Aβ oligomer concentration in mouse and human brain and its drug-induced reduction ex vivo

Summary The elimination of amyloid beta (Aβ) oligomers is a promising strategy for therapeutic drug development of Alzheimer’s disease (AD). AD mouse models that develop Aβ pathology have been used to demonstrate in vivo efficacy of compounds that later failed in clinical development. Here, we analyze the concentration and size distribution of Aβ oligomers in different transgenic mouse models of AD and in human brain samples by surface-based fluorescence intensity distribution analysis (sFIDA), a highly sensitive method for detecting and quantitating protein aggregates. We demonstrate dose- and time-dependent oligomer elimination by the compound RD2 in mouse and human AD brain homogenates as sources of native Aβ oligomers. Such ex vivo target engagement analyses with mouse- and human-brain-derived oligomers have the potential to enhance the translational value from pre-clinical proof-of-concept studies to clinical trials.


In brief
Eliminating amyloid beta (Ab) oligomers is a promising strategy for therapeutic drug development of Alzheimer's disease (AD). Here, Kass et al. quantitate Ab oligomers in brain homogenates from various AD murine and human tissue samples and demonstrate the dose-and timedependent disassembly of Ab oligomers by the compound RD2.

INTRODUCTION
For 2020, the number of worldwide dementia cases was estimated to exceed 50 million, 1 with Alzheimer's disease (AD) being responsible for 60%-80% of all cases of dementia. 2 The disease's pathology is characterized by plaques consisting of amyloid beta (Ab) fibrils in the extracellular space, neurofibrillary tangles composed of hyperphosphorylated tau protein fibrils inside neurons, and neurodegeneration. Still, no curative treatment of AD is available. There is agreement that by the time first cognitive symptoms become noticeable, the disease process has already been going on for decades. 3,4 Soluble oligomeric forms of Ab are thought to be the most toxic species and have been described to be especially synapto-and neurotoxic. 5,6 Ab oligomers are, therefore, a very attractive target for curative therapy approaches as well as for early diagnosis.
During the last years, we have developed compounds that are designed to stabilize Ab monomers in their native, intrinsically disordered conformation. Thereby, the drug candidates destabilize Ab oligomers and other Ab assemblies and ultimately disassemble them directly into native Ab monomers. In order to achieve this mode of action, we use all-D-enantiomeric peptides, which are known to be protease-resistant 7 and non-immunogenic. 8,9 The lead compound, D3, was selected by mirror-image phage display 10 and was shown to reduce Ab aggregation and neuroinflammation and to improve cognition in a mouse model of AD even when applied orally. 11, 12 Since then, numerous derivatives of D3 have been developed in order to optimize its binding properties and pharmacokinetic properties. [13][14][15] The most promising and clinically most advanced candidate is RD2. It is well characterized in terms of binding mode, target engagement, efficiency, 16- 18 and pharmacokinetics. 19 Oral treatment with RD2 improved cognition in different mouse models of AD, 18,20 even in old-aged mice with full-blown pathology. 21 In the latter study 21 , we demonstrated that RD2 treatment significantly reduced the concentration of Ab oligomers, as measured by surface-based fluorescence intensity distribution analysis (sFIDA) in brain homogenates.
sFIDA realizes absolute specificity for Ab aggregates over Ab monomers. It achieves single aggregate particle sensitivity by combining the biochemical principle of a sandwich-ELISA with the readout of fluorescence intensity per pixel as obtained from fluorescence microscopy. Originally developed for the detection of prion protein aggregates, 22 sFIDA has been adapted for the quantitation of Ab oligomers in cerebrospinal fluid (CSF) 23 and blood 24 and is in further development as a general tool for quantitating all possible protein aggregates. 25 sFIDA is specific for aggregates by using capture and detection antibodies that recognize overlapping or identical epitopes of the aggregated protein of interest. Mostly, two different fluorescence-labeled detection antibodies are used, and total internal reflection fluorescence (TIRF) microscopy images are recorded in both channels directly at the glass surface, providing superior single particle sensitivity compared with the ensemble signal used in ELISA-type assays. Only pixels above a certain intensity threshold that are co-localized in both channels are counted (indicated as sFIDA readout), thereby ruling out possible unspecific signal of any of the used antibodies. Based on the sFIDA readout, concentrations can be calculated using a calibration standard, such as silica nanoparticles (SiNaPs) of a defined size, covalently coated with the capture and detection antibody-relevant epitopes. 26,27 Here, we set out to characterize the amounts and the size distributions of Ab oligomers in various amyloid-based animal models to compare them with each other and with humanbrain-derived Ab oligomers. Also, we demonstrate the usefulness of sFIDA to measure Ab oligomer target engagement of the oligomer-eliminating compound RD2 in human brain homogenates. Such ex vivo target engagement based on patientderived brain tissue (ex vivo) may well be suitable to enhance the translational value of pre-clinical in vivo experiments toward clinical trials.

RESULTS
Comparison of the concentrations of Ab oligomers in density gradient centrifugation (DGC) fractions and in unfractionated brain homogenates from different mouse models of AD Recently, sFIDA assay was adapted for quantitative detection of Ab aggregates in complex matrices, such as brain homogenate, to demonstrate in vivo target engagement and validate the mechanism of action for RD2. 21 Brain homogenates were fractionated by density gradient centrifugation (DGC) prior to analysis by the sFIDA assay. In the current study, we analyzed specimens of three mouse models expressing different human APP variants based on human familial mutations in comparison with wild-type mice. Ab oligomer concentrations were calculated based on SiNaP calibration standards and are displayed in Figure 1A as concentrations in undiluted brain homogenate or DGC-obtained fractions, resulting in an apparent concentration of 1 pM in unfractionated wild-type brain homogenate. The average sFIDA readout observed in wild-type DGC fractions barely exceeded that of the buffer control. The antibodies IC16 and Nab228 used in this assay are specific for human Ab, which is absent in wild-type mice. This outcome matched the expectations for wild-type mice as a negative control. Wild types were, therefore, not included in the calculation of relative oligomer concentrations, shown in Figure 1B. Western blot analysis performed with antibody 6E10, which is also specific for human Ab, also did not yield specific bands in wild-type DGC fractions, as displayed in Figure 1C.
In transgenic mouse samples, the highest Ab oligomer titers were found in fractions 9 (APP SwDI ) or 10 (APP/PS1 and APP Lon ), with mean concentrations of up to 16.0 ± 9.7 pM in APP SwDI , 790 ± 190 pM in APP/PS1, and 300 ± 150 pM in APP Lon samples. Samples showed high inter-individual heterogeneity, and the respective peak fractions made up 26.6% ± 16.2% (APP SwDI ), 41.8% ± 9.9% (APP/PS1), or 30.3% ± 14.5% (APP Lon ) of all oligomers measured in DGC fractions of the respective mouse models Figure 1B. A local maximum was identified in all transgenic mouse models, covering fractions 4 and 5. The amounts of oligomers found in these two fractions together made up 13% (APP SwDI ), 2.8% (APP/PS1), and 6.9% (APP Lon ) of the total oligomer concentration. In this type of density gradient, calibrated with globular proteins, fractions 4 and 5 would correspond to a size of 66-150 kDa, and fractions 9 and 10 to a size of at least 400 kDa, respectively. 28 The overall distribution of total Ab detected via western blot with antibody 6E10, shown in Figure 1C, was similar to the sFIDA results of the different mouse models. However, APP SwDI Ab bands were of equal or higher intensity than those of the other two mouse models, although the Ab oligomer concentration measured by sFIDA in this mouse model was substantially lower. All three antibodies worked equally well in western blot ( Figure S1), so major differences in the general detectability of denatured DI-Ab by IC16 and Nab228 in comparison with 6E10 were ruled out as a possible reason for this observed difference. Recovery was calculated as the ratio of the total amount of oligomers measured by sFIDA assay in all fractions to the total amount of oligomers in the corresponding unfractionated homogenates. The respective recovery rates were 0.986 for APP SwDI , 1.634 for APP/PS1, and 0.975 for APP Lon samples. Overall, this shows that the chosen dilutions of 1:10 for DGC fractions and 1:100 for 10% brain homogenates were well suited for quantitation of oligomers in APP SwDI , APP/PS1, and APP Lon mice, while applying less diluted wildtype samples did not cause any artifacts. Small amounts of sample would therefore be sufficient to investigate in vivo target engagement of oligomer-eliminating compounds in several different mouse models, in a similar fashion to Schemmert et al. 21 Notable differences of the concentration of Ab oligomers between human AD samples and nondemented controls The previously analyzed transgenic mouse models of AD play an important role in the development of therapeutic compounds that would ultimately be used in human patients. In order to mimic the clinical situation more accurately, we investigated the concentrations and size distributions of Ab oligomers in post mortem brain homogenates of human AD patients as well as in age-matched, non-demented control subjects (NCs). Details of human brain donors can be found in Table 1. While the four NCs had similar oligomer concentrations, the AD samples showed large differences between each other, as indicated in Figures 2A and 2B. In general, the Ab oligomer concentration in the NC group was more than 10-fold lower than the lowest oligomer concentration in the AD group but clearly exceeding the limit of detection (LOD) of 7.6 fM and lower limit of quantitation (LLOQ) of 9.7 fM With the exception of sample AD3, DGC fraction 10 contained the highest Ab oligomer concentrations, ranging from 19.9 ± 3.8 pM (AD1) to 1820 ± 191 pM (AD6) and 0.9 ± 0.7 pM in NC samples. These oligomer particles, corresponding to a calibrated size of more than 450 kDa, 28 made up 33.9% ± 6.5% to 54.5% ± 5.7% in all AD samples, except for AD3, and 23.4% ± 1.8% in NC samples, as depicted in Figure 2C. In sample AD3, the highest oligomer concentration (50.3 ± 9.1 pM) was found in fraction 11 and a concentration of 42.7 ± 1.1 pM in fraction 10, which corresponded to 33.8% ± 6.1% and 28.6% ± 0.3% of all oligomers, respectively. Similar to the observations in transgenic mouse samples, local maxima were identified. A local maximum was found in fraction 5 in almost all samples, AD1 being the only exception, with a local maximum in fraction 4. The size of these oligomers agrees well with the size of artificially prepared Ab oligomers that have not yet elongated and do not contain other components besides Ab. 28 The percentage of total oligomers found in the respective local maximum fraction was in the range of 0.6% ± 0.1% to 2.3% ± 0.1% for AD samples and 5.5% ± 0.6% for NC samples. In a corresponding western blot of total Ab, no bands were detectable in the NC2 sample, with the distribution of total Ab across the density gradient matching the sFIDA results for the AD4 sample ( Figure 2D; uncropped western blot images can be found in Figure S2). Recovery rates were 1.15, 2.02, 1.61, 2.40, 0.86, and 4.06 for AD1 to AD6, respectively. Here, the sFIDA assay has demonstrated its usefulness to determine concentrations of Ab oligomers, even in complex samples like brain homogenates with a wide range of oligomer particle sizes. While the diversity of oligomer concentrations found in human samples was much larger compared with transgenic mice, the overall size distribution was very similar to that of aged APP/PS1 mice. Recovery rates exceeding 1 by a larger margin could be due to underestimation of the oligomer concentration in unfractionated brain homogenates. This effect occurred mostly in the three AD samples with the highest oligomer concentration. As a consequence, greater dilution factors of unfractionated homogenate have been used in the following experiments to reduce the possible influence of other proteins and to avoid saturation effects, while still staying in the quantifiable, linear range of the assay.
Ex vivo treatment with RD2 results in a dose-and incubation-time-dependent reduction of Ab oligomers derived from APP/PS1 mouse or human brain homogenate with RD2 The all-D-enantiomeric compound RD2 has been developed to stabilize Ab monomers in their native conformation, thus destabilizing Ab assemblies and ultimately disassembling them into Ab monomers. RD2 has been shown to improve cognition in several AD mouse models, 18,20,21 and its oligomer-eliminating effect has been well characterized using synthetic Ab. 17,18 RD2 did not display any toxic effects in mice or in cell cultures, indicating that the possible disruption of HMW oligomer species by RD2 did not produce smaller, toxic oligomer species but non-toxic monomers. 18,21 In a study conducted in APP/PS1 mice, a reduction of Ab oligomer concentration was measured by the sFIDA assay in brain homogenates of RD2-treated mice, demonstrating in vivo target engagement and suggesting elimination of these oligomers by RD2. 21 That study showed that oligomers were significantly reduced in brain homogenates of APP/PS1 mice after oral treatment with RD2 in comparison with placebo. DGC fraction 10 contained the highest concentration of Ab oligomers. We therefore first investigated the effect of RD2 on this isolated fraction 10 in an ex vivo approach. RD2 or buffer was added to 1:2 diluted fraction 10 of APP/PS1 mouse brain homogenate, and samples were drawn after the indicated incubation times. In cases of ''0 min'' incubation time, samples were mixed and immediately flash-frozen. As shown in Figure 3, RD2 was able to reduce the concentration of oligomers in a dose-and time-dependent manner. After the maximum incubation time of 20 h, an 81% reduction of oligomers was achieved with 50 mM, and 67% reduction was observed with 20 mM. The incubation time dependence clearly indicates that the RD2 dose dependence of oligomer reduction is not due to competition with the detection antibodies. A slight effect could also be seen with 10 mM RD2, yielding a reduction of about 8% after 20 h, but this change was not statistically significant. It must be noted that samples incubated with 20 and 50 mM RD2 showed a remarkably reduced oligomer concentration even without additional incubation time ( Figure 3B, 0 min): the baseline oligomer concentration of the sample incubated with buffer was 88 ± 6 pM, whereas samples with the addition of 20 or 50 mM RD2 had Ab oligomer concentrations of 48 ± 3 pM and 36 ± 2 pM, respectively. A possible explanation could be an ongoing reaction of RD2 with oligomers during sample incubation on the capture surface, thereby prolonging the effective reaction time by about 2 h. Another possible explanation would be that the initial elimination of oligomers is a fast process and that the delay of several minutes during the sample preparation steps due to freezing and thawing the samples before analysis would make observation of a true baseline value difficult. To address these questions and in order to rule out any delays during sample preparation, freeze-thaw cycles were omitted in further experiments involving short incubation times and 0 min marks. The dose-dependent reduction of Ab oligomers from APP/PS1 mouse brain strongly suggests successful ex vivo target engagement of RD2.
Next, we wanted to investigate whether RD2 shows similar ex vivo target engagement on Ab oligomers in brain homogenates derived from AD patients. Based on the aforementioned sFIDA analysis of several human brain homogenates (Figure 2), homogenate sample AD2 was chosen for further analysis, because this sample had one of the largest quantities of Ab aggregates, allowing robust detection of signal across a range of dilutions, even considering the possible drastic reduction of sFIDA signal by the addition of RD2. The 10% homogenate sample was serially diluted before being incubated with different concentrations of RD2, ranging from 0.31 to 20 mM. In addition to buffer as a negative control, we chose another all-D-enantiomeric compound of similar size, D1, as a negative control peptide. D1 was originally selected for binding Ab fibrils, but not for specific elimination of oligomers, 29 and was further developed as a positron emission tomography (PET) tracer rather than a therapeutic compound. [30][31][32] After an incubation time of 24 h, samples were subjected to sFIDA assay, and concentrations were calculated with SiNaP standards. Refer to Figure S3A for representative TIRF images. The 1:10 diluted brain homogenate that was incubated only with buffer, shown in Figure 4A, had a concentration of 47.1 ± 20.0 pM Ab oligomers. Samples incubated with D1 showed an overall reduced concentration of 30.2 ± 4.1 pM to 37.1 ± 3.7 pM, but no dose-dependent effect was observed. The same could be observed for the three lowest concentrations of RD2. The addition of 20 mM RD2 resulted in a drastic, significant reduction of the Ab oligomer concentration down to 3.3 ± 0.4 pM, corresponding to a reduction by 93% compared with the buffer control. The Ab oligomer concentration in 1:20 diluted homogenate was 12.8 ± 1.8 pM in the sample incubated with buffer (Figure 4B). Addition of D1 of any concentration or of 0.31 and 1.25 mM RD2 resulted in a similar reduction as observed in the 1:10 diluted homogenate sample, with no significant dosedependent effect. Effects of 5 and 20 mM RD2 were distinct, with a reduction of oligomers to 4.9 ± 1.0 pM (61%) and 0.6 ± 0.1 pM (95%), respectively. The results for 1:40 diluted brain homogenate are shown in Figure 4C: after 24 h incubation with 5 and 20 mM RD2, the concentration of Ab oligomers was reduced to 0.7 ± 0.1 pM (84%) and 0.2 ± 0.04 pM (96%) from 4.1 ± 1.1 pM. A ten-dency of oligomer reduction was also observed with 0.31 and 1.25 mM RD2, with a reduction to 2.7 ± 0.1 pM (36%) and 2.1 ± 0.1 pM (50%), respectively, of which only 1.25 mM RD2 caused an oligomer reduction that was significantly different from that observed with all concentrations of the control peptide D1. Different degrees of signal reduction were observed with the control peptide D1, ranging from 4% to 27% with no evidence of a dose-effect. Despite a certain reduction of Ab oligomers in D1treated samples, this effect was not dose-dependent, leading to the conclusion that only RD2 showed a specific reduction effect on Ab oligomers in human AD brain homogenate.
Time-and dose-dependent elimination of oligomers in human brain homogenate by RD2 but not by control Dpeptides D1 and QB37 To gain more insight into the dynamics of the ex vivo target engagement of RD2 in human AD brain homogenate, we . Dose-and incubation-time-dependent effect of RD2 on fractionated APP/PS1 mouse brain homogenate DGC fraction 10 of APP/PS1 mouse brain homogenate, representing the peak of sFIDA signal, was diluted 1:2 and was incubated with 0, 5, 10, 20, or 50 mM RD2. Before analysis, samples were further diluted 1:5, resulting in a total dilution of 1:10, and the Ab oligomer concentration was immediately analyzed by sFIDA assay. The final dilution factor during image acquisition was 1:20. monitored the Ab oligomer concentration in diluted homogenates from four different AD patients with incubation times ranging from 0 to 23 h. These samples were chosen because of their high concentration of Ab oligomers, as determined previously, to allow monitoring of changes over a wide signal range. In addition to the previously tested peptide D1, QB37 was added as an additional control D-peptide of similar size to RD2 and D1. The homogenate was used at a dilution of 1:20; both control peptides were used at a concentration of 20 mM; and RD2 concentrations were 1.25, 5, and 20 mM. Figures 5A, 5C, 5E, and 5G show baseline concentrations measured by sFIDA assay after 0 min incubation time and endpoint concentrations after 1,365 min (AD2); 1,3,57 min (AD4 and AD5); or 1,425 min (AD6) of incubation; for representative TIRF images of samples AD2 and AD6 at their endpoints, see Figure S3B. At baseline, none of the peptides, including RD2, caused a reduction of Ab oligomers in comparison with the buffer controls. The design of sFIDA assay includes an incubation time of at least 1.5 h after the indicated pre-incubation times. Without any pre-incubation time (0 min), no sFIDA signal reduction was observed. This indicates that the signal reduction observed after longer incubation times was attributed to the reaction of the Ab oligomers with RD2 but that no additional reaction took place once the oligomers were captured on the sFIDA plate. After approximately 23 h of incubation, samples incubated with 5 and 20 mM RD2 showed a significant decrease in oligomer concentration to 1.0 ± 0.3 pM (57% decrease) and 0.1 ± 0.1 pM (93%) for AD2, 0.41 ± 0.03 pM (81%) and 0.19 ± 0.03 pM (91%) for AD4, 0.44 ± 0.05 pM (73%) and 0.18 ± 0.02 pM (89%) for AD5, or 3.6 ± 0.4 pM (64%) and 1.0 ± 0.2 pM (90%) for AD6 compared with the buffer control. Results for additional incubation times of these two RD2 concentrations are shown in Figures 5B, 5D, 5E, and 5H, demonstrating a rapid decrease of oligomer concentration within the first 40 to 50 min of the reaction with 20 mM RD2, followed by a considerably slower further reduction of the oligomer concentration in the following hours. In order to describe our observations using a global kinetic fit based on a pseudo-first-order reaction, a double exponential decay function was used, essen-tially reflecting a combination of two reactions taking place at different rates. Importantly, a threshold concentration of RD2 was assumed for these calculations, because an effect was observed only with a dose of 5 mM but not with 1.25 mM RD2 or lower. Due to possible binding of RD2 to various other components of brain homogenate, the effective RD2 concentrations were expected to be considerably smaller than the total concentration. The global fits shown in Figures 5B, 5D, 5F, and 5H are therefore based on a threshold concentration of 4 mM, meaning that the remaining effective RD2 concentrations were 1 and 16 mM instead of 5 and 20 mM, respectively. Reaction rate constants k 1,fast of 2,329; 2,928; 2,374; and 1,275 L*mol À1 *min À1 were calculated for AD2, AD4, AD5, and AD6, respectively. The reaction rate constants k 1,slow were 76, 37, 50, and 48 L*mol À1 *min À1 for AD2, AD4, AD5, and AD6. Similar to our previous observation, shown in Figure 4, with samples from AD2, three of the samples showed a reduction of Ab oligomers with 20 mM D1 as well. The effect was much weaker than with 5 mM RD2 and is most likely not specific, due to the absence of a dose-dependent effect, as stated before (Figure 4).
Effect of RD2 on different DGC fractions of human brain homogenate Fractions 5 and 10 of AD brain homogenate were identified as local and total Ab oligomer concentration peak fractions, indicating distinct sizes and possibly different types of aggregates. To investigate the effect of RD2 on these fractions while staying in the quantifiable concentration range of the assay, fraction 10 was diluted 1:10, and fractions 4 to 6 were pooled with no further dilution. While fractions 4 to 6 would possibly represent oligomers with sizes similar to that of synthetic Ab oligomers in absence of other proteins, in fraction 10, larger Ab assembly species are present that possibly also consist of additional proteins and form co-aggregates. Fraction 10 was of particular interest because the reduction of oligomers found in this fraction correlated with improved cognition in transgenic mice treated with RD2. 21 For incubation with fraction 10, RD2 was used at 5 and 20 mM, and pooled fractions 4 to 6 were incubated with 20 mM  Figure S4A. After 0 min of pre-incubation, baseline concentration of fractions without peptides was around 11 pM for AD2 and 8 pM for AD6. A slight decrease in concentration of these samples was observed after 327 and 1,380 min to around 9 pM (AD2) and 6 pM (AD6). In contrast to the previous findings on unfractionated homogenates, fraction 10 of both samples showed a reduction of oligomer concentration by approximately 50% already at 0 min pre-incubation time with 5 mM RD2 and over 60% with 20 mM RD2. The concentration of Ab oligomers in RD2-treated samples decreased further over the course of 1,380 min, down to 1.7 ± 0.1 pM (5 mM RD2, 81%) and 0.7 ± 0.1 pM (20 mM RD2, 92%) in AD2 and 1.4 ± 0.1 pM (5 mM, 77%) and 0.7 ± 0.1 pM (20 mM, 88%) in AD6. With 20 mM RD2, the endpoint values were comparable with those found in unfractionated homogenate, whereas 5 mM RD2 had an overall slightly larger effect on fraction 10 than on homogenate. The time-and dose-dependent response observed here with DGC-derived fraction 10 of human AD brain homogenate as well as the seemingly instant reduction of oligomers without additional pre-incubation time is very similar to the one observed in fraction 10 of APP/PS1 mouse brain homogenate before (Figure 3). The pooled fractions 4 to 6 showed low baseline values of 0.35 ± 0.03 pM (AD2) and 0.51 ± 0.04 pM (AD6), as presented in Figures 6C and 6D (Figure S4B: representative TIRF images). Baseline concentrations were identical for the samples incubated with buffer and samples with 20 mM RD2. After 1,380 min, the concentrations of the samples incubated with 20 mM RD2 were notably reduced to 0.12 ± 0.01 pM (AD2, 64%) and 0.14 ± 0.01 pM (AD6, 62%).

DISCUSSION
Size distribution of Ab aggregates in relevant AD mouse models and patient-derived samples Transgenic mouse models of AD are commonly used for understanding disease development and testing potential diseasemodifying drugs. However, there is no single mouse model of choice that displays the complete and holistic pathology of AD, including Ab plaques, tangles, neurodegeneration, and cognitive decline. Therefore, several different mouse models are used to investigate the efficacy of potential drug candidates. We analyzed the Ab oligomer size distributions in brain homogenates of aged specimens of three different mouse models of AD and human AD cases, using the aggregate-specific sFIDA assay in combination with DGC. Our choice of extraction and analysis methods was intended to keep all aggregates as native as possible. In general, a majority of Ab aggregates were found in fractions 9 or 10, corresponding to a calibrated size of more than 400 kDa in all transgenic mice and in human AD as well as NC samples. Ab aggregates of similar size have been reported in transgenic mice and human AD cases before, using methods, such as size-exclusion chromatography (SEC), blue native blots, or density gradients. They have been found to have oligomeric or protofibrillar conformation [33][34][35] and have been postulated to serve as a reservoir for smaller, more diffusible toxic oligomers rather than being toxic themselves. 36,37 Fractions 4 to 6 were of special interest because they contained especially neurotoxic oligomers between 66 and 150 kDa in experiments with synthetic Ab 28 and presented a local maximum in all transgenic mice and human samples. Oligomers in this size range might therefore have particular biological significance despite their lower relative concentration in the samples.
It is also likely that different co-aggregates of Ab are found in different fractions. Oligomer binding proteins, such as apolipoprotein E (ApoE), influence the apparent aggregate size of Ab assemblies consisting of otherwise similar numbers of Ab units. 38 The investigation of the composition and function of potential co-aggregates in these fractions is a subject of further research.
In comparison with the strong bands observed in the western blot using the antibody 6E10, the concentrations of Ab oligomers from APP SwDI mouse brains measured by sFIDA assay using the antibodies Nab228 and IC16 were lower than expected, albeit still within the dynamic range of the assay. A control of the western blot using antibodies Nab228 and IC16 also yielded strong bands ( Figure S1). These findings suggest that Ab oligomers from APP SwDI brain homogenate samples have less accessible epitopes for the antibodies used in the sFIDA but that these epitopes were released by the denaturing conditions of the SDS-PAGE. In contrast to APP Lon and APP/PS1 mice, the Ab variant expressed in APP SwDI mice contains two amino acid residue replacements (E22Q/D23N). This highly artificial APP variant, which does not exist in humans, possibly results in Ab monomers that form oligomers with conformations that are different from those obtained from other transgenic models and humans. Such different conformations possibly result in reduced accessibility of the epitopes for the respective antibodies in non-denaturing conditions, even though the epitopes as such are not affected by the APP mutations. Still, we were able to clearly discriminate APP SwDI samples from wild-type samples in all fractions at a 1:10 dilution. All transgenic mice tested here showed full-blown pathology, which would be relevant for curative studies like the one conducted by Schemmert et al., 21 in which sFIDA assay was used for the first time to monitor in vivo target engagement. The development and possible changes in aggregate size can be monitored in mice of different ages to further characterize existing or novel mouse models of AD.
Data are presented as mean ± SD N = 3 (technical replicates). *between groups, p < 0.05; #versus buffer, p < 0.05; xversus both control peptides D1 and QB37, p Human AD brain homogenates showed much larger inter-individual variability of their total Ab oligomer concentrations compared with those of transgenic mice, which was expectable due to the generally larger variation among human patients compared with inbred mouse strains. It is also possible that the location from which the samples were taken for homogenization affected the total amount of aggregates later found in the sample: All human samples used in the present study were from the superior parietal lobule, but inhomogeneity of the local distribution of Ab might still occur, which would not be noticeable in mice of which the whole brains were used. Although mostly lacking Ab pathology, NC samples also showed low amounts of oligomers with a maximum in fraction 10 as well. The occurrence of these miniscule amounts of oligomers is in line with the fact that sporadic accumulation of Ab can generally be found long before cognitive symp- Figure 6. Effect of RD2 on different DGC fractions of human brain homogenate (A and B) DGC fraction 10 of human AD brain homogenate (AD cases 2 and 6), representing the peak of sFIDA signal, was diluted 1:10 and was incubated with 0, 5, or 20 mM RD2. After incubating for the indicated duration, the Ab oligomer concentration of the samples was analyzed by sFIDA assay. Concentrations given here reflect the actual concentrations in the prepared sample after dilution and are not directly related to the concentrations in undiluted samples, shown in Figure 2. Data are presented as mean concentration ±SD N = 3 (technical replicates). *between groups, p < 0.05; #versus buffer, p < 0.05 (Kruskal-Wallis one-way ANOVA on ranks with Student-Newman-Keuls post hoc analysis). (C and D) DGC fractions 4 to 6 of human AD brain homogenate (AD cases 2 and 6), representing a local maximum in the size distribution of Ab oligomers, were pooled and were incubated with 0 or 20 mM RD2 overnight. Samples were analyzed by sFIDA assay. Concentrations given here reflect the actual concentrations in the prepared sample after dilution and are not directly related to the concentrations in undiluted samples, shown in Figure 2. Data are presented as mean ± SD N = 3 (technical replicates). *between groups, p < 0.05 (two-tailed t test); TIRF images are displayed in Figure S4. toms would occur, if the afflicted person were to develop AD at all. 39 Regardless of the differences in absolute concentrations and in amyloid plaque pathology, the relative distribution of Ab oligomer in AD samples was mostly comparable with that of APP/PS1 mice, possibly indicating similar types of aggregates.
Ex vivo target engagement of the oligomer-eliminating compound RD2 There is an urgent need to assess the value of pre-clinical animal experiments for their predictability in clinical studies, especially in the field of AD. Previously, we described that oral treatment with RD2 in aged APP/PS1 mice with full-blown pathology yielded improvement of cognition, memory, and behavior. We rationalized that this outcome was based on the efficacy of the compound to directly eliminate toxic Ab oligomers, a mode of action that RD2 was designed and developed for. Indeed, a significant reduction of Ab oligomer concentration in fraction 10 of fractionated brain homogenate was found. 21 To further substantiate the rationale that RD2 will show the same efficacy also in AD patients, we demonstrated that the Ab oligomers obtained from brains of the APP/PS1 mouse model and from human patients do have very similar size distributions, suggesting that the target Ab oligomer has similar properties in both sources. The range of Ab oligomer concentrations found in AD-affected human brain samples overlapped with that of the APP/PS1 mouse model. This similarity in Ab oligomers in the animal experiment and in human patients may already give some degree of confidence to the translation of the pre-clinical efficacy data for clinical trials.
Even more confidence is obtained from the direct observation of RD2-eliminating Ab oligomers from APP/PS1 mouse brain homogenate ex vivo ( Figure 3) and from AD patient-derived brain homogenate (Figures 4 and 5). During the ex vivo target engagement of RD2 in unfractionated brain homogenates of AD patients, we found a dose-and time-dependent reduction of Ab aggregates. Two further D-peptides of similar size as RD2 were chosen as controls, D1 and QB37. None of these two peptides showed any dose-dependent effect, indicating that the observed reduction of Ab oligomer concentration was indeed specific for RD2. A closer look at earlier time points of homogenate at a 1:20 dilution with 20 mM RD2 ( Figure 5) revealed that the majority of the reaction took place in the first 40 min of pre-incubation time. Importantly, no reaction was observed without preincubation time, which means that no further reaction took place during the incubation with capture antibody on the sFIDA assay plate. It also shows that the observed effect is not attributed to any interaction of RD2 with the assay setup as such. The brain homogenates that we used here present a very complex matrix due to the abundance of different proteins that become released during the homogenization procedure. A considerable portion of RD2 is possibly bound to proteins other than Ab, while its positive net charge makes it also likely to interact with nucleic acids, glycoproteins, and membrane constituents, such as phospholipids. 40 In order to describe the time and dose dependency of the reduction of Ab oligomers by RD2 by global kinetic parameters, a concentration of 4 mM of RD2 considered as bound (''threshold dose'') was subtracted from the respective total RD2 concentrations. The best fit was achieved with a double exponential decay model, suggesting a fast and a slow reaction. This could be indicative of different types of Ab aggregates present in the sample that have different susceptibility to RD2. The calculated reaction rate constants, while generally in the same range, were global only within each sample but could potentially be used to compare different compounds. We observed ex vivo target engagement of RD2 in pooled DGC fractions 4 to 6 and in a 1:10 dilution of fraction 10 in human samples as well. Unlike fractions 4 to 6 and unfractionated homogenate, fraction 10 showed a notable decrease of Ab oligomer concentration without pre-incubating with RD2, indicating that the majority of the reaction either happened within less than a minute or during the interval of about 120 min, in which the capture and first washing steps were completed. The latter, however, seems less likely due to the aforementioned observations in unfractionated brain homogenate. On the one hand, this might be due to high susceptibility to RD2 of the particular oligomers in fraction 10. On the other hand, possible matrix effects should be taken into account. While the overall oligomer concentration was almost equally high in a 1:10 dilution of fraction 10 and a 1:20 dilution of homogenate, DGC fractionation generally results in a reduction of complexity and in a dilution of the total protein concentration.
The dose and time dependence of RD2 activity resembles that of an enzyme. This is well in agreement with the proposed mode of action of RD2. RD2 is designed to stabilize Ab monomers in their native intrinsically disordered conformation. Ab oligomers, therefore, are destabilized by RD2, which is ultimately disassembling Ab oligomers into monomers. Complexes of RD2 with Ab monomers are transient, but of high affinity in the nanomolar K D range, 17 and may be called fuzzy complexes. 41 The observed dose-and time-dependent Ab oligomer elimination by RD2 agrees well with the hypothesis that RD2 is acting similar to a chaperone 42 that folds Ab oligomers back into natively folded Ab monomers ( Figure S6).

CONCLUSION
sFIDA reliably allowed the reproducible measurement of Ab oligomer concentrations in post mortem brain tissues from transgenic AD mouse models and from human AD patients. The combination of sFIDA with particle-size-dependent fractionation of the respective brain tissue homogenates by DGC allowed the quantitative analysis of the particle size distribution of Ab oligomers. Although the absolute oligomer concentrations varied between individual AD patients and between the different mouse models by about two orders of magnitude, oligomer size distributions were very similar among human AD samples as well as between human AD samples and transgenic mice. The similarity of size distributions of Ab oligomers in AD mouse models and humans supports a translational value of beneficial effects for cognition observed in the respective animal model, at least for drug candidates that eliminate Ab oligomers, especially when animal models have been used that express human wild-type Ab.
Also, animal and human brain tissue can be used to assay for ex vivo target engagement of drug candidates designed for direct Ab oligomer elimination. Our results on ex vivo target engagement of RD2 support the findings of earlier in vitro and in vivo studies demonstrating the Ab oligomer elimination activity of RD2. 18,21 After ex vivo treatment with RD2, the concentrations of Ab oligomers in brain homogenates, as measured by sFIDA assay, were reduced. Due to the specificity of sFIDA assay for multimeric Ab assemblies, reduction of sFIDA-obtained oligomer concentrations suggests monomerization of existing, native Ab oligomers in the samples. The combination of DGC fractionation and sFIDA provides a general and sensitive tool for characterization and identification of different Ab (co-)aggregates and further Here, effective target engagement of the compound RD2 with human samples could be a promising indication of its efficacy in human patients. Safety in humans has been demonstrated already. 43 The principle of using ex vivo-obtained oligomers to test efficacy of oligomer-eliminating compounds using sFIDA can be translated to other disease-relevant oligomers, such as a-synuclein or tau oligomers.
Limitations of the study Our analysis of a limited number of human brain samples revealed variability concerning the individual Ab oligomer concentrations. Yet, we reliably observed ex vivo target engagement of RD2 in samples derived from different donors, supporting the mode of action and the general concept of RD2's ex vivo target engagement. The reduction of effective peptide concentrations by potential matrix effects in native human brain homogenates is also a point of concern. In future studies, further attempts may be undertaken to reduce this effect, for example, by using isolated DGC-obtained fractions with reduced matrix content.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Calculation of Ab oligomer concentration
The sFIDA readout of the SiNaP standards was used to perform a linear regression analysis. Concentrations were calculated based on this linear regression, reflecting the concentration of oligomer particles of a certain defined size and number of epitopes. Excel 2010 (Microsoft, USA) and OriginPro 9.4 (OriginLab, USA) were used for calculations and graphs.

Statistical analysis
All data are presented as mean ± standard deviation over triplicate wells in single sFIDA measurements (technical replicates), or mean ± standard deviation over the indicated number (N) of biological replicates. In cases where only few different biological samples were analyzed, or variation between samples was high, technical replicates are shown for each sample separately. The limit of detection (LOD) and lower limit of quantification (LLOQ) were defined as the concentration exceeding that of the blank sample by 3 or 10 standard deviations, respectively. Further statistical analyses were carried out in SigmaPlot 11.0 (Systat Software, Germany) and are summed up in Table S1. The reaction rate constant k 1 was fitted in SigmaPlot 11.0 based on the assumption that RD2 was present in large excess compared to the Ab oligomer concentration [O], so that the principles of a pseudo-first order reaction apply: To calculate k 1 directly and globally, a modified formula for a double exponential decay was used: ½O t = ½O fast Ã e À k 1; fast Ã½RD2 free Ãt + ½O slow Ã e À k 1; slow Ã½RD2 free Ãt Taking into account the observation that a certain threshold concentration of RD2 had to be exceeded to show any effect on oligomers in brain homogenate, the assumed free RD2 concentration [RD2] free was used for kinetic fits instead of the original concentrations.
[RD2] free was calculated by subtracting 4 mM from each of the used concentrations, yielding effective concentrations of 1 and 16 mM. Fits were calculated using the mean concentrations of three technical replicates.