Proteolytically Inactive Insulin-Degrading Enzyme Inhibits Amyloid Formation Yielding Non-Neurotoxic Aβ Peptide Aggregates

Insulin-degrading enzyme (IDE) is a neutral Zn2+ peptidase that degrades short peptides based on substrate conformation, size and charge. Some of these substrates, including amyloid β (Aβ) are capable of self-assembling into cytotoxic oligomers. Based on IDE recognition mechanism and our previous report of the formation of a stable complex between IDE and intact Aβ in vitro and in vivo, we analyzed the possibility of a chaperone-like function of IDE. A proteolytically inactive recombinant IDE with Glu111 replaced by Gln (IDEQ) was used. IDEQ blocked the amyloidogenic pathway of Aβ yielding non-fibrillar structures as assessed by electron microscopy. Measurements of the kinetics of Aβ aggregation by light scattering showed that 1) IDEQ effect was promoted by ATP independent of its hydrolysis, 2) end products of Aβ-IDEQ co-incubation were incapable of “seeding” the assembly of monomeric Aβ and 3) IDEQ was ineffective in reversing Aβ aggregation. Moreover, Aβ aggregates formed in the presence of IDEQ were non-neurotoxic. IDEQ had no conformational effects upon insulin (a non-amyloidogenic protein under physiological conditions) and did not disturb insulin receptor activation in cultured cells. Our results suggest that IDE has a chaperone-like activity upon amyloid-forming peptides. It remains to be explored whether other highly conserved metallopeptidases have a dual protease-chaperone function to prevent the formation of toxic peptide oligomers from bacteria to mammals.


Introduction
The formation of amyloid is a process by which disordered peptides or specific regions of partially unfolded proteins selfassemble to yield fibrils with a typical morphology, a cross-b structure and affinity to dyes such as Congo red. A growing number of proteins capable of forming amyloid are associated with human pathology, including high prevalence diseases such as type 2 diabetes mellitus (DM2) and dementia. In vitro, amyloid fibril formation follows a typical nucleation-dependent kinetics [1] that includes key variables such as critical concentration, nuclei formation, lag phase, exponential growth and steady state. One or more pathways may lead to the conversion of monomeric soluble proteins or peptides into amyloid fibrils [2,3,4]. Alternatively, ''off-pathway'' oligomeric products relative to amyloid formation may be formed after the initiation of the self-assembly process. This has been shown by using compounds that inhibit amyloid b (Ab) oligomer formation without affecting fibrillogenesis [5,6].
Recent work has raised the possibility that fibril formation may be protective to cells by stabilizing non-amyloid cytotoxic intermediates, as proposed for several neurodegenerative diseases [7,8,9]. Furthermore, amyloid fibrils may play a physiological role serving as a scaffold for the fast fibrillization of the Ma protein within melanosomes [10]. Yet, the existence of cytotoxic oligomers and protofibrils in vivo implies that at least one pathway (either ''on or off'' relative to fibril formation) of proteins/peptides aggregation is likely a pathological process [9,11,12].
Ab, both in fibrillar and non-fibrillar forms, massively accumulates in the brain in sporadic and hereditary Alzheimer's disease (AD) and amyloid angiopathies [13,14,15]. Ab aggregates are capable of inducing dysfunction or death on neural, glial and vascular cells. Moreover, several coding missense mutations within or around the Ab sequence in its precursor protein, APP, are associated with familial forms of these diseases [16,17,18,19].
At the subcellular level, Ab generation has been tracked along the late secretory and endocytic pathways, consistent with the localization of the so called b and c secretases, major protease complexes that release Ab peptides from APP [20,21,22]. Under normal conditions, cells shed soluble Ab species to the extracellular compartment without membrane damage [23], yet the sites of their initial oligomerization under pathological conditions remain elusive. Although the bulk of Ab fibrils in the brain seem to be extracellular, recent studies support that Ab oligomers may be formed intracellularly [24,25,26]. Since self-assembly of Ab is dependent on monomer concentration and time, the overall rates of its production and removal in a given cellular compartment are key factors in determining the initiation of the amyloidogenic pathway(s).
Insulin-degrading enzyme (IDE), neprilysin (NEP) and endothelin-converting enzyme (ECE) are among the major Ab proteases with physiological relevance, as shown by gene knockout and transgenic animal models [27,28,29]. Alternatively, the aggregation of Ab in vivo may be modulated by its interaction with molecular chaperones. Studies in cell cultures, transgenic C. elegans and in vitro biophysical data, support that heat-shock proteins (Hsp) 70-90 and small heat shock proteins (sHsp) with chaperone-like activity interact with Ab and are capable of reducing fibril formation [30,31]. In addition to these canonical chaperones, Ab may interact extracellularly with other proteins with chaperonelike activity. Three secreted glycoproteins, namely, clusterin, haptoglobin and a 2 -macroglobulin (a 2 M), are known to have ATP-independent chaperone-like activity in vitro [32,33,34,35,36]. Among these, the highly stable interaction of Ab with clusterin, a multi-funcional protein, leads to diverse effects, depending on molar ratios that range from prevention of aggregation to enhancement of fibril formation. Interestingly, HSPs, sHSPs and clusterin expression is increased in AD brains and co-localize with Ab deposits in neuropil and vessels or in surrounding astrocytes, supporting that they participate in some way in the amyloidogenic process in vivo [37,38,39,40].
IDE is a highly conserved Zn 2+ metallopeptidase ubiquitously expressed in tissues and at the subcellular level. IDE hydrolyzes insulin and Ab in vivo [41] as well as short bioactive peptides with amyloidogenic potential such as glucagon, atrial natriuretic peptide and amylin (reviewed in [42]). We have recently shown that natively folded IDE is capable of forming a highly stable complex (IDE-AbSCx) with Ab in vitro and in vivo. The shortest competent Ab fragment for IDE-AbSCx formation contained the central LVFF sequence, which seems to be crucial for amyloid formation [43,44,45]. Interestingly, this sequence is included in the Ab segment that acquires b-conformation inside the IDE chamber close to the active site [46].
In this study we present results supporting that the catalytically inactive mutant IDEQ drastically modifies the Ab self-assembly pathway, halts amyloid fibril formation and yields non-toxic aggregates through a chaperone-like activity.

Results
Native IDEQ is highly efficient in forming a stable complex with Ab1-42 We first confirmed that there were no structural differences between purified IDE wild type (IDEwt) and IDEQ, as determined by static light scattering (SLS) and circular dichroism (CD) ( Figure S1). The loss of enzymatic activity of IDEQ was assessed by four different methods: 1) The absence of fluorescent signal increase after co-incubation with the fluorogenic substrate V; 2) The lack of significant degradation as determined by trichloroacetic acid (TCA) precipitation of 125 I-insulin (,1.5% residual activity); 3) Western blots with anti-Ab and 4) Enzyme-linked immuno-sorbent assay (ELISA) quantification of monomeric Ab1-42 that remained measurable after 5 days of incubation with IDEQ (94.4%), IDEwt (10.4%) and IDEwt with EDTA (88.9%) at a 1:10 molar ratio (enzyme:Ab) ( Figure S2). In a prior work [44] we showed that co-incubation of Ab1-28 with IDEwt yielded higher levels of IDE-AbSCx when the protease was inhibited with EDTA. Thus, we tested if natively folded IDEQ produced a similar effect. A time-course of IDE-AbSCx formation was determined by quantifying the immunoreactivity of the ,120 kDa band with anti-Ab by Western blot. Although the half-times of IDE-AbSCx formation were similar (IDEQ = 32min; IDEwt = 26 min), the yield of IDEQ-AbSCx was ,3 -fold higher relative to IDEwt (Figure 1). To test whether Ab was interacting with IDEQ substrate-binding site, as previously shown for IDEwt [44], competition was performed with insulin. Preincubation with increasing concentrations of insulin decreased the formation of IDEQ-AbSCx reaching a final value of ,22% of the maximum formed in the absence of insulin ( Figure 1). The fact that prevention was not complete may be attributed to several factors including the k d (,8 mM) for insulin dimerization (the size of an insulin dimer is not compatible with the volume of the IDE chamber), the non-equilibrium reaction that leads to IDEQ-AbSCx formation and that binding site may not be fully shared [44]. IDEQ has no effect upon insulin conformation and its activity on cultured cells Although insulin does not aggregate under physiological conditions and low concentrations, it has a high affinity for IDE and undergoes partial unfolding inside the IDE chamber before hydrolysis [47]. Having shown that IDEQ was natively folded and had minimal residual activity, we tested its possible effect on insulin conformation or aggregation. To address this question, two experiments were performed. Firstly, insulin was incubated alone or with IDEQ at 1:100 (enzyme:insulin) for 24 h and secondary structure determined by CD. Signal intensities were similar and no differences were seen in the two a-helix minima of insulin with or without IDEQ at dead time and after 24 h of incubation ( Figure 2). Secondly, human astrocytic U-87 cells were treated with insulin with or without pre-incubation with IDEQ. The activation of insulin receptor was assessed by quantifying Akt phosphorylation. Insulin pre-incubated with IDEwt was used as a positive control for the loss of insulin activity due to hydrolysis. Insulin treatment increased phospho-Akt levels relative to total Akt. Pre-incubation of insulin with IDEQ did not result in a significant loss of hormone activity, while pre-incubation with IDEwt abolished insulin receptor activation ( Figure 2). When cells were previously treated with wortmannin, an inhibitor of PI3K, the increase in Akt phosphorylation was abrogated, indicating the specificity of insulin pathway activation in the insulin-IDEQ experiment. These results support that IDEQ, despite binding insulin, does not induce a conformational change on the hormone or promotes its aggregation.

Inverse relationship between hydrolysis and IDE-AbSCx formation as a function of pH
The physical association between IDE and Ab in vivo seems to be ubiquitous, including intracellular organelles with acidic pH [48,49]. Therefore, we tested the effect of a range of pH in the formation of IDE-AbSCx. The activity of IDEwt was drastically reduced in vitro below pH 6 while the formation of IDEwt-AbSCx increased from low levels at neutral pH reaching a maximum at pH 5.5, likely related to reduced degradation of Ab ( Figure S3). Instead, IDEQ was capable of forming comparable amounts of IDE-AbSCx across the range of pH studied, as shown in Figure 1. These and our previously reported results prompted us to explore a functional significance of the stable interaction between IDE and Ab unrelated to peptide hydrolysis.

IDEQ reduces the size of Ab aggregates
The effect of IDEQ upon Ab self-assembly was first studied by dynamic light scattering (DLS), which allows a rough estimate of the size of aggregates in solution. Ab1-42 showed species with increasing hydrodynamic diameter (D H ) and intensity of scattered light with peaks around 1000 nm after 6.5 days of incubation ( Figure 3). Although the total amount of IDE in the brain is likely much higher than monomeric Ab [50,51], we used IDEQ at substoichiometry to minimize non-specific hydrophobic interactions in vitro. Ab1-42 incubated in the presence of IDEQ at 1:10 molar ratio (enzyme: Ab) resulted in the detection of two distinct populations: one around 10-13 nm, as expected for IDEQ dimers, and a second population with a peak around 150 nm. Notably, only two discrete populations with D H of 150 and 400 nm were seen at the end of incubation, as opposed to Ab1-42 alone ( Figure 3). Similar results were obtained using the highly amyloidogenic ''Dutch'' variant of Ab peptide 1-28 in which Glu is replaced by Gln at position 22 (Ab28Q) (not shown).

IDEQ abolishes the formation of Ab fibrils
To determine the morphology and size of the aggregates detected by DLS, samples were analyzed by transmission electron microscopy (TEM). After 6.5 days of incubation, Ab1-42 ( Figure 3) and Ab28Q ( Figure S4) formed typical amyloid fibrils. The effect of co-incubation with IDEQ at 1:10 (enzyme: Ab molar ratio) was remarkable, with a complete absence of amyloid fibril formation by Ab1-42 or Ab28Q. In the presence of IDEQ, much smaller structures were seen. These ranged from thin, short rods or protofibrils to coalescent annular structures with a diameter of ,5 nm. Similar structures were seen when IDEQ was incubated alone, although the tendency to form large aggregates was higher in the Ab-containing samples ( Figure 3). When IDEwt inhibited with EDTA was incubated with Ab1-42, non-fibrilar products were seen, reinforcing that blockage of amyloid formation was unrelated to proteolysis ( Figure 3). To test if the lack of amyloid formation by Ab in the presence of IDEQ was due to a retardation of the fibrillization pathway, samples were co-incubated for 30 days. Amyloid fibrils with the typical features were observed with Ab1-42 alone. In contrast, only small aggregates were found in the presence of IDEQ ( Figure 3). Typical amyloid fibrils were formed by Ab28Q in the presence of bovine serum albumin (BSA) at the same molar ratio ( Figure S4), indicating a specific effect of IDEQ rather than a non-specific, hydrophobic peptide-protein interaction. Overall, DLS and TEM results indicate that in the presence of IDEQ at sub-stoichiometry, Ab forms small aggregates that lack the morphological features of amyloid fibrils. In addition, by using the shorter Ab1-28Q, we showed that the hydrophobic transmembrane C-terminal region of Ab was not necessary for the underlying peptide-enzyme interaction that halted amyloid formation.

Kinetics of Ab aggregation in the presence of IDEQ and effect of ATP
To further characterize the effect of IDEQ on Ab1-42 selfassembly, the kinetics of aggregation were followed by scattering at 340 nm. Ab1-42 self-assembly showed an initial lag phase of ,12 h, followed by exponential growth of 24-36 h until a steady state was reached after 48-72 h (the shorter periods of time for these phases as compared with the rest of the experimental conditions was due to the repeated agitation every 30 min, as described in Materials and Methods). Typical amyloid fibrils were obtained at steady state, as assessed by TEM (Figure 4). At a molar ratio of 1:10 (enzyme: Ab), IDEQ slowed the rate of aggregation and the scattered intensity. This effect was concentrationdependent and it could still be observed at a substoichiometry (IDEQ: Ab) of 1:200. Interestingly, co-incubation with ATP, which binds IDE and modulates its conformation and catalytic activity upon short substrates [52,53], showed an increase in the anti-aggregation effect ( Figure 4). Such enhancement was independent of ATP hydrolysis as confirmed by using the nonhydrolysable ATP-PNP (data not shown). Moreover, it could not be attributable to an increase in the minimal residual proteolytic activity of IDEQ ( Figure S2). To further characterize the products in the presence of ATP, these were visualized by TEM at steady state. A homogeneous population of protofibrils of ,10 nm was seen in the Ab samples co-incubated with IDEQ and ATP in contrast to the typical amyloid fibrils formed by Ab alone (Figure 4). The incubation of IDEQ and Ab at a 1:1 stoichiometry did not yield a lower absorbance than IDEQ-Ab at 1:10 with ATP (data not shown). However, this result was not unexpected in light of the rather large aggregates seen under TEM.
IDEQ cannot dissociate Ab fibrils and Ab aggregates formed in its presence are incompetent for seeding Ab1-42 was incubated alone for 48 h and then fresh IDEQ was added at a 1:10 molar ratio. Turbidity measurements after IDEQ addition were similar to those of Ab1-42 alone, suggesting that no disaggregation took place and that IDEQ was not capable of dissociating amyloid fibrils ( Figure 4). We next addressed whether Ab aggregates formed in the presence of IDEQ were competent to seed the self-assembly of monomeric Ab. An aliquot of Ab1-42 aggregated in the presence of IDEQ (at 1:10 molar ratio, enzyme:peptide) was taken at steady state, sonicated and added to a fresh Ab1-42 solution. As expected, seeds obtained by sonication of fibrillar Ab1-42 shortened the lag phase and increased the light scattering plateau. Strikingly, the ''seeds'' obtained from the co-incubation of Ab1-42 with IDEQ had no effect on the aggregation rate when compared with Ab1-42 selfassembly alone, as shown by the super-imposable scattering curves in both conditions ( Figure 4). When end-products obtained from steady state were analyzed by TEM, typical amyloid fibrils were observed in case of Ab1-42 with or without seeds, yet it tended to be longer in the former (not shown). These results suggest that IDEQ not only modifies the kinetics of Ab self-assembly but that non-fibrillar Ab species formed in the presence of IDEQ lack the ability to act as templates for monomeric Ab to follow the amyloidogenic pathway.
Differential solubility and secondary structure of Ab species formed in the presence of IDEQ To study the effect of IDEQ upon Ab1-42 solubility, samples were co-incubated for 5 days, centrifuged at low speed and analyzed by Western blot. Total immunoreactivity of Ab1-42 in the supernatant was 2-fold higher after incubation with IDEQ as compared to Ab1-42 alone. Accordingly, a higher proportion of Ab1-42 partitioned to the pellet in the absence of IDEQ ( Figure 5). Major differences in Ab pellets were accounted for by trimers, tetramers and high molecular mass SDS-resistant oligomers. To determine the secondary structure of soluble Ab oligomers formed alone or in the presence of IDEQ, Ab1-42 was incubated at 1:300 molar ratio (to minimize enzyme signal) and far UV-CD spectra analyzed. The effect of IDEQ in preventing amyloid fibril formation at this molar ratio was assessed by TEM ( Figure S5). IDEQ alone showed a spectrum with minima at 208 and 222 nm, consistent with its reported structure [46]. In contrast, the spectra of fresh Ab1-42 alone or Ab1-42 with IDEQ were mainly unstructured, in accordance with the negligible contribution of signal by IDEQ ( Figure 5). After 6 days of incubation, samples were centrifuged to separate large aggregates and far UV-CD spectra obtained from the supernatants. Soluble Ab1-42 in the absence of IDEQ showed a minimum at ,213 nm, but the bulk of the peptide was unstructured with one third of its original intensity, consistent with the results obtained by Western blots ( Figure 5). Notably, Ab1-42-IDEQ spectrum was at least 2-fold more intense than Ab1-42 alone displaying minima at 208 and

Ab1-42 aggregates formed in the presence of IDEQ are not neurotoxic
The Ab aggregates seen by DLS and TEM together with the CD data raised the possibility that co-incubation with IDEQ resulted in the generation of toxic Ab non-fibrillar species. When Ab was incubated alone for 4 days, mainly spherical oligomers, rods and protofibrils were seen under atomic force microscopy (AFM) (Figure 6). In the presence of IDEQ, aggregates were more heterogeneous and slightly larger as compared to Ab alone ( Figure 6). Primary neurons exposed to Ab1-42 aggregates showed a remarkable retraction and nicking of their processes and a ,40% reduction in viability. In contrast, cells exposed to Ab1-42 aggregates obtained from co-incubation with IDEQ had no morphological alterations and were ,95% viable as compared with control cells (Figure 6). Co-incubation of Ab1-42 with BSA showed a trend toward lower toxicity although it was not statistically different from that induced by Ab1-42. To rule out a possible trophic effect upon neuronal viability of IDEQ unrelated to its effect on Ab1-42 aggregation, cells were exposed to IDEQ alone and showed no significant differences with vehicle treatment.

Discussion
The catalytically inactive IDEQ was capable of forming a highly stable complex with Ab1-42 more efficiently than IDEwt. In agreement with our prior work using IDEwt, competition with insulin suggested that at least part of the substrate binding site was involved [44]. IDEQ-AbSCx was generated in a broad range of pH, including that compatible with cellular secretory and endocytic vesicles in which both IDE and Ab are present in vivo [49,54]. These properties of IDEQ suggested another inherent function of IDEwt unrelated to hydrolysis. Purified IDEwt aggregates at pH,6 after long incubation times required for full Ab self-assembly and retains a ,10% of activity. Thus, to rule out hydrolysis-dependent effects we used the inactive mutant IDEQ. It is important to remark that the ELISA assay is not capable of detecting Ab in a stable complex with IDEQ. Therefore, the residual activity of IDEQ (,6%) as determined by this method, is likely overestimated. At neutral pH, the size and morphology of Ab aggregates were drastically modified by IDEQ. Ab aggregates were smaller and did not form amyloid fibrils in the presence of IDEQ, as shown by DLS and TEM. Instead, small annuli, rods and protofibrils were found, even after 30 days of co-incubation, ruling out retardation or a purely kinetic effect upon Ab self-assembly. The time-course of turbidity experiments showed that 1) IDEQ at substoichiometry down to 1:200 was sufficient to slow down the lag and exponential phases with a lower turbidity at steady-state; 2) the addition of IDEQ at the end of the exponential growth phase was ineffective to prevent amyloid formation or to modify its kinetics; 3) sonicated end-products of Ab-IDEQ co-incubation were incompetent to induce the ''seeding'' of Ab monomers, indicating that they did not form amyloidogenic templates. These results are consistent with IDEQ exerting its effect at early stages of Ab aggregation, most likely before the exponential phase of amyloid fibril elongation. Interestingly, the presence of ATP at its k d enhanced the anti-amyloidogenic effect of IDEQ. This effect was independent of ATP hydrolysis and may be related with the reported stimulation of the catalytic activity of IDEwt upon very short peptide substrates by polyphosphate anions (see below). Western blots of insoluble and soluble aggregates showed a 2-fold increase of Ab species in the supernatant in the presence of IDEQ, including intermediate and high molecular mass SDS-resistant oligomers with a likely content of b structure. The potential neurotoxicity of these Ab oligomers and aggregates was ruled out by the exposure of primary neuronal cultures in which Ab, in the presence of IDEQ, yielded non-toxic species as assessed by neuronal viability and morphology.
Regarding the possible mechanism underlying the anti-amyloidogenic effect of IDEQ, we propose that it relies on a chaperonelike activity. Similar effects upon Ab aggregation have been described for Hsps 70/40, Hsp90, Hsp104 and small heat shock proteins (sHsp20) [30,31,55]. These effects include the blockage of the amyloid pathway resulting in non-fibrillar aggregates, the competence at substoichiometry, and their activity at early stages of the aggregation process, except for Hsp104 which shows activity at different stages [55]. In some cases, Ab oligomers formed in the presence of chaperones have been proven to be non-toxic to neuronal cell lines [31].
The three-dimensional structure of IDEQ crystallized with several substrates [46] including insulin and Ab1-40 displays features that are relevant to the interpretation of our findings: 1) the presence of a chamber that has the approximate volume of a native insulin monomer, enclosed by the amino-terminal half (domains 1 and 2) and the carboxyl-terminal half (domains 3 and 4); 2) the existence of a substrate binding ''exosite'' (,30 Å away from the catalytic residues) that contributes to guide the partial unfolding of substrates; 3) the stabilization of short b strands common to all substrates (not present in the free state in solution) at the ''exosite'' and at the catalytic site; 4) the alternation between closed and open states of the IDE chamber articulated upon a flexible loop connecting IDE halves which may be the rate limiting step for the catalytic cycle; 5) an inner acidic surface of domains 3 and 4 that contributes to stabilize the ''closed'' conformation in the absence of substrate and 6) the relatively slow Kcat (in the order of min 21 ) for amyloidogenic peptides [56,57].
Our initial observation of an IDEQ-Ab highly stable complex with an apparent 1:1 stoichiometry (after SDS treatment) and occupying at least part of the binding sites led us to speculate that after binding, a population of Ab alone or together with a partner sequence on IDE may provide a template for additional binding of Ab monomers or small oligomers. IDE-Ab complexes may then progress to form non-amyloid aggregates. Although the closed IDEQ chamber may be limiting in this mechanism, the binding of at least another Ab monomer would impose a steric hindrance for ''IDE closure'', facilitating the assembly process upon the IDEQ-Ab template. The major polyanion binding site has been recently proposed by crystallographic analysis to be located at a positively charged inner surface contributed by IDE domains 3 and 4 [53]. The enhancement of the anti-amyloidogenic effect by ATP may be related with a reduced electrostatic attraction between IDE halves, favoring the open state and the aggregation of Ab upon IDEQ-Ab templates. We propose that such aggregation takes place much faster (,30 min) than the lag phase of amyloidogenic Ab self-assembly and lowers free monomer concentration, a limiting factor for nuclei formation.
Although insulin is non-amyloidogenic under physiological conditions, the high affinity between IDE and insulin, the assisted partial unfolding of insulin inside the IDE chamber and the lack of proteolysis by IDEQ, raised the question whether IDEQ could induce stable insulin conformational changes and/or self-aggregation. Notably, despite a long co-incubation, IDEQ had no effect upon insulin conformation as assessed by CD and receptor activation on cultured cells.
Taken together, these results are consistent with IDEQ having a chaperone-like activity upon Ab aggregation with a holdase or mixed (''foldase-holdase'') type of mechanism. Proteases and chaperones share the capability of binding partially folded or unfolded polypeptides. Both activities have been extensively studied in proteins of the C1pP, FtsH and Lon families; highly conserved ATP-dependent proteases with chaperone activity [58,59]. Proteases in which the catalytic site is required for a chaperone-like activity include the aspartic proteases pepsin and HIV protease, as shown by following the aggregation of denatured citrate synthase or rhodanese [60]. In the same work, thermolysina typical M13 Zn 2+ metallopeptidase-did not show chaperone activity but this remains to be tested with relatively short peptides such as Ab. In gram-negative bacteria, the IDE ortholog, pitrilysin (PTR) is targeted to the periplasmic space in which it is believed to degrade small peptides derived from leader sequences [61]. In eukaryotes, IDE has been found in cytosol, plasma membrane, endocytic pathway, mitochondria, exosomes and extracellular space [42]. Several cellular factors may impact upon IDE proteolytic activity. Acidic pH, oxygen reactive species, oxidized glutathione and nitric oxide can reduce the catalytic function of IDE [62,63]. Moreover, it has been shown that naturally occurring IDE missense mutations in a rat model of DM2 reduce IDE proteolytic activity upon insulin and Ab [64]. Therefore, under physiological (low pH in endocytic and exocytic vesicles) or pathological conditions (oxidative stress, inflammation, mutations), a latent chaperone activity of IDE may become relevant to prevent Ab toxic aggregation.
The M16 family members presequence protease (PreP) in chloroplasts and human metalloprotease 1 (MP1) in mitochondria are capable of degrading Ab [65,66,67]. PreP crystal structure and its proposed mechanism of recognition and catalysis are similar to those of PTR and IDE, including a ''peptidasome'' chamber that encloses peptide substrates [68]. A chaperone-like activity of IDE and related M16 proteases such as PTR or PreP and MP1 under suboptimal conditions for hydrolysis may enhance their efficiency in preventing the self-assembly of small peptides into toxic oligomers. By driving the formation of non-toxic and more soluble aggregates, such activity may facilitate their cellular disposal by general degradation systems.

Site-directed mutagenesis
Rat IDE cDNA 42-1059 was kindly provided by Dr. Richard Roth, Stanford University). The Glu by Gln substitution at position 111 was introduced by PCR using the forward primer: 59CCTGGCTTAAGCCATTTTTGTCAGCATATGCTGTT-TTTGGGAACA39 and the complementary reverse primer. After cloning in plasmid pET 30 a (+) (Novagen), whole plasmid was amplified and the PCR product treated with Dnp I. Digestion of the methylated or hemi-methylated DNA template allowed the selection of the mutation-containing DNA. To confirm that the desired mutation was correctly introduced, the IDE insert was sequenced using a primer for the T7 promoter.

Expression and purification of IDEwt and IDEQ
IDEwt and IDEQ were expressed in E. coli BL21 transformed with pET 30 a (+) vector containing the corresponding cDNAs. Bacteria were cultured in TB medium containing 100 mg/mL kanamicin at 37uC and protein expression induced with 0.5 mM isopropyl-b-D-1-thiogalactopyranoside (IPTG). The first purification step was done with a His-Trap Ni 2+ affinity chromatography (GE Healthcare) in 50 mM Tris-HCl, pH 8.0, 300 mM NaCl followed by elution with a linear gradient from 20 to 250 mM imidazole in the same buffer. The second purification step was size-exclusion chromatography (SEC) with a Superdex 200 columm (Amersham) in 50 mM NaCl, 20 mM Tris-HCl, pH 7.4 (working buffer). After SEC, protein purity was determined by SDS-PAGE and quantified by measuring absorbance at 280 nm using the IDE extinction coefficient (115810 M 21 cm 21 ). Concentration was expressed as monomeric IDE.

Ab peptides preparation and IDE-AbSCx formation
Ab1-40 (Sigma) and Ab1-42 (American Peptide Co) were acquired as lyophilized powder. Ab1-28Q was kindly provided by Jorge Ghiso (New York University). Peptides were first dissolved in 1,1,1,3,3,3-hexa-fluoro-2-propanol (HFIP) at 1 mg/ml, incubated at room temperature for 8 h and stored at 280uC until use. To disaggregate possible Ab nuclei, prior to each experiment an aliquot was taken from the stock solution and sonicated for 10 min on ice. After removing HFIP with gentle streaming of nitrogen, peptides were dissolved in working buffer and sonicated for 10 min on ice. Finally, peptides were centrifuged for 30 min at 10,0006 g to pellet insoluble material and the supernatants filtered through 0.22 mm spin filters (Sigma). To obtain IDEwt or IDEQ AbSCx, purified enzymes and peptide were incubated at 1 mM and 35 mM, respectively, in working buffer at 25uC for the indicated times. At the end of incubation, samples were analyzed by Western blots as described below. Data were fitted to a single exponential equation to calculate half-lives as described [44].

IDE activity assays
IDEwt and IDEQ enzymatic activities were assessed by four different methods. 1) Hydrolysis of the fluorogenic peptide 7methoxycoumarin-RPPGFSAFK-2, 4-dinitrophenyl (''substrate V'', R&D Systems) containing the bradykinin sequence. The enzymatic activity of IDEwt (10 nM) was determined using increasing concentrations of substrate V in working buffer, pH 7, for 10 min at 37uC. The sigmoidal plot of V 0 as a function of substrate concentration allowed us to calculate a K 0.5 ,12.5 mM. To compare the activities of IDEwt and IDEQ, proteins were incubated at 10 nM with substrate V at 10 mM in working buffer at 37uC for up to 10 min. The increase in fluorescence was monitored with an Aminco-Bowman spectroflu-orometer at 320 nm and 405 nm as excitation and emission wavelengths, respectively. 2) TCA precipitation of porcine insulin degradation. Briefly, 20,000 cpm of 125 I-insulin (specific activity 300 mCi/mg, kindly provided by Edgardo Poskus, University of Buenos Aires) were incubated for 30 min at 37uC in working buffer containing 0.25% BSA and 10 ng of IDEwt or IDEQ. ATP was added at 0.5 mM to IDEQ, as indicated. The reaction was stopped by addition of 0.1 ml of 15% TCA. Samples were centrifuged at 12,0006 g for 30 min at 4uC and c.p.m. determined in pellets and supernatants using an automatic gamma-counter. 3) Western blots after incubation of Ab1-42 with or without IDEQ were performed as described below. 4) ELISA quantification was performed on samples with Ab1-42 (15 mM) alone or in the presence of IDEwt, IDEwt containing 5 mM EDTA or IDEQ at 1:10 molar ratios (enzymes: peptides). At ''dead-time'' or after 5 days of incubation at room temperature, aliquots of 2 ml were mixed with 2 ml of distilled water and 1 ml of 5M guanidine-HCl in water. Ab42 levels were quantified by using a commercial ELISA kit following the manufacturer's instructions (Invitrogen).
IDEwt activity assays in pH range 5.5 to 7.5 The effect of pH on IDEwt activity was determined in a pH range from 5.5 to 7.5 using the fluorogenic peptide substrate V as described above. One hundred mM citric acid or 100 mM PO 4 NaH 2 / PO 4 Na 2 H were used to reach the specified pH in each case.

Circular dichroism and thermal denaturation of IDEwt and IDEQ
Far UV-CD spectra measurements were carried out in a Jasco J-815 spectropolarimeter equipped with a Peltier temperaturecontrolled sample holder device. Spectra of IDEwt or IDEQ at 0.5 mM in working buffer at 25uC were acquired before and after thermal denaturation at 1uC/min using a 0.2 cm path length quartz cuvette. Final spectra were obtained after the subtraction of background readings of working buffer. Thermal denaturation was followed by plotting the signal at 222 nm.

Static light scattering assays (SLS)
The average molecular masses of IDEwt or IDE-Q were determined with a Precision Detector PD2010 light scattering instrument. Briefly, 200 mg of a purified batch of IDEwt or IDEQ were injected into a Superdex G-200 column (Amersham) connected to an HPLC system Bromma LKB 2248 and eluted in working buffer at 0.5 ml/min. To estimate the average molecular mass of proteins, the 90u static light scattering signal and UV absorbance or refractive index of the eluting material were recorded and analyzed with the Discovery 32 software (Precision Detectors). The 90u light scattering detector was calibrated using bovine serum albumin (66.5 kDa) as a standard. Average values obtained from four independent measurements of the apparent molecular masses of IDE dimers and tetramers were ,250 and ,500 kDa, respectively.

Dynamic light scattering (DLS)
Ab monomeric species, obtained as described above, were incubated at 50 mM in working buffer at 25uC for the indicated times with or without IDEQ at a molar ratio 1:10 (enzyme:peptide). The same molar ratio was used for BSA as a control for specificity. Particle size measurement was assessed using the 633 nm excitation laser and detector at 173u with a Zetasizer Nano-S (Malvern Instruments). Results were expressed as the intensity of scattered light as a function of D H and time of incubation.

Transmission electron microscopy (TEM)
Aliquots were taken from the same samples used for DLS measurements or from samples at steady state of turbidity experiments (see below). Ten ml were adsorbed for 1 min onto formvar-coated nickel grids (Electron Microscopy Science) and stained with 1% uranyl acetate aqueous solution for 1 min. Grids were visualized on a Zeiss 10C Transmission Electron Microscope.

Turbidity assessment of Ab aggregation kinetics and seeding
One hundred and fifty ml of Ab1-42 at 10 mM were loaded per well in a 96-multiwell polystyrene plate with flat bottoms (Nunc). Co-incubation with IDEQ was done at the indicated molar ratios. Plates were placed in a TECAN Genios multi-well reader and peptide aggregation followed by measuring absorbance at 340 nm every 30 min with a 10-s mixing shake before each measure. Preformed seeds were obtained from samples at steady state by sonication for 5 min on ice. The seeding assay was carried out by adding 1% v/v of seeds to a fresh preparation of monomeric Ab. All experiments were performed at constant 25uC.

AFM characterization of non-fibrillar Ab1-42 species
Aliquots of Ab1-42 or Ab1-42 with IDEQ were collected after 4 and 6 days of incubation, deposited onto freshly cleaved mica and left for 1 min at room temperature. The specimens were then rinsed with deionized water and dried with a nitrogen stream (1 bar). Images were collected after dynamic scanning force microscopy in air by using a Veeco Explorer AFM fitted with a tube-type 2 micrometer scanner and antimony doped silicon cantilever tips (spring constant 1-5 N/m; Veeco). Tapping/Noncontact mode using a FESP-MT tip and a gain of 2.0 V were used. Three sample regions covering 4 mm 2 were randomly documented and 100 to 1500 elements from each region were analyzed. Particles' heights, widths and lengths were measured by using the Nanotec WSxM 4.0 Image Browser software.

CD assessment of Ab and insulin-IDEQ co-incubation
Ab1-42 at 15 mM in working buffer alone or in the presence of IDEQ at 1:300 molar ratio (enzyme:peptide) were incubated for 6 days at 25uC. Spectra were acquired at dead-time and at the end of incubation over a wavelength range of 260-202 nm using a Jasco J-815 spectropolarimeter in 10-mm path cuvettes. Data interval was set at 0.2 nm and the average of five scans was obtained for each experimental conditions. Insulin secondary structure was determined in the presence or absence of IDEQ at a 1:100 molar ratio (enzyme:peptide) incubated for 24 h at 25uC. In both experiments, UV-CD spectra were smoothed with an algorithm provided by JASCO after the subtraction of working buffer background signal.

SDS-PAGE and Western Blots
Proteins were subjected to electrophoresis on 7.5 or 10% polyacrylamide SDS-Tris-tricine gels, stained with Coomassie blue or transferred to PVDF membranes (GE Healthcare). For Western blots, membranes were blocked with 5% low fat milk in PBS containing 0.1% Tween for 2 h at room temperature and incubated with the indicated primary antibodies overnight at 4uC. Anti-Ab monoclonal 6E10 and 4G8 were obtained from Signet. Anti-rabbit or anti-mouse horseradish peroxidase-labeled antibodies were used for detection using ECL Plus reagent (Amersham Biosciences). Enhanced chemiluminescence signals were scanned with a STORM 860 fluorometer and analyzed with ImageQuant 5.1 software (Molecular Dynamics). To quantify IDE-AbSCx formation, the amount of total IDE in each lane was estimated by Coomassie blue staining of the same PVDF membranes used for fluorescence detection.

Primary neuronal cultures
Neuronal cultures were obtained from rat embryos as described [69]. Briefly, hippocampi were dissected free of meninges, suspended in 0.5 ml of Ca 2+ and Mg 2+ -free Hank's balanced solution and treated with 0.25% trypsin (Invitrogen) for 25 min at 37uC. Immediately after trypsinization, tissue was washed with Neurobasal medium (Gibco) containing 10% fetal bovine serum supplemented with B-27 and N-2, 50 U/ml penicillin and 50 mg/ ml streptomycin (NBSM). A cellular suspension was obtained by using fire-polished Pasteur pipettes. For immunofluorescence, isolated cells were plated at 60/mm 2 on 12-mm coverslips. For viability assessment, cells were plated at 3610 4 on P-96 wells. After 3 h of incubation under 5% CO 2 at 37uC to permit cell attachment, medium was replaced with fresh medium without serum. Neuronal cultures with a purity of 90-95% were obtained. Cultures were maintained in NBSM without serum for 14 days to allow neuronal differentiation.

Inmunofluorescence and neurotoxicity assay
To assess neuronal damage induced by Ab1-42, cells were examined by immunofluorescence (IF). The remaining viability was determined using an (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. Neurons were exposed for 20 h to 10 mM Ab1-42 in NBS, washed with PBS and fixed with 4% paraformaldehyde. For immunofluorescence, cells were permeabilized with 0.1% Triton X-100 in PBS and blocked with PBS containing 0.1% BSA. To recognize specifically neuronal morphology, anti-b III tubulin monoclonal IgG (Sigma) was incubated for 1 h, followed by Cy3-labeled goat anti-mouse IgG. Nuclei were labelled with the blue-fluorescent Hoechst acid stain. Images were captured using an Olympus BX60 microscope with a DP Controler 3.1.1.267 PC program. The average length of neuronal processes was determined using Image J software. Quantitative estimation of remaining cellular viability was assessed by the reduction of MTT by mitochondria of living cells. MTT was incubated at 0.5 mg/ml for 2 h at 37uC and products of reaction dissolved in 0.1N HCl and 2% SDS in water. Absorbance was measured at 570 nm and viability expressed as the O.D. relative to neurons treated with vehicle alone. As a positive control for neuronal damage, cells were exposed to 0.05% H 2 O 2 (not shown). Quantitative data of three independent MTT assays were analyzed by one-way ANOVA and a post-hoc Tukey test with Graph Pad Prism v. 4.0. All results represent the mean 6 SEM and p,0.05 was considered statistically significant.

Insulin receptor activation assay
Human glioma U-87 cells (ATCC HBT-14) were cultured at 37uC with 5% CO 2 in DMEM high glucose media supplemented with non-essential amino acids (PPA) and 2 mM piruvic acid. Cells at 80% confluence were deprived of serum for 24 h and exposed to 100 nM insulin for 30 min. Insulin was co-incubated with IDEQ or IDEwt for 24 h at 25uC before cellular exposure. After treatments, cells were washed with cold PBS, harvested and lysed by passage through a 25-gauge needle 20 times in 150 mM PO 4 H 2 Na, pH 7.5 containing 150 mM NaCl, 0.5% deoxycholic acid, 1% Triton X-100 and a cocktail of protease inhibitors (Sigma). Homogenates were centrifuged for 10 min at 12,0006 g and protein concentration in the supernatant determined with a Pierce BCA kit. To detect Akt and phosphorylated Akt at Ser 473 under different conditions, 65 mg of total protein were separated on 10% SDS-PAGE and transferred to PDVF membranes. Polyclonal rabbit anti-Akt and monoclonal mouse anti-p-Akt antibodies from Cell Signaling were used for Western blots. To assess the specificity of signaling, phophatidyl-inositol-3 kinase (PI3K) was inhibited by incubating cells with 100 nM wortmannin (Millipore) for 30 min prior to insulin exposure.

Ethics Statement
The protocol for the care, handling and use of animals followed the ARRIVE guidelines [70] and was approved by the Fundación Instituto Leloir Institutional Animal Care and Use Committee.