The Aβ(1–38) peptide is a negative regulator of the Aβ(1–42) peptide implicated in Alzheimer disease progression

The pool of β-Amyloid (Aβ) length variants detected in preclinical and clinical Alzheimer disease (AD) samples suggests a diversity of roles for Aβ peptides. We examined how a naturally occurring variant, e.g. Aβ(1–38), interacts with the AD-related variant, Aβ(1–42), and the predominant physiological variant, Aβ(1–40). Atomic force microscopy, Thioflavin T fluorescence, circular dichroism, dynamic light scattering, and surface plasmon resonance reveal that Aβ(1–38) interacts differently with Aβ(1–40) and Aβ(1–42) and, in general, Aβ(1–38) interferes with the conversion of Aβ(1–42) to a β-sheet-rich aggregate. Functionally, Aβ(1–38) reverses the negative impact of Aβ(1–42) on long-term potentiation in acute hippocampal slices and on membrane conductance in primary neurons, and mitigates an Aβ(1–42) phenotype in Caenorhabditis elegans. Aβ(1–38) also reverses any loss of MTT conversion induced by Aβ(1–40) and Aβ(1–42) in HT-22 hippocampal neurons and APOE ε4-positive human fibroblasts, although the combination of Aβ(1–38) and Aβ(1–42) inhibits MTT conversion in APOE ε4-negative fibroblasts. A greater ratio of soluble Aβ(1–42)/Aβ(1–38) [and Aβ(1–42)/Aβ(1–40)] in autopsied brain extracts correlates with an earlier age-at-death in males (but not females) with a diagnosis of AD. These results suggest that Aβ(1–38) is capable of physically counteracting, potentially in a sex-dependent manner, the neuropathological effects of the AD-relevant Aβ(1–42).


Results
Biophysical assays. We monitored secondary structures or aggregation potential of our Aβ peptide mixtures using established biophysical techniques, including AFM, ThT fluorescence, CD, DLS, and SPR.
AFM measurements (Fig. 1A-G) of freshly prepared Aβ  and Aβ  reveal particles primarily with < 2 nm heights (Fig. 1B), which are generally associated with monomeric species 23 . Though the average volume values for the Aβ particles will be larger due to the AFM tip convolution effects, the general comparison between Aβ(1-38) and Aβ  shows that Aβ(1-38) has a larger particle volume (Fig. 1E), suggesting a different morphology or modest initial aggregation when compared to Aβ(1-42) particle morphology. With time, Aβ  progresses through to amyloid fibrils. while Aβ  co-incubated with Aβ(1-38) shows a dynamic profile that also begins relatively homogenously small and progresses (at 24 h) through a heterogeneous mix of heights and volumes (e.g. with two peaks at 2 and 6 nm in height and a shallow, but broad, volume peak at 50,000 nm 3 ) (Fig. 1C,F), to return to a more homogenous mixture of particles that remain smaller and shorter after 48 h incubation (Fig. 1D,G).
For our Western blotting experiments, we used increasing concentrations (2,10, and 20 µM) to monitor potential aggregation of the various peptides. First, our blots confirm that HFIP-treated Aβ , Aβ , and Aβ  preparations contain significant amounts of monomeric species (bottom panels, Fig. 2), as expected 23,24 . In isolation, there are no detectable aggregates in the Aβ  or Aβ  solutions, even at concentrations of 20 µM (top left panel, Fig. 2). In contrast, the 10 and 20 µM solutions of Aβ  provide evidence of significant high molecular weight (HMW) aggregates. Co-incubation with increasing concentrations of Aβ  progressively lessens the amount of HMW aggregate associated with 20 µM Aβ(1-42) (top middle and right panels, Fig. 2). Mixtures of Aβ(1-38) and Aβ(1-40) do not present any evidence of HMW complexes. These observations confirm that Aβ  can interfere with Aβ  aggregation and, as importantly, suggest that although Aβ  and Aβ  aggregates are detected by AFM (see Fig. 1), these aggregates have profoundly different physicochemical properties, with, for example, Aβ  aggregates being stable on SDS-PAGE, whereas Aβ(1-38) aggregates are not.
There is a time-dependent increase in ThT fluorescence for Aβ  and Aβ(1-42) (Fig. 3A). While Aβ  in isolation causes an anomalous initial decrease in ThT fluorescence, this stabilizes over time. In a mixture, Aβ  exerts distinct effects on the other peptides, e.g. increasing ThT fluorescence with Aβ(1-42) (Fig. 3B), but decreasing ThT fluorescence with Aβ(1-40) (Fig. 3C). These data suggest differences in β-sheet content depending on whether the peptides are incubated in isolation versus in a mixture.
In isolation, Aβ  and Aβ  show a strong negative peak around 200 nm in the far-CD spectrum indicating an initial disordered state 25,26 , but with time the canonical anti-parallel β-sheet content emerges ( Fig. 3D-G). A 20:1 Aβ(1-42):Aβ  solution has much less β-sheet growth, whereas a 20:1 Aβ(1-40):Aβ  solution appears to show evidence of some β-sheet growth that was not observed when Aβ  was incubated in isolation. These observations appear to contrast with our ThT binding data (above), but appear to align with our DLS measurements (Fig. 3H). Indeed, the scattering intensity, e.g. particle size, is lower over time in the Aβ :Aβ  mixture when compared to either peptide alone, whereas Aβ  particles in isolation are far smaller and tend to be modestly larger when co-incubated with Aβ(1-38) (Fig. 3H).
These cumulative biophysical data confirm that the individual peptides do demonstrate varying degrees of aggregation potential, but that co-incubation with Aβ(1-38) can trigger dramatically different aggregation behavior in Aβ  and Aβ . This further suggests that conclusions drawn from studies of Aβ peptides in isolation, while meaningful to understanding the behaviour of that particular peptide, likely cannot be generalized to the peptide's behaviour in more complex biological mixtures. We chose to examine whether these biophysical trends extended to functional paradigms. Functional assays. Freshly prepared Aβ(1-42) decreases mitochondrial respiration (e.g. MTT conversion) in a concentration-dependent manner [P < 0.0001], whereas Aβ(1-38) on its own has no effect (Fig. 4A). Cotreatment with Aβ(1-42) and differing ratios of Aβ(1-38) does not produce any effect that cannot be simply attributed to a titration of the Aβ(1-42) effect. Interestingly, a combination of Aβ(1-38) and subequimolar concentration of Aβ(1-42) appears to reduce mitochondrial respiration compared to Aβ(1-38) alone, although the effect does not reach statistical significance (Fig. 4A). In contrast, the effect of Aβ  [P < 0.0001) is completely inhibited by co-treatment with Aβ(1-38) in HT-22 cells (Fig. 4B).
To determine whether APOE ε4, a risk allele for late-onset AD in women (discussed in 27 ), might influence outcomes, we tested the peptides in two human fibroblast cell lines from female donors that differed in their APOE ε4 status. MTT conversion tends to be marginally affected in the APOE ε4/ε4 fibroblast cell line by Aβ(1-40) (P = 0.0629) and Aβ(1-42) (P = 0.0814) (Fig. 4C). Post-hoc analysis shows that Aβ(1-38) exerts no effect on its own in this cell line, but reverses the modest effects exerted by both Aβ  and Aβ . In contrast, the peptides do not exert any effect in the APOE ε2/ε3 fibroblast cell line (Fig. 4D), although significance (P < 0.01) across the treatment groups is detected; post-hoc analysis reveals that a subequimolar concentration of Aβ  exacerbates the effect of Aβ , while a similar subequimolar concentration of Aβ(1-40) does not (Fig. 4D). This intriguing observation warrants further investigation.
Patch-clamping reveals that hippocampal neuron current densities, e.g. currents normalized to the cell capacitance and an index of membrane permeability, are highest (18.97 pA/pF ± 2.15) in non-Aβ exposed neurons (CTL) (Fig. 5C). Exposure to Aβ(1-38) causes a significant decrease in current density (9.21 pA/pF ± 3.85, P = 0.007) and exposure to Aβ  leads to an even lower current density (3.10 pA/pF ± 1.91, P = 0.01), and yet, as with the LTP paradigm, co-treatment with Aβ(1-38) rescues the effect of Aβ . Interestingly, the decrease in current density observed for Aβ  and a subequimolar concentration of Aβ(1-42) is more in the range of current density measured when neurons are exposed to Aβ(1-42) alone. As Aβ peptides decrease the current, the membrane resistance is increased and the membrane potential that is measured is -80 mV (reflecting the closure of K + channels). www.nature.com/scientificreports/ These two electrophysiological paradigms confirm that Aβ  has the capacity to negatively regulate Aβ(1-42)-mediated deficits at the synapse and the cell membrane. We used the C. elegans worm to test whether any 'protection' afforded by Aβ(1-38) might extend to an in vivo context.
The C. elegans GMC1010 strain expresses full length Aβ  in body wall muscle cells and exhibits a paralysis phenotype 28 . We used 'thrashing rate' as a proxy for compromised muscle function due to Aβ accumulation. The CL2122 strain, which expresses GFP in the intestine, but produces no Aβ peptide, was used as a control. The presence of muscle-specific Aβ(1-42) and/or Aβ  in these worms is inferred with detection of gut GFP and neuronal DsRed, respectively (Fig. 6A). Western blotting confirms the expression of Aβ  in the GCM101 strain and Aβ  in the CEC220 strain (Fig. 6B). However, both peptides show a similar mobility on Urea/PAGE, which suggests an increase in hydrophobicity of the Aβ(1-38) species 29 in this worm, presumably through some post-translational modification. Samples resolved on standard 15% SDS-PAGE ( Fig. 6B) reveal that a putative Aβ(1-38) dimer migrates higher than an Aβ(1-42) dimer, confirming a modified, e.g. heavier, Aβ(1-38) species.
At 12-h post-L4 larval stage, the thrashing rate in synchronized populations of Aβ(1-42)-expressing worms (GMC101) was reduced (P < 0.0001) as compared to the control CL2122 strain, but not compared to Aβ(1-38)expressing worms (CEC220) (Fig. 6C). The motor defect observed in GMC101 worms is consistent with a previous report 28 . At 18 h, the thrashing rates is significantly higher in CEC222 worms, which express both Aβ  and Aβ  peptides, when compared to worms that express Aβ(1-42) alone [two-way ANOVA, Interaction P = 0.0004]. At 24 h, this difference is less pronounced, but remains significant, and by 36 h any difference is lost (Fig. 6C).
The C. elegans data suggest that any potential benefit attributable to Aβ(1-38) might be masked if Aβ(1-42) accumulates within the same tissue. The AD brain tends to accumulate numerous Aβ species and we recently demonstrated that levels of insoluble (guanidine-extractable) Aβ peptides differed in a region-and sex-dependent manner in brain samples from autopsy-confirmed cases of AD 27 . We chose to examine how levels of soluble Aβ peptides relate to each other in these same samples.
Clinical autopsy samples. Early-onset (EO) and late-onset (LO) AD donor statistics are presented in Table 1. Western blotting of RIPA-extracts of cortical samples clearly reveals bands corresponding to Aβ  For ease of interpretation, the relative proportions of Aβ(1-42) to either Aβ  or Aβ(1-40) are presented as gnu plots (Fig. 9). In both regions, the relative abundance of Aβ(1-38), Aβ(1-40), and Aβ(1-42) is greater in EOAD samples. There is generally more of the three peptides in males with LOAD in both regions. In contrast, while all three peptides are detectably higher in female cortical EOAD (vs. control) samples, levels in the corresponding hippocampal LOAD samples are unchanged from those in female controls (Fig. 9).
Parenthetically, it is known that protein accounts for 10% of brain wet weight 30 and that brain density (e.g. g per cm 3 ) is approximately '1' 31 . Using these factors, we were able to convert our data, expressed in 'ng per mg protein' , to 'ng per gram wet weight tissue' and our rough estimates for the cortical samples are: 0.  (C) Average current densities (± sem) from whole cell voltage-clamp recordings in primary hippocampal neurons exposed to Aβ(1-38) and Aβ(1-42) (in µM, 24 h). Cells were held at -60 mV and currents were elicited by voltage ramps, e.g., step depolarized between -80 and + 100 mV (in 20 mV increments). Current density was measured at 0 and 20 mV, 100 ms after the voltage step. Currents were divided by cell capacitance and reported as current densities (pA/pF). **: P < 0.01; ***: P < 0.001; ****: P < 0.0001 between indicated groups. www.nature.com/scientificreports/ Finally, so as to determine whether APOE ε4 status might be influencing any of these autopsy-derived data, we re-analyzed the data by stratifying for APOE ε4 status and sex (independent of diagnosis as samples sizes were too small). In the cortical samples, any significant effects (data not shown) were limited to: an increase in Aβ There were no effects of APOE ε4 status on soluble peptide levels in the hippocampal samples.
Our observations indicate region-and sex-dependent proportions of these Aβ peptides in the AD brain. Up to this point, we have examined the effect of peptides one-on-one. We chose to determine what might transpire should different proportions of all three of the Aβ peptides be allowed to interact in a mixture. To do so, while still keeping the experimental design manageable, we monitored peptide-peptide interactions in real-time using surface plasmon resonance.
Surface plasmon resonance (SPR). Approximately 1600 relative units (RUs) of either Aβ(1-40) or Aβ(1-42) were immobilized on a biosensor chip. First, we demonstrate that Aβ(1-38) binds more with Aβ(1-   in the GMC101 strain and Aβ  in the CEC220 strain. The protein ladder (in kDa) is indicated on the left. Extracts were also resolved by standard 15% SDS-PAGE and reveal putative monomers (M) and dimers (D). A band in the CEC220 extract (identified with white asterisk '*') likely represents a modified Aβ(1-38) dimer, which is seen more clearly in a longer exposure (lower panel). (C) Time-dependent changes in the thrashing rates in CL2122 (control) worms (filled circle) and worms expressing Aβ(1-38) (filled square) (relative to the '30-50' segment on the right Y-axis) as well as in the Aβ(1-42)-expressing GMC101 strain (○), and worms co-expressing Aβ  and Aβ(1-42) phenotype (□). Two-way ANOVA shows that all groups were significantly different from their respective control groups (n = 60-90; mean ± sem). **: P = 0.01; ****: P = 0.0001 between indicated groups. These observations confirm that the three peptides can interact in a complex mixture and that Aβ(1-38), under these circumstances, might interfere with Aβ(1-40) self-aggregation, but could promote the interaction between Aβ(1-40) and Aβ . This could be fundamental for our understanding of amyloid plaque formation in the AD brain.

Discussion
Although there are countless reports regarding the behaviour of Aβ peptides in isolation, very little is known of their properties when incubated as complex mixtures or when studied immediately upon reconstitution of HFIP-treated stocks, when the peptides would have the lowest percentage possible of β-sheet structure 24 , which would more closely reflect the state immediately upon synthesis in vivo.
Our biophysical analyses confirm that these peptides exhibit significant differences in size distribution and time-dependent genesis of secondary structural elements 23,33 , with the properties of Aβ  lying between those of Aβ(1-40) and Aβ   22 . We also demonstrate that co-incubation of Aβ(1-38) with Aβ(1-42) mitigates fibril length and aggregate size as well as overall β-sheet content of any interacting complex, and these differences in biophysical profiles generally align with differences in functional profiles.
Aβ  in the form of higher molecular weight aggregates 18,34 as well as low molecular weight and toxic oligomers composed of dimers, trimers, and tetramers 33 has been shown to inhibit LTP, by way of interactions with the phospholipids of the plasma membrane 35 or with specific receptors, e.g. the insulin receptor 36 or cholinergic receptors 37 . The Aβ peptides also directly alter conductance centered on calcium homeostasis, with the latter implicating specific calcium channels 38 , NMDA 39 or AMPA 40 receptors, or a role for channel formation by Aβ(1-42) itself 41,42 . Longer, soluble Aβ peptides, including Aβ(1-42), alter synaptic plasticity and impair hippocampal LTP 43,44 , while the shorter peptides, including Aβ(1-37/38/39/40), are less likely to elicit any overt effect on synaptic function 21 . Our current studies confirm a significant inhibition of LTP by soluble Aβ(1-42) and a modest (~ 20%), albeit not statistically significant, inhibition of LTP by Aβ(1-38); however co-treatment with the two peptides completely rescues the impaired LTP phenotype observed with Aβ(1-42) alone. Similarly, patch-clamping shows an Aβ(1-42)-dependent loss of current density in primary hippocampal neurons, which is reversed by an equimolar concentration of Aβ . Although the actual mechanism needs to be defined, our observed changes in current density suggest that Aβ(1-38) might be mitigating a disruption of membrane permeability that has been demonstrated elsewhere by changes in intracellular Ca 2+ (also discussed above) or increased influx of dye (e.g. ethidium bromide) following treatment with Aβ(1-42) 45 . Another possibility might be that Aβ(1-38) is disrupting self-association of Aβ(1-42) into an ion-conducting channel or pore structure that is not evident with shorter peptides, e.g. Aβ   42 , in such acute treatment paradigms. Our observations have significant implications for the influence of Aβ length variants on neuronal membrane integrity, synaptic plasticity, and memory formation.

Control
Early-Onset AD Late-Onset AD www.nature.com/scientificreports/ Our two human fibroblast cell lines, both from female donors, but differing in their APOE ε4 status (a risk for AD in women), also yielded intriguing results. This APOE ε4/ε4 fibroblast line is modestly sensitive to both Aβ(1-40) and Aβ , and this is reversed by co-treatment with Aβ(1-38). In contrast and somewhat counterintuitively, the APOE ε2/ε3 cell line does not respond to either Aβ    The relation between these ratios and age of the donor at autopsy were examined by regression (Pearson's) analysis. *: P < 0.05; **: P < 0.01; ***: P < 0.001 vs. corresponding controls (CTL). EO: Early-Onset AD; LO: Late-Onset AD. An example of a full-length blot is presented in Supplementary Fig. 1 47 , potentially through a direct interaction with phospholipids 35 , while larger aggregates are thought to trigger the pro-inflammatory reactions often associated with the AD brain 47 . Furthermore, soluble, primarily N-terminally truncated Aβ extracts from the AD brain can induce amyloidosis when injected intracerebroventricularly in mice, whereas soluble Aβ extracts from CSF (containing both C-terminally and N-terminally truncated species) do not 48 . Our Western blotting observations suggest that unlike the stable Aβ(1-42) aggregate we observe, any potential Aβ(1-38) aggregate is likely not stable under SDS-denaturing conditions. This inherent difference between aggregates of Aβ(1-38) and Aβ(1-42) is supported by the anomalous loss of ThT fluorescence we observed with Aβ(1-38) alone. Part of this could be explained by the fact that ThT, while a valid probe for monitoring protein folding and amyloid fibril behaviour, actually binds non-covalently (reversibly) to cross-β-strand structures, rather than to the β-sheet region of amyloid structure 49 . Thus, the ThT fluorescence associated with Aβ(1-38) in isolation may be depicting a dynamic flux in conformation of this peptide over time, whereas those traces associated with complex mixtures containing Aβ(1-38) may be depicting more stable conformations owing to Aβ variant interactions. Whatever the mechanism, it is clear that studying these peptides in isolation is biasing our understanding of their influence in pathophysiological mixtures.
The different stages of AD progression have been associated with different proportions of Aβ peptide in the RIPA/SDS-soluble versus the insoluble (plaque-associated) fractions 50 . We recently reported significant Aβ  and Aβ(1-42) levels in the guanidine-extractable, plaque-associated fraction in autopsy AD brain samples 27 and we now reveal sex-dependent differences (with some influence of APOE ε4 status) in levels of soluble Aβ peptides in these same samples. The levels of soluble Aβ(1-38), Aβ , and Aβ  were all increased in samples from donors with a diagnosis of EOAD, confirming a previous study based on aggressive, genetic forms of AD 12 and cell-based studies of familial mutations in the gene encoding presenilin-1, the catalytic core of the γ-secretase complex 51 . This suggests an indiscriminate processing of the APP precursor through to a heterogeneous pool of Aβ peptides in these more aggressive cases of the disease. In contrast, there were increases in the levels of all three peptides in the LOAD cortical samples, but only Aβ(1-42) was significantly elevated over control levels. www.nature.com/scientificreports/ Although similar patterns emerged in the hippocampal samples, they were not statistically significant and, in fact, levels in female LOAD samples remained remarkably unchanged from those in control samples. These observations continue to support differences in the male and female LOAD brain, and indirectly support the temporal pattern of amyloid burden that has been associated with AD progression, i.e., increases in amyloid in the cortex precede those in the hippocampus 52 . A temporal relevance to the interaction between Aβ(1-38) and Aβ  was confirmed by our C. elegans experiments, in which Aβ(1-38) was able to mitigate an Aβ(1-42)-mediated phenotype at earlier time-points, but any 'protection' was gradually lost as both peptides continued to accumulate.   www.nature.com/scientificreports/ These data remain cross-sectional and while they cannot inform on whether any observed changes were adaptive or causative, they certainly do support differences in the male and female AD brain, potentially suggesting differences in Aβ clearance mechanisms between the sexes, and supporting the consideration for different therapeutic strategies based on sex. We previously observed higher levels of Aβ  in older male (vs. female) J20 (APP Swe/Ind ) mouse brains 13 and potential sex-dependent differences in clearance (female > male) from the brain have been shown in the APP Swe /PS1ΔEx9 mouse 55 . Sex-dependent differences in CSF levels of Aβ  have been shown to correlate with differences in cognitive function, for example, based on the Mini-Mental State Examination (MMSE) 56 or the Word List Delayed Recall 57 , while higher levels of plasma Aβ(1-42) have been detected in women with preclinical sporadic AD 58 . γ-Secretase inhibitors capable of shifting the cleavage of APP to yield Aβ  at the expense of Aβ(1-42) 14 could have translational relevance, yet it is important to note that shorter Aβ length variants are not necessarily all beneficial; for example, an increase in CSF levels of Aβ(1-34), a BACE1-mediated fragment 59 found in AD CSF and brain extracts 9,48 , is a putative marker for conversion from mild cognitive impairment to AD 59 , while Aβ(1-24), the APP fragment ostensibly tied to MPP9 cleavage, can act as a seed for Aβ(1-42) fibrillogenesis and trigger behavioral and cognitive phenotypes in the wildtype mouse similar to those observed in an age-matched APP/PS1 mouse 60 .
In reality, the interaction of Aβ peptides is likely far more complicated than suggested herein. Indeed, our SPR results reveal that a subequimolar concentration of Aβ  interferes far more with the ability of Aβ  [vs Aβ ] to recognize Aβ  or Aβ , but that the same subequimolar concentration of Aβ  promotes the interaction between Aβ(1-40) and Aβ . These SPR results (which are reminiscent of our observation that Aβ(1-38) exerts opposite effects on Aβ(1-42)-and Aβ(1-40)-mediated ThT fluorescence) clearly expose complex interactions that are relevant to the emerging interest in understanding the clinical impact of a heterogeneous pool of Aβ variants.
It is often simplistically presumed that any Aβ length variant, or its accumulation, exacerbates the pathological progression associated with AD. Unfortunately, this misconception has underscored AD research for so long that evidence to the contrary, e.g. Aβ burden in cognitively intact elderly individuals or any beneficial roles reported for Aβ peptides, is often viewed as an anomaly 1 . Yet, the possibility that soluble Aβ length variants could be exerting a multitude of roles, with some being neuroprotective rather than 'amyloidogenic' and neurotoxic, would support a neurobiological 'benefit' for heterogeneity within the pool of Aβ peptides and could help to explain why the indiscriminate targeting of Aβ peptide(s) in AD clinical trials has met with a succession of negative outcomes 1,61 . (H-1368) were obtained from Bachem Americas, Inc., and the amino acid composition was confirmed by mass spectrometry. All peptides were reconstituted in hexafluoroisopropanol (HFIP) so as to disrupt any preexisting β-sheet structures 24 and residual HFIP was evaporated prior to peptide use in any assay. Different commercial lots were used to avoid the possibility that our results were biased by a particular batch of synthetic peptide(s). The anti-β-amyloid antibody [clone 6E10: targets residues 1-16 of the Aβ peptide: cat# 803016] was obtained from BioLegend. The antibody raised against the C-terminal region of human APP695 [targets residues 676-695: cat# A8717] was obtained from Sigma-Aldrich Ltd. Protein-A/G sepharose was obtained from GE Healthcare Bio-Sciences Inc.
Biophysical experiments. Atomic Force Microscopy (AFM) measurements were used to monitor changes in fibril morphology in our peptide preparations and were carried out on a PicoSPM instrument (Molecular Imaging) operating in intermittent contact mode. A silicon cantilever (NSG_L, K-TEK Nanotechnology) with tip curvature of radius < 10 nm, a force constant of approximately 58 N/m, and a resonant frequency of approximately 190 kHz was used for each measurement. Experiments were conducted at a set-point ratio of approximately 0.8-0.85 from the free-amplitude of the cantilever and all measurements were obtained in a vibration isolation system. The scan rate was 0.5-1.0 Hz (512 pixels per line) for all images. Data were analyzed using SPIP V5.1.6 software (Image Metrology).
Mica surfaces were prepared by applying 25 µL of poly-L-lysine (0.01% 70-150 kDa) for 3 min. Surfaces were then rinsed three times with Millipore water and gently dried under nitrogen gas. Aβ peptide solutions (0.1 mg/ mL in PBS) were incubated for 0, 24, or 48 h and then deposited onto the freshly coated surfaces for 3-5 min, rinsed with water, air dried, and stored in a dust-free environment until imaged.
Gel electrophoresis (Western blotting) was used to visualize the aggregation potential of Aβ peptide mixtures. Peptide solutions were incubated at room temperature for 24 h. Aliquots were resolved using either standard 15% SDS-PAGE or a discontinuous 8 M urea/12% SDS-PAGE system as we have done for Aβ peptides isolated from the insoluble (guanidine-extractable) fraction of these same tissues 27 and then transferred to nitrocellulose membrane. We found that boiling the membrane was critical for detection of the monomeric Aβ peptides (urea gel electrophoresis), but hindered the detection of the higher molecular weight Aβ aggregates (and thus was avoided for those blots). Membranes were blocked in TBS containing 1% BSA and probed overnight (4 °C) with the 6E10 antibody.
Thioflavin T (ThT) fluorescence is thought to reflect binding of ThT dye to putative β-sheet structures associated with amyloid fibril formation 62  The analysis of protein secondary structure using Circular Dichroism (CD) spectroscopy is based on the differential absorption of polarized light by optically active molecules. CD measurements were carried out on a Pistar-180 CD spectrometer (Applied Photophysics Ltd.) at 25 °C using a 0.1 cm optical path-length quartz cuvette. Aβ solutions were scanned from 260-190 nm in 0.5 nm steps at a scan rate of 5 nm/min and a bandwidth of 6 nm. The CD spectrometer was calibrated with 10-camphorsulphonic acid and spectra were backgroundsubtracted using PBS, pH 7.4. Depicted CD spectra are the average of three spectra smoothed using a five-point Savitsky-Golay smoothing algorithm 63 . Deconvolution was performed using BeStSel 64 .
Dynamic Light Scattering (DLS) determines the size distribution of particles based on the proportion of incident light scattered, e.g. the larger the particles, the greater the scattering. DLS measurements were carried out in a quartz cuvette (Hellma Analytics) on a Dyna-Pro MS800 instrument (Wyatt Technologies) at 25 °C using an 824.8 nm (55 mW) laser diode. Scattered light was collected at 90° with an Avalanche photodiode detector. Data were acquired for 5 s and analyzed with DYNAMICS software (Wyatt Technologies).
Surface Plasmon Resonance (SPR) is a cell-free, optical technique used for monitoring real-time molecular interactions between a ligand immobilized on the surface of a flow cell and an injected analyte solution. SPR experiments were performed on a Proteon XPR36 (Bio-Rad) instrument. Standard amine-coupling chemistry was used to immobilize Aβ peptides onto a GLC sensor chip 65 . Briefly, sensor surfaces were activated using a solution of 20 mM 1-Ethyl-3-(3-Dimethylaminopropyl) carbodiimide: 5 mM sulfo-N-hydroxysulfosuccinimide injected for 5 min at 30 µL/min. After activation, individual Aβ peptides (50 µg/mL in 10 mM acetate buffer, pH 4.0-4.5) were injected at 25 µL/min for 5 min, after which any unoccupied succinimide sites were deactivated with an injection of ethanolamine (1 M, pH 8; 5 min, 30 µL/min). A reference surface was generated in the absence of Aβ peptide(s). The analyte solution was injected over the flow cell at 5 µL/min (120 s) and then replaced with wash buffer (20 min); any interaction between the test and immobilized proteins yields a sensorgram with association and dissociation phases.
Functional experiments. Mitochondrial metabolic activity based on the MTT conversion assay was used as an index of cell viability. The immortalized mouse hippocampal HT-22 cell line 66 was cultured in DMEM/ low glucose medium containing 10% fetal bovine serum (FBS). Human skin fibroblasts from female in-patients without metabolic disease were obtained from the Montreal Children's Hospital Cell Repository and have been characterized elsewhere 67  www.nature.com/scientificreports/ Cells (10,000/well) were treated with Aβ peptides (24 h) and the conversion of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; 0.5 mg/mL; 2 h; 37 °C; 5% CO 2 ) to the formazan product was quantified by spectrophotometry (absorbance = 570 nm) 68 .
Electrophysiological paradigms, e.g. LTP (synaptic plasticity) and whole cell patch-clamping (membrane conductance) were used to assess the functional impact of Aβ peptide mixtures.
LTP in acute hippocampal slice preparations: These procedures were approved by Memorial University's Animal Care Committee (PI: MPP). Four-to six-week old male C57BL/6 mice (Charles River Laboratories, Inc) were anesthetized using isoflurane and brains were quickly removed into ice-cold oxygenated slicing solution (in mM): 125 NaCl; 2.5 KCl; 25 NaHCO 3 ; 1.25 NaH 2 PO 4 ; 2.5 MgCl 2 ; 0.5 CaCl 2 ; and 10 glucose 69 . Transverse hippocampal slices (350 µm) were immediately transferred to artificial cerebrospinal fluid (aCSF), which had the same formulation as the slicing solution except for MgCl 2 (1 mM) and CaCl 2 (2 mM). Slices were allowed to recover (90 min, RT) and then transferred to a recording chamber. Oxygenated aCSF was continuously perfused at a flow rate of 1-2 ml/min (25 °C). Glass pipettes were pulled using a Narishige PB-7 pipette puller to a resistance of 1-3 MΩ when filled with aCSF.
Stimulation was applied to the Schaffer collaterals through a glass pipette using an Iso-flex stimulator and field excitatory postsynaptic potentials (fEPSPs) were recorded by placing a glass recording electrode in the CA1 stratum radiatum, approximately 400 µm from the stimulating electrode. A quick input/output plot was generated for each individual slice by increasing the stimulation intensity and an intensity that elicited 30-40% of maximum slope was used for the experiment. A stable baseline was established using 0.1 ms pulses at a frequency of 0.33 Hz before bath-applying Aβ  or Aβ(1-38), either alone or in combination, for 20 min before LTP induction, which consisted of a standard theta burst stimulation protocol of 10 bursts of 4 pulses at 100 Hz, with 200 ms interburst intervals. Aβ application continued for 5 min after LTP induction. Recordings continued for 60 min after LTP induction and the percent potentiation was analyzed as the average percent increase in the initial 1-2 ms of the fEPSP slope for the last 5 min of the experiment (55-60 min post-induction) compared to the 10 min of stable baseline prior to induction. All data were collected and analyzed using pClamp 10 software (Molecular Devices) and GraphPad PRISM.
Whole Cell Patch-Clamp Electrophysiology in isolated hippocampal neurons: These protocols conformed to the guidelines approved by the President's Committee on Animal Care, University of Regina (PI: JB). One-year old mice were sacrificed by pentobarbital overdose (120 mg/kg) and hippocampi were removed to PBS containing 4% Penicillin-Streptomycin and dissociated in sterile collagenase solution (1 h; 37 °C) 70 . Isolated neurons were plated onto Matrigel-coated coverslips and astrocyte-conditioned medium was provided to increase concentration of growth factors. Cultures were maintained at 37 °C in humidified 5% CO 2 .
Nystatin-perforated patch recordings were made using an AxoPatch 200B, and signals were filtered with a low-pass 5 kHz filter, digitized (Digidata 1550 series) and analyzed using Clamfit 10.7 software (Molecular Devices) as previously described 71 . The series resistance was compensated and junction potentials were cancelled in all experiments. Cells were held at − 60 mV, step depolarized to the indicated test potential (between − 80 and + 100 mV in 20 mV increments) for 50 ms, and sampled at a frequency of 1000 Hz. Patch pipettes were pulled (PC-10, Narishige International) to a resistance of 6-8 MΩ when filled with an internal pipette recording solution containing (in mM): 135 KCl; 5 NaCl; 2 CaCl 2 ; 10 HEPES (pH 7.2), and nystatin (250-500 µg/mL). The extracellular recording solution contained (in mM): 135 NaCl; 5 KCl; 2 CaCl 2 ; 2 MgCl 2 ; 10 D-glucose; 10 HEPES (pH 7.4). The hippocampal neurons were treated with Aβ  or Aβ(1-38) alone or in combination, and current density (e.g. current normalized to the cell capacitance) 24-h post-exposure to Aβ peptides was measured at 0 and 20 mV, 100 ms after the voltage step, in terms of mean I-V curves of steady-state currents.
The GMC101 strain 28 expresses the AD-related Aβ  under the control of the muscle-specific unc-54 (heavy chain muscle myosin) enhancer, and switching these worms from 20 to 25 °C triggers intracellular accumulation of Aβ(1-42) within the body wall muscle, eventually leading to paralysis 73 . Such defects in movement can be quantified using 'thrashing behaviour' , e.g. the number of complete bends of the dorsal or ventral side of the animal. Basically, a worm is transferred to a drop of M9 buffer, allowed to acclimate (20 s), and then the number of body bends in a 30-s test period is recorded manually. Thus, the thrashing rate can be used as a proxy for compromised muscle function due to Aβ accumulation. Each test group relied on a minimum of 60 worms per time point.