the exposed and non-amyloidogenic hAChE T40 oligomerisation domain and the consequences for heterologous seeding of Ab, a peptide thought to be the key player during AD. IDE and NEP may not be the only or major enzymes involved in the formation and/or clearance of hAChE species. Nonetheless, these two enzymes, which have already been implicated in turnover of amyloidogenic peptides, are present and active in the relevant compartment (namely the extracellular space of the brain) to attack an exposed part of a substrate (hAChE) that is also present and known to associate with plaques. hAChE T40 is exposed in the monomer and appears to remain vulnerable to proteolysis even in assembled tetramers [33,32]. An outcome of such proteolytic attack may be hAChE fragments that are able to interact with Ab, promoting fibril assembly. IDE degraded the non-amyloidogenic and a-helical T40, to generate CSR species, but also b-sheet forms of CSR. Upon binding to IDE, substrates undergo drastic conformational changes from ahelical to b-strands [45]. Thereafter, cleavage occurs at the bstrand sites [45]. Thus, the cleavage of the a-helical T40 by IDE implies that part of T40 is able to convert to b-conformation, which is supported by the identification within T40 of a unique predicted CSR (W 585 to K 599 ) with high propensity for conversion to non-native (hidden) b-strand [30]. Although IDE preference for cleavage is after aromatic and hydrophobic residues, multiple alignments for substrate binding can occur [46]. All IDE cleavage sites on T40 or AChE 586-599 are consistent with the preferences previously described [46]. The T40/IDE cleavages indicate both distinct and secondary cleavages of an initial product. Indeed, CSR species terminating at Ser 19 may have been generated by an initial cleavage at Ser 19 -Tyr 20 of T40 followed by a second cleavage at the N-terminus. However, species encompassing the T40 N-terminus and terminating after Tyr 20 may have been generated by distinct cleavage events. Both IDE and NEP appeared to digest primarily from the C-terminus of either T40 (IDE) or AChE 586-599 peptide (IDE and NEP). Although NEP hydrolyses Ab at specific sites, the enzyme digested AChE 586-599 at almost every peptide bond in a non-specific manner. In contrast to Ab, IDE acted on both monomeric and oligomeric AChE 586-599 species (compared with only monomeric species for Ab), whereas NEP acted mainly on monomeric species (compared with monomeric and oligomeric species for Ab) [10,11]. The fact that IDE was able to digest some oligomeric species of AChE 586-599 is not inconceivable since the enzyme was shown to degrade substrate above 50 amino acids [47,45]. However, such big substrates are less likely to be entrapped by IDE catalytic cleft and their degradation would be much slower. This could explain the 'lack' of efficiency of IDE towards some AChE 586-599 oligomers and the relatively slow digestion process when compared to T40 or monomeric AChE 586-599. Very few cleavage sites on Ab and hAChE peptides are in common and are as follows; His 12 -Trp 13 and Trp 13 -Lys 14 for IDE, and Ser 8 -Tyr 9 for NEP (AChE 586-599 peptide numbering)( Figure S2 supporting information).
Significant differences were observed between the degradation capability of IDE and NEP. In contrast to NEP, IDE digested T40 and a bigger variety of AChE 586-599 oligomers. However in conditions that allowed complete degradation of the monomeric AChE 586-599 species, small AChE 586-599 oligomeric species were slowly digested by IDE and untouched by NEP and some AChE 586-599 oligomers were more resistant to degradation by both IDE and NEP. IDE was also more efficient (1 IDE:1200 peptides versus 1 NEP:52 peptides). One could postulate that IDE independently mediates the formation of CSR species and the clearance of soluble and some insoluble aggregates, whereas NEP could be involved in the clearance of the newly formed soluble CSR species. Indeed, we have demonstrated that CSR species generated from the T40 by IDE are a substrate for NEP. Therefore, if modification of hAChE by IDE occurs in vivo, both IDE and NEP deficiencies could alter the brain levels of CSR species and increase the risk of oligomerisation and fibril deposition. Indeed, reduced levels of these enzymes would still allow the formation of pathological species (albeit at a reduced rate) that could assemble into insoluble oligomers and fibrils, whereas clearance of oligomers (which is already inefficient when these enzymes are abundant) would be severely compromised. The small oligomer species resistant to both IDE and NEP could be initiating-agents in early pathological reactions. IDE activity was decreased in soluble fractions from the brain of AD patients compared to normal control brains [48], which suggests that a decrease in enzyme activity could be responsible for the increased accumulation of pathologic amyloid peptides during AD. It was also reported that in AD brain, IDE is less effective because it is oxidized [49]. NEP mRNA levels were reduced in amyloid affected areas of sporadic AD brain, which could be the cause of Ab deposition [50].
IDE cleavage of the non-amyloidogenic T40 triggered a conformational change from ato predominantly b-structure, a transition that was also observed for other amyloid proteins (e.g. insulin and prion protein) [51,52]. Although native Ab is unordered, a-helix formation is a key step for fibril assembly [53]. Several amyloid proteins, a-helical in the native state, contain stretches of a-helix in places that are predicted to form b-strand. These helices could form b-strands by unfolding into intermediates less likely to refold into a helical conformation [54]. Moreover, helical aggregates may convert short-range to long-range interactions triggering a b-transition [55]. Computational identification of non-native (hidden) b-strand propensity in protein sequences has predicted the minimal amyloidogenic fragments for Ab and asynuclein [30]. When applied to T40, two regions with a strong propensity for conversion to b-strand were recognized, YMVHWK the strongest and AEFHR more weakly (Figure 1), which is consistent with the fact that fragments (peaks e-f) containing these regions are the most surface-active. In the larger context of protein aggregation, our results support the importance of gatekeeper residues in preventing the conformational switch that leads to the formation of b-sheets and amyloid fibrils [56]. Evolutionary pressure may have sequestered CSR within T40 to maintain conformational integrity and to protect against deleterious misfolding.
Several intermediates of amyloid fibril formation have been identified with the first stage being spherical structural units that could associate to form beaded protofibrils from which fibrils nucleate and elongate [41,43,1]. Spherical oligomers from Ab and a-synuclein specifically increase membrane conductivity [57]. Ab (Arctic variant) and a-synuclein also form annular spheres resembling bacterial pore-forming toxins [42,43]. The formation of pores in membranes may be one mechanism for the cytotoxicity seen in neurodegenerative diseases. In the case of Ab, there is still controversy regarding the identity of the pathological species (monomers, small oligomers, large oligomers or fibrils) [58,59]. However, several recent studies suggested that cognitive dysfunction correlates better with cortical levels of soluble oligomeric rather than insoluble (fibrilar) Ab [60,61,62]. In our case, the surfactant and amyloidogenic CSR species generated by IDE might lead to an increased concentration of potentially toxic oligomeric forms, whether these are homo-oligomers of b-strand CSR species or hetero-oligomers also containing Ab. The existence of such heterologous interactions is established in this study by the demonstration that the lag phase of Ab assembly is reduced by the CSR species. However, the molecular details of the heterologous interaction remain to be characterized.
IDE catalysis generated CSR species that are highly dependent on pH for their surface-tension activity, which appeared to be solely the consequence of the presence of some CSR species and could not be explained by only the presence of hydrophobic or aromatic residues within the peptide sequences. In the case of Ab, the surfactant properties were proposed to be detergent-like and linked to lysosomotropic activity resulting in accumulation of Ab in lysosomes, release of lysosome contents and cell death [38]. AChE 586-599 and CSR species from the T40/IDE digest were also surface-active. Thus, one could propose that both AChE 586-599 and CSR species are detergent-like and could potentially permeabilize membranes. For AChE 586-599 , surfactant activity was directly linked to the threshold concentration for fibril formation [39], suggesting that surface-active T40 generated CSR species might also assemble into higher oligomeric species. Indeed, like other amyloid proteins, the T40/IDE digest formed amyloid protofibrils (spheres, annular spheres and ''beaded rods''), which could contribute to neuronal toxicity in AD. Thus unlike the islet amyloid polypeptide for example, of which nested amyloidogenic peptides that formed fibrils were originated from an already amyloidogenic parent peptide [63], IDE digestion converted T40 (a-helical and non-amyloidogenic) into internal fragments that form b-sheet and are amyloidogenic.
hAChE was reported to increase Ab fibrilization, an effect that was not mediated by isolated T40 [40]. Under the same experimental conditions, we confirmed that the non-amyloido-genic T40 does not promote Ab fibril formation. In contrast, products generated from a T40/IDE digest and AChE 586-599 seeded Ab, an effect measured as a reduction in lag phase in a fibril formation assay (by 2 and 11 fold respectively) and the T40/IDE digest also increased the rate of Ab fibril formation (by 17 fold). This result underlines the importance of the T40 CSR, which is included within the T40/IDE products and AChE 586-599 , and provides an insight into the identity of an AChE domain that may cooperate with Ab during AD pathogenesis.
In conclusion, we have clearly demonstrated that IDEdependent cleavage of the non-amyloidogenic hAChE oligomerisation domain leads to a conformational switch to b-structure and liberates surface-active peptides that assemble into amyloid protofibrils and seed the aggregation of hetero-oligomers comprised of Ab and CSR species. Therefore, IDE-mediated formation of amyloidogenic hAChE fragments may provide useful targets for the identification of fibrilogenic hAChE-derived species in the brain, which has so far been impossible due to the lack of information about relevant hAChE species. While the CSR peptides themselves may offer a potential target in the struggle to prevent abnormal protein aggregation in the brain, our results suggest that simply increasing IDE and NEP activity may not be as beneficial as anticipated. Furthermore, the role of the nonamyloidogenic AChE T40 domain in heterologous seeding interactions may have to be re-appraised in light of the proteolytic events reported here. To our knowledge, this study represents the first evidence of heterologous amyloid seeding by a proteolytic fragment from another protein. Such seeding could represent a novel initial trigger for not only AD but also other neurodegenerative diseases sharing common characteristics, in which the abundance of the major amyloidogenic specie may not be the only important factor.

Preparation of AChE 586-599 oligomers
AChE 586-599 oligomers were covalently cross-linked by photoactivation using the photo-induced cross-linking of unlabelled proteins (PICUP) [64]. In the dark, 0.3 nmoles of AChE 586-599 and 94 mM tris-bipyridyl ruthenium salts in 500 mM NaH 2 PO 4 buffer pH 7.2 were incubated with 1.9 mM ammonium persulfate (30 sec). The reaction mixture was exposed to light ( Lyophilized HPLC separated T40/IDE products were used for  surface tension and seeding experiments. 60 mM substance P or T40 were incubated with or without 1.2 mM NEP (4 hours), stopped (0.5% trifluoroacetic acid) and subjected to RP-HPLC.

RP-HPLC
Reaction products were resolved by RP-HPLC with a Sephasil C4 column (5 mm, 4.66250 mm; Amersham Biosciences, UK) using a 5-95% linear gradient of acetonitrile in 0.1% trifluoroacetic acid over 25 min (flow rate 1 ml/min). The eluent was monitored by UV absorption at 280 nm.

Western-blot
Nitrocellulose membranes were blocked with 5% (w/v) non-fat milk in PBS and incubated with KD69 anti-T40 antiserum or with the Mab 105A recognizing AChE 586-599 in b-sheet conformation [28], followed by anti-rabbit or anti-mouse IgG conjugated to horseradish peroxidase (HRP). Products were visualized by enhanced chemiluminescence.

Surface tension measurement
Analyses were performed in a 96-well plate format, as described [39]. Briefly, HPLC purified/lyophilized products were resuspended in 80 mL 200 mM sodium acetate pH 3 and surface tension measured at 450 nm (BMG Polarstar plate reader) before and after neutralization (20 mL 1M NaH 2 PO 4 , pH7.2). DO-D = (OD offset position -OD central position ) neutral pH -(OD offset position -OD central position ) acidic pH . At least three independent assays were performed and analyzed with the two-sample t-test.

Circular dichroism
CD-spectra were recorded from 250 to 190 nm at 20uC in a quartz cuvette (1 mm path length) using a Jasco J-720 spectropolarimeter. The spectrum of 100 mM T40 in buffer A was recorded before addition of 83 nM IDE. The reaction mixture was incubated at 37uC for various times, cooled briefly on ice before spectrum recording. The mean spectra of multiple scans (scan speed of 50 nm min 21 and response time 4 sec) were deconvoluted with Selcon3 [37].

Seeding experiments
HPLC separated/lyophilized T40/IDE products were treated as for surface tension measurement, and then re-lyophilized, re-suspended in buffer B and incubated for 2 hours at 37uC. Control experiments showed that there was no carry over of IDE activity under the sample preparation conditions (RP-HPLC and lyophilizations). The quantity of products was normalized to the height of the RP-HPLC peaks. AChE 586-599 seeds were prepared by incubating 200 mM AChE 586-599 in PBS for 3 hours under continuous agitation. Individual T40/IDE products or the total digest (,10 mM starting T40), 15 mM AChE 586-599 seeds and 15 mM T40 were dispensed in a 96-well plate (black wall, clear bottom; Greiner, UK) with 15 mM Ab and 165 mM ThT in PBS. 15 mM Ab in buffer B was used as a control for fibrilization. ThT fluorescence (excitation 450 nm, emission 480 nm) was measured at 37uC every 20 min, with 5 min shaking after every measurement, on a BMG Polarstar plate reader. The values of peptide-ThT were subtracted from the values of peptide-Ab-ThT. At least three independent assays were performed and analyzed with the two-sample t-test.

Electron microscopy
T40/IDE digest (30 min), as prepared for the seeding experiment, was adsorbed onto Formvar-coated 400 mesh copper grids, air dried, washed with distilled water, negatively stained with 2% aqueous uranyl acetate and viewed with a Zeiss Omega 912 microscope. Figure S1 Cleavage map after complete digestion of T40 by IDE. 60 mM T40 was incubated with 50 nM IDE for 30 min at 37uC. The products were loaded onto a C4 reverse-phase HPLC column and separated using a 5-95% linear gradient of acetonitrile. HPLC product peaks were collected manually and their identities were determined by mass spectrometry. The fulllength T40 sequence is shown at the top of the complete map. CSR species are shown in white letters on a black background.