An antibody scanning method for the detection of α-synuclein oligomers in the serum of Parkinson's disease patients

Misfolded α-synuclein oligomers are closely implicated in the pathology of Parkinson's disease and related synucleinopathies. The elusive nature of these aberrant assemblies makes it challenging to develop quantitative methods to detect them and modify their behavior. Existing detection methods use antibodies to bind α-synuclein aggregates in biofluids, although it remains challenging to raise antibodies against α-synuclein oligomers. To address this problem, we used an antibody scanning approach in which we designed a panel of 9 single-domain epitope-specific antibodies against α-synuclein. We screened these antibodies for their ability to inhibit the aggregation process of α-synuclein, finding that they affected the generation of α-synuclein oligomers to different extents. We then used these antibodies to investigate the size distribution and morphology of soluble α-synuclein aggregates in serum and cerebrospinal fluid samples from Parkinson's disease patients. Our results indicate that the approach that we present offers a promising route for the development of antibodies to characterize soluble α-synuclein aggregates in biofluids.


Introduction
The aggregation of a-synuclein is a hallmark of a family of neurodegenerative diseases collectively known as synucleinopathies, the most common one being Parkinson's disease (PD), 1 which is estimated to affect over 6 million people worldwide. 2 PD is characterized by the presence of Lewy bodies in the midbrain of affected individuals, which are proteinaceous deposits that contain a-synuclein aggregates. 3 Its close association with PD has made a-synuclein a central target for therapeutic and diagnostic interventions. [4][5][6] As the pathology of other misfolding diseases, soluble prebrillar aggregates, known as oligomers, rather than mature aggregates, have been identied as the driver of cytotoxicity. 7-10 However, not much is currently known about the structure and abundance of a-synuclein oligomers in normal and diseased brains. 11,12 In recent studies a-synuclein oligomers have been detected in the blood and cerebrospinal uid (CSF) of PD patients, [13][14][15][16] opening the way for the potential use of these particles as biomarkers both for diagnostics and surrogate endpoints in clinical trials. 12 Methods to detect these aggregates 12,17 include immunoassays such as ELISA [18][19][20] or single-molecule immunoassays, [21][22][23] which rely on the use of highly specic antibody pairs. [9][10][11] Commonly available antibodies used in these immunoassays target either linear epitopes at the highly immunogenic C-terminal region of a-synuclein, including phosphorylated forms, or aggregate conformations of this protein (e.g. mouse monoclonal antibodies targeting residues 121-125 of a-synuclein (SC), 18,19 a-synuclein phosphorylated at Ser129 (PS-129), 20,23 or oligomeric a-synuclein (Syn-O2) 20 ). However, the development of conformation-specic antibodies against pathogenic a-synuclein assemblies, including oligomers, remains challenging. 24 It has also proven difficult to target regions outside the C-terminus because of their low immunogenicity, in particular the central non-amyloidb component (NAC), which plays an important role in the aggregation process and formation of the intermediate aggregate species.
A recently introduced method of rational design of antibodies offers an opportunity to scan protein sequences to address this problem. 25,26 In previous studies, a battery of antibodies was generated using this approach against the amyloid b peptide (Ab), and subsequently graed onto a single-domain antibody scaffold. 27 Aer expression and purication, the designed antibodies (DesAbs) were characterized in different assays. 27 By selecting DesAbs with low affinity for monomeric and brillar forms, a conformation-specic antibody to Ab oligomers was identied. 28 It was also shown that, as different DesAbs bind different Ab epitopes, they can be used to characterize different Ab aggregates with different mechanisms of toxicity. 29,30 In this study, we applied this antibody design method to rationally design a panel of antibodies scanning the sequence of a-synuclein. Characterizing the binding of each DesAb in various biophysical assays including a kinetic aggregation assay and by uorescence microscopy allowed us to identify a DesAb capable of recognizing a-synuclein oligomers in the serum of PD patients. By using super-resolution microscopy, we estimated the size distribution of these oligomers, consistent with recent ndings that these species are conformationally heterogeneous. 31

Results and discussion
Generation of a library of DesAbs to scan the sequence of asynuclein In this study, we sought to apply a rational antibody design approach recently used to identify an oligomer specic antibody against Ab to scan the sequence of a-synuclein. Nine designed peptides to be used as paratopes targeting selected epitopes in the amphipathic and NAC regions of a-synuclein were thus selected. This approach produced, for each target epitope, complementary peptides (paratopes) by combining sequence fragments found to form b-sheet pairs with fragments of the target epitope in protein structures available in the Protein Data Bank (PDB). 25,26 To rank complementary peptides, each sequence is given a score comprising factors such as the frequency the b-sheet pairs are found in the PDB or the speci-city for each binding pair. 25,26 The solubility for the binding partners is evaluated by employing a sequence-based solubility calculation with the CamSol method 32 and DesAbs scoring a high solubility score are favored. The designed complementary peptides and the target sequence are shown in Fig. 1 with the position of the epitope labelled with regards to the sequence of a-synuclein. The fragments forming each peptide design are shown in Fig. S1. † Each DesAb was expressed and puried ( Fig. S2 †), correct folding (Fig. S3 †) and thermal stability (Fig. S4 †) were veried by circular dichroism.

Different DesAbs have distinct effects on the aggregation of asynuclein
We measured the activity of the nine DesAbs described above in a kinetic aggregation assay of a-synuclein under mildly acidic and quiescent conditions. 33 Under these conditions, a-synuclein undergoes autocatalytic nucleation on existing bril surfaces, rather than elongation of brils by incorporating monomeric protein on to bril ends. Aggregation reactions were initiated by the addition of a xed concentration of preformed brils. The formation of brils was then tracked by a uorescent readout of the amyloid-specic dye thioavin T (ThT). Performing the aggregation assay in the presence of varying concentrations of antibodies and tting all the data simultaneously (global t) enables the assessment of the effects of the antibodies on the microscopic steps in the aggregation of a-synuclein. 33,34 All DesAbs were rst tested in a high seeded assay (15% of monomeric a-synuclein concentration) to decouple the nucleation and elongation mechanisms. Under high seed conditions, the effect of primary and secondary nucleation can be neglected, thus enabling the quantication of the effects on bril elongation. Each of the DesAbs which was incubated at varying concentrations in the highly-seeded assay had an insignicant effect on the aggregation and the elongation rate, k + , as quantied by tting the initial, linearly increasing data points with a linear function (Fig. 2).
Next, we tested the DesAbs in a low seeded assay (0.06% of monomeric a-synuclein concentration) (Fig. 3) under mildly acidic conditions to induce secondary nucleation. We found that the DesAbs targeting the region of the a-synuclein sequence that is incorporated closer to the core region of brils tended to be more effective at inhibiting the aggregation reaction, in particular aS46, aS49 and aS61. The data also indicate that the scaffold (DesAb-none) and a DesAb designed against Ab oligomers (DesAb-O) have a minimal effect on the aggregation kinetics ( Fig. 2 and S5 †). We also checked that the DesAbs do not aggregate themselves (Fig. S6 †). Here, X4 and X2 refer to separate batches of DesAb-none. We also conducted transmission electron microscopy (TEM) of the end points from the low seeded aggregation reactions to analyse the morphology of the brils. Aggregates from a reaction with and without DesAbs appear similar in length and structure (Fig. S7 †), indicating no clear morphological differences, consistent with our nding that the DesAbs have minor effects on elongation. By tting the aggregation curves obtained from both low seeded and high seeded aggregation assays with the master equation framework, we decoupled the effect of the DesAbs on a-synuclein secondary nucleation from its effect on bril elongation.
Furthermore, by using the changes in the effective microscopic rate constants, the oligomer ux could be calculated. 33 This parameter describes the tendency of the system to form oligomeric species over time and is informed by both the apparent rate of secondary nucleation, which acts as a source of oligomeric species, and the apparent rate of elongation, which acts as a "sink" of oligomeric species as they elongate to form mature brils. The ux is depicted as a Gaussian function, f, for each antibody at a stoichiometry of 1 : 125 and may be used to extract three key parameters relating to the population of oligomeric species, namely, the peak height (f Peak ), area (f Area ) and peak time (f Time ) of the function. These parameters describe the maximum magnitude of the ux with which oligomeric species formation occurs (f Peak ), the total number of oligomers Error bars represent AEs.d. of three technical replicates. Data were analysed using a one-way analysis of variance (ANOVA) and all p-values are given compared to the mean of the control result (no antibody present in aggregation reaction). *p < 0.1, **p < 0.01, ***p < 0.001, ****p < 0.0001. The bars are normalized with respect to the control measurement and are displayed at the position of the target epitope along the a-synuclein sequence. The panel below the a-synuclein sequence shows residues which have been discovered to arrange in the core of a-synuclein fibrils using cryo-electron microscopy studies (black). Structures 1-2 refer to different conformations obtained from in vitro aggregation reactions, whereas MSAI and MSAII show two folds found in the brains of multiple system atrophy (MSA) patients. [47][48][49][50] Antibody-free aggregation is referred to as control. Data were analyzed using a one-way analysis of variance (ANOVA) and p-values are given compared to the mean of the control result. **p < 0.01, ***p < 0.001, ****p < 0.0001. formed over the time course of the aggregation (f Area ), and the time at which the ux towards oligomeric species formation is maximal (f Time ), respectively. Each of the parameters of the function (f Peak , f Area , and f Time ) are summarized in a bar plot (Fig. 3b). An ideal DesAb would reduce f Peak and f Area while maximally delaying f Time . Our analysis conrmed that aS46, aS49, aS61, aS68 and aS70 have the largest effect on the oligomer ux. These DesAbs are designed to bind to a region that is in the core of a-synuclein brils and thus the effect on the oligomer ux to some extent mirrors the structure of a-synuclein brils obtained from cryo-EM (Fig. 3b).
Finally, we tested aS46 in an aggregation assay involving the use of lipid vesicles to induce the aggregation of a-synuclein in a neutral pH buffer. 35 Similar stoichiometries to the seeded assays were used, and the recorded ThT traces indicate that aS46 also affects a-synuclein aggregation under these experimental conditions (Fig. S8 †).
Although we were able to identify DesAbs that potently alter secondary nucleation rates in the kinetic aggregation model, it remained challenging to trace these effects back to actual binding events. For instance, DesAbs binding to the catalytic site of a-synuclein brils may be the cause of a decrease in secondary nucleation rate. Equally, the DesAbs may also be able to bind to an oligomer's conformation in a secondary nucleation event. While the results obtained from the kinetic analysis provide a good indication of the activity of the antibodies, we sought to further assess the binding to a-synuclein monomers and aggregates in specically designed binding experiments.

DesAbs show low binding activity to monomeric and brillar a-synuclein
To determine the binding affinity to monomeric a-synuclein, we used microscale thermophoresis (MST) with uorophorelabelled a-synuclein (aSyn-N122C Alexa Fluor™ 488) at varying concentrations of DesAbs. The measurements generated K D values in the low mM range ( Fig. 4a and Table 1), which are comparable to those found for DesAbs designed to bind Ab. 27,28 To test whether a low pH buffer affected the binding affinity, the measurement was repeated at low pH with three of the DesAbs (aS28, aS46, aS68) and resulted in an approximately ten-fold increase in binding affinity towards monomeric a-synuclein. A conformational change of monomeric a-synuclein 36,37 or a change in charges of the DesAbs leading to a different binding mechanism could potentially explain these results.
The binding to a-synuclein brils was determined by biolayer interferometry (BLI) by immobilizing a-synuclein brils to AR2G biosensors to reduce the effect of avidity in the measurement. No binding signal was observed at DesAb concentrations between 100 nm and 1 mM, as compared to a strong signal obtained from a commercial monoclonal antibody (mAb) (Fig. 4b). The lack of binding signal to brils suggest that the epitopes are inaccessible to the DesAbs in a brillar conformation (at least in the BLI setup that we used here), which might also be enhanced by the disordered tails of a-synuclein covering the brils. 38 Both experiments are consistent with ndings from the kinetic aggregation assay that the DesAbs most likely interact more strongly with conformations of a-synuclein other than the monomeric or brillar.
Binding affinity measurements using BLI and MST are powerful high throughput methods for measuring the binding affinities of two isolated binding partners. The measurements work well with a homogeneous molecular target, e.g., monomeric a-synuclein or brillar a-synuclein aggregates, since both are bulk measurements. It becomes challenging, however, to quantify affinity when the target species are difficult to isolate or are present in low concentrations. Using classical bulk binding affinity characterization methods prove to be less benecial for these types of experiments. Firstly, because a-synuclein oligomers have a tendency to be hydrophobic, the aggregates may unspecically bind to the biosensor or capillary. Furthermore, in order to capture an adequate binding curve for bulk measurements, the experiment needs to use signicant amounts and concentrations. A homogeneous cover of the BLI biosensor is desired when lling the device, however this is challenging given the low amounts of a-synuclein oligomers that are formed during an in vitro aggregation reaction. Therefore, we next employed single-molecule imaging techniques to study the binding to oligomers.
Detection of a-synuclein aggregates in PD serum with uorescence microscopy We next sought to employ uorescence microscopy to characterize the binding of the DesAbs to individual a-synuclein aggregates. Seven of the DesAbs (aS13, aS28, aS35, aS46, aS49, aS61, aS68) were labelled with Alexa Fluor™ 647 by N-hydroxysuccinimide (NHS) ester chemical conjugation to primary amines aer conrming that the complementary epitopes did not contain lysine residues. The serum of PD patients was used to assess the ability of DesAbs to detect a-synuclein aggregates present in these samples.
To specically detect a-synuclein aggregates in the serum, a single-molecule pull-down (SiMPull) assay was used in combination with total internal reection uorescence (TIRF) microscopy. 39,40 This method has previously been used to study asynuclein aggregates from human brain tissue using an a-synuclein mAb. 40 Here, we used a commercial anti-a-synuclein antibody raised against residue 121-125 of monomeric a-synuclein (named SC in the following) acting as capturing antibody immobilized on the coverslip. PD serum samples of three different patients were mixed (1 : 1) for this experiment to balance out a potential variability between patient samples. Aer incubation with the PD serum and several blocking and washing steps, a-synuclein aggregates were detected to different extents with the uorophore-labelled DesAbs (Fig. 5). The results are shown in comparison to their corresponding blanks (PBS used instead of PD serum) and the uorophore labelled SC antibody ( Fig. 5b and c). Data for the four different conditions (serum, brils, PBS and blank) were obtained by averaging over the intensity of 50 stack images at one eld of view (Fig. 5d). Some of the DesAbs showed an increased binding signal compared to their corresponding blank, with aS28, aS46, aS49 and aS68 exhibiting high signicance (p < 0.0001). The blanks showed essentially no localizations, in contrast to the sample slide detected with either aS46 or SC, which showed remarkably similar distributions of localizations while only the intensity of spots differs. According to how many lysines are available for labeling and therefore the difference in labeling degree between those two types of antibodies, explains the difference in intensity.
All DesAbs showing an increased binding signal were then imaged in an experimental setup where a-synuclein brils were captured with SC. Conrming the results obtained using BLI, the DesAbs did not show signicantly more binding to a-synuclein brils (Fig. S9 †) or monomers (Fig. S10 †) as compared to the PBS control. SC, however, showed signicantly higher amounts of localizations in the sample slide (Fig. S9 †) since the SC also binds brils. These ndings conrm that the DesAbs do not bind brils at the concentrations tested and further, that the a-synuclein aggregates present in PD serum are not similar in conformation to in vitro a-synuclein brils.
To investigate the composition of serum samples and test the specicity of aS46 for a-synuclein aggregates, further experiments with control samples were conducted. We used two types of control samples, a serum mimic composed of bovine serum albumin (BSA) at a concentration similar to that of human serum albumin in human serum (50 mg ml −1 ), 41 and in vitro Ab brils. In both instances, the number of spots per eld of view that could be detected was far less compared to the number found in the serum sample ( Fig. 5d and e). We also tested whether aS46 shows an increased signal if an antibody specic to Ab (6E10) was used as capture antibody. Here, fewer spots were observed, but they were nonetheless more numerous  than the number of spots found in the PBS sample, while SC detected a level that was comparable to PBS. This nding could be explained by the occurrence of co-aggregates formed from Ab and a-synuclein monomers, 42 which would have a different epitope presented than the one SC is targeting. Another possibility is that aS46 binds to a higher extent of non-specically to certain other species found in human serum. By using a biotinylated version of aS46 as capture antibody, we also tested the extent to which aS46 is able to pull down aggregates from serum and using uorescently labelled aS46 as detection antibody, we observed an increased number of detected aggregates (Fig. S11b †). We also estimated the binding affinity towards aggregates in serum (K D ¼ 41 nM; R 2 ¼ 0.96) by changing the concentration of the detection antibody and tting a binding curve to the number of detected aggregates for each concentration used (Fig. S11a †). Finally, we sought to test whether our ndings could be applied to other biouids. We thus analyzed CSF samples pooled from three PD patients, where we detected signicantly more a-synuclein aggregates in a SiMPull set-up when using SC as a capture and aS46 as a detection antibody as compared to the PBS blank (Fig. S12 †). Notably, the level of a-synuclein aggregates found in PD CSF is about 14-fold lower on average than in PD serum (number of detected spots per eld of view: N PD_CSF ¼ 27 AE 16 vs. N PD_Serum ¼ 370 AE 130), which is consistent with previously reported results. 43 Detection of a-synuclein aggregates in PD serum with dSTORM microscopy The uorescence spectroscopy method described above could not go beyond the diffraction limit, so we could not use it to establish the size of the a-synuclein aggregates in the PD serum. Therefore, we next analyzed with the DesAbs the size and morphology of the a-synuclein aggregates by direct stochastic optical reconstruction microscopy (dSTORM) imaging, a superresolution imaging technique. The imaging buffer was changed to a STORM buffer (GLOX buffer, PBS-Tris (50 mM), glucose (0.5 mM), glucose oxidase (1.3 mM), catalase (2.2 mM), and mercaptoethylamine (MEA) (50 mM)), and samples were imaged at a higher laser power and additionally with an activation laser at 405 nm. Thus, the characteristic blinking of uorophores was achieved, and eld of views were imaged for up to 6 min at 33 frames per second. Images were reconstructed using a customwritten Python code accessing the Thunderstorm module in ImageJ. Fig. 6 depicts representative diffraction-limited and reconstructed super-resolved images for serum samples imaged with SC and aS46. We used aS46 since throughout previous characterization it consistently showed strong binding characteristics. Analyzing the length of the a-synuclein aggregates for both conditions revealed a comparable size distribution for both conditions with most a-synuclein aggregates appearing with a length smaller than 200 nm (aS46 mean length ¼ 100 AE 80 nm; SC: mean length ¼ 120 AE 90 nm, p < 0.0001 (unpaired ttest)) (Fig. 6).

Conclusions
In this work, we have reported designed antibodies (DesAbs) for the characterization of the conformational properties of soluble aggregates of a-synuclein. The approach is based on the computational design of a panel of antibodies that target epitopes that scan the sequence of a-synuclein, and on their characterization in terms of their ability to bind a-synuclein oligomers. Our results indicate that this approach can lead to the generation of antibodies capable of characterizing heterogeneous populations of soluble a-synuclein aggregates in human serum. We were further able to use one of the designed antibodies in a super-resolution imaging setting and detected asynuclein aggregates both in serum and CSF from PD patients. Our approach thus offers a route for the development of diagnostic tools for PD and related synucleinopathies.

In silico antibody design
The computational method used to identify complementary peptides was described previously. 25,44 Briey, complementary peptides are obtained from a library of b-sheet fragments from the protein data bank (PDB). To design a complementary peptide fragments are merged from the PDB library that are found to face segments of the target epitope in a b-strand conformation. This method usually generates more than one complementary peptide for a given epitope. To select the complementary peptide with the best predicted affinity and specicity, DesAbs are ranked based on the following properties of the different fragments: (1) the number of times a pair of fragments is found in native proteins, (2) the length of the fragment, and (3) how many times a fragment faces other peptides which are not in the epitope.

Protein expression and purication
Plasmids with the corresponding epitope graed onto the DesAbs scaffold's CDR3 loop were bought commercially (Genscript). In E. coli SHuffle T7 competent cells from New England BioLabs, the various antibody plasmids were produced and puried with the addition of ampicillin (100 g ml −1 ). Using Overnight Express Instant TB Medium (Merck Millipore) treated with ampicillin (100 g ml −1 ), cells were cultured overnight for 15 hours at 29 C and 200 rpm. Cells were collected by centrifugation aer an additional hour of incubation at 37 C, and aer being resuspended in 40 ml of TRIS solution that had 10 mM imidazole added, they were sonicated to lyse. Using centrifugation at 18 000 rpm, the supernatant containing the protein was isolated from cell debris (JA-20 rotor, Beckman Coulter). A Ni-NTA Superow column (Qiagen) that had previously been equilibrated with PBS containing 10 mM imidazole was loaded with the cleared lysate. DesAb that had been his-tagged was eluted in PBS that contained 200 mM imidazole. Finally, employing size exclusion chromatography on a HiLoad 16/600 Superdex 75 column, the protein-containing fractions were combined and further puried (GE Healthcare Life Sciences). By measuring the absorbance at 280 nm and utilizing theoretical extinction coefficients generated by ExPASy ProtParam, the nal concentration was computed. Aliquots were kept at −80 C aer being frozen in liquid nitrogen.

Circular dichroism
A Jasco J-810 spectropolarimeter was used to measure the farultraviolet UV CD spectra of several DesAbs using a cuvette with a 0.1 cm-pathlength. At a protein concentration of 10 M, spectra were taken between 200 and 250 nm, at 20 C, and a baseline of the buffer was systematically subtracted from each spectrum. At a xed CD wavelength of 207 nm, the temperature denaturation of each DesAb was monitored from 20 to 95 C at a rate of 0.5 C min −1 . The temperature probe was placed inside the cuvette to measure the temperature. The fraction of unfolded protein was used to standardize the measured ellipticity values.

Preparation of a-synuclein seeds
Monomeric a-synuclein was buffer-exchanged to MES buffer (10 mM 2-(N-morpholino)ethanesulfonic acid, 1 mM EDTA, pH 5.5) and concentrated to 200-300 mM using 10 000 kDa MWCO centrifugal spin lters (Amicon Ultra, Millipore). Concentrated a-synuclein monomers were incubated in protein low-binding tubes (Eppendorf) for 72 hours at 40 C and 400 rpm with a magnetic stirrer. In a benchtop centrifuge, the solution was centrifuged at 21 130 Â g to determine the bril concentration (monomeric equivalent) (Eppendorf). By measuring absorbance with a NanoDrop 2000 (Thermo Scientic), the concentration of the remaining -synuclein monomer detected in the supernatant was calculated and subtracted from the initial concentration. Aer MES buffer was added to the supernatant's volume, the stock was aliquoted and kept at −80 C. The bril stock was diluted to a nal concentration of 5 mM in protein low binding tubes and sonicated for 15 seconds to create -synuclein seeds.

Lipid vesicle preparation
2,3-Dimercapto-1-propanesulfonic acid (DMPS) powder was dissolved to a nal concentration of 5 mM in 20 mM sodium phosphate buffer at pH 6.5 and stirred at 900 rpm at 50 C for 2-2.5 h followed by ve freeze-thaw cycles in dry ice and a hot water bath at 45 C to ensure unilamerallity. To form vesicles, the solution was sonicated by a Sonopuls HD 2070 probe sonicator (Bandelin) for 3 Â 5 min cycles on a 50% cycle at 10% maximum power, and centrifuged at 15 000 rpm for 30 min at 25 C to remove residue formed during sonication.

Plate reader protocol
A Superdex 75 10/300 GL column (GE Healthcare) was used to gel lter puried a-synuclein while it was equilibrated in MES buffer (10 mM 2-(N-morpholino)ethanesulfonic acid, 1 mM EDTA, pH 5.5). The peak corresponding to monomeric a-synuclein peptide was then collected in a low-binding test tube (Corning) on ice. Monomeric a-synuclein peptides were used to prepare solutions at a protein concentration of 40 mM in the presence of increasing amounts of the specied DesAb variant in MES buffer supplemented with 5 mM ThT and 0.06% a-synuclein seeds. Then, 150 ml of sample was pipetted into each well of a 96-well half-area plate of black polystyrene with a clear bottom and polyethylene glycol coating (Corning). Evaporation was prevented by sealing the plates. Utilizing a CLARIOstar plate reader (BMG Labtech), aggregation experiments were carried out at 37 C in quiescent conditions. With a 440 nm excitation lter, the ThT uorescence was monitored through the plate's bottom at 480 nm every minute.
In the case of vesicle induced aggregation, the measurement was taken with monomeric a-synuclein at a concentration of 20 mM in the presence of increasing amounts of aS46 in MES buffer (10 mM 2-(N-morpholino)ethanesulfonic acid, 1 mM EDTA, pH 6.5) supplemented with 5 mM ThT and DMPS at 100 mM at a temperature of 30 C as previously described. 35

Determination of rate constants
In experimental conditions under which a-synuclein bril elongation is favoured, and where primary and secondary nucleation events are negligible, the initial rate of bril elongation mat be approximated with a linear function: where M(t) is the bril mass concentration, P(0) is the initial number concentration of brils, m(0) is the initial monomer concentration and k + is the rate of bril elongation. Since P(0) and m(0) are universal for each condition tested, the relative rate of elongation for each DesAb may be found by normalising the value of 2k + P(0)m(0), that is, the slope of the initial linear points of the elongation reaction, of a given DesAb to that of the control condition. Under conditions where a-synuclein secondary nucleation is favoured, the aggregation may be described by tting a generalised logistic function to the normalised aggregation data: where m tot refers to the total a-synuclein monomer concentration present within a reaction. The terms a, k and c are tting parameters with and with c xed at a value of 0.3. This corresponds to a reaction order of n 2 ¼ 4, and represents secondary nucleation behaviour as observed in the case of the islet amyloid polypeptide precursor (IAPP) protein. 45 The terms l and k represent combinations of primary and secondary nucleation rate constants, respectively. 45 Fitting the generalised logistic function (eqn (2)) to normalised secondary nucleation data provides an analytical approximation for the monomer concentration at time t, m(t), since This, in turn, facilitates the approximation of the bril number concentration, P, over time by rearranging eqn (1) in terms of the rst derivative of P(0), which corresponds to the change in bril number over time, dPðtÞ dt : This is solved at the half-time of each reaction, where it is assumed to me maximal, to yield dPðtÞ dt ; obtained at the half-time of the reaction, is used as an effective rate of bril amplication and is reported for each DesAb relative to the control condition.

Determination of the oligomer ux
Data were tted as previously described by Staats et al. 33 The ux towards oligomeric species over time, f(t), was obtained analytically by means of the following equation derived from the linear polymerisation model (eqn (1)): where M(t) and m(t) are represented analytically in terms of the t of the generalised logistic function (eqn (2)) to secondary nucleation data, while r + denotes the effective elongation rate obtained for each DesAb as approximated by eqn (1).
To approximate the parameters of the ux towards oligomeric species, eqn (2) may be substituted into eqn (7) to yield an expression for f(t) in terms of the generalised logistic function: from which the area under the curve of the ux towards oligomeric species, f Area , may be calculated by calculating the integral of eqn (8): Eqn (8) may also be used to approximate the peak time and peak height of f(t), since the peak time may be calculated as a solution to df dt t¼t Peak which yields a description for the peak height of f as The extracted values for f Peak , f Area , and f Time of each condition, i, were then normalised to the corresponding mean value for three replicates of a-synuclein only (control): Normalised

BLI measurements
Amine reactive second generation (AR2G) biosensors from FortéBio were used for the measurements on the BLI platform Octet K2. 200 ml were used for each well's sample volume. Biosensors were soaked in PBS for at least 10 minutes in a 96well plate prior to each cycle. Initially, the pH of the buffer and the concentration of -synuclein brils were changed to provide the best immobilization conditions for -synuclein aggregates. 5 mM a-synuclein bril concentration was determined to be the ideal condition (pH 7.4). PBS was utilized as an equilibration buffer and baseline buffer. Biosensors were activated using a freshly prepared EDC/sNHS mix (20 mM EDC and 10 mM sulfo-NHS in 10 mM acetate buffer, pH 5) before the plate was set up. Tips were quenched in 1 M etanolamine aer being loaded with -synuclein brils (pH 8.5). A baseline was recorded before each sample's association was evaluated. One tip served as a reference by having PBS in the matching well instead of immobilized -synuclein to control for non-specic antibody binding to the biosensor. Reference curves were subtracted from sample curves and plotted using a customwritten code in order to examine binding of each DesAb (MATLAB).  The supernatant (serum) was aliquoted and stored at −80 C until use. Lumbar puncture was performed under sterile technique with 1% lignocaine local anesthetic (Table 1). 5 ml of CSF was collected and centrifuged for 10 min at 300 G at 4 C before samples were aliquoted and stored at −80 C.

Serum and CSF samples
Before imaging a 1 : 1 dilution of serum or CSF samples as listed in Table 1 was prepared.

Preparation of in vitro a-synuclein brils for SiMPull
Monomeric recombinant a-synuclein was incubated at a concentration of 200 mM in PBS for 72 hours at 37 C with constant agitation of 200 rpm in an orbital shaker (Eppendorf). Before imaging brils were diluted to a nal concentration of 500 nM monomer equivalent.

Preparation of in vitro Ab for SiMPull
Monomeric recombinant Ab (Abeta42, rPeptide) was incubated at a concentration of 4 mM in PBS for 8 hours at 37 C. Before imaging brils were diluted to a nal concentration of 500 nM monomer equivalent.

Fluorescence microscopy
Sample preparation for uorescence microscopy. All TIRF experiments were conducted using covalently PEGylated glass coverslips. Coverslips were prepared following a previously published protocol, 46 with minor modications. The glass cover slips (26 Â 76 mm, thickness 0.15 mm, Thermo) were cleaned using an ultrasonic cleaner (USC100T, VWR) in a series of solvents for 10 minutes each: 18.2 MU cm MQ water, acetone, and MeOH. Aer being washed, the cover slips were etched with 1 M KOH under 20 minutes of ultrasonication, rinsed with MeOH, 18.2 M MU cm MQ water, and dried with nitrogen ow before being cleaned with argon plasma for 15 minutes (Femto Plasma Cleaner; Diener Electronic). Clean coverslips were salinized for 20 min with 60 s of ultrasonication at the beginning and middle of the process using 5 ml of 3-aminopropyl triethoxysilane (Fisher Scientic UK, cat. no. 10677502), 8.3 ml AcOH in 166.7 ml MeOH (10 min aer the start point). Following a rinse in MeOH, 18.2 MU cm MQ water, and MeOH, the salinized coverslips were dried with nitrogen ow. The coverslips were then cleaned and salinized, and 50-well PDMS gaskets (Sigma, GBL103250-10EA) were aerwards applied. A 100 : 1 aqueous combination of MPEG-SVA-5000 (110 mg ml; MW 5000; Laysan Bio Inc.) and biotin-PEG-SVA-5000 (1.1 mg ml; MW 5000; Laysan Bio Inc.) was added to the wells to passivate them, along with 1 ml of 1 M NaHCO 3 (pH 8.5). In a humid chamber, the coverslips were incubated with PEG solution for an entire night. Aer being rinsed with 18.2 MU cm MQ water, they were dried with nitrogen ow. In order to further treat the passivated wells, 9 ml of MS(PEG)4 methyl-PEG-NHS-ester (10 mg ml −1 , Thermo Fisher, cat. no. 22341) and an additional 1 ml of 1 M NaHCO 3 (pH 8.5) were added. The coverslips were placed in a humid chamber for an overnight PEG solution incubation, rinsed with 18.2 MU cm MQ water, and then dried with nitrogen ow. Glass coverslips that had been PEGylated were kept in a desiccator at −20 C until needed.
Before slide preparation, the following three buffers for washing, dilution and imaging were prepared (all at pH 7.4): 1 Â PBS (PBS), 1 Â PBS + 0.05% Tween 20 (PBS-T), 1 Â PBS + 0.05% BSA + 0.0.5% Tween 20 + 10% salmon sperm (PBS-B). 10 ml of NeutrAvidin (0.2 mg ml −1 ) was added to each well and was le to incubate for 10 min. Each well was then washed with PBS-T three times before adding 10 ml of the biotinylated capture antibody sc-12767 (SC, Santa Cruz) at 12.1 nM for 15 min. With PBS-T, the extra antibody was removed three times. The target sample was then added to the well, such as a 500 nM bril sample or an undiluted serum sample, and allowed to incubate for 60 min. PBS was placed into blank wells for the same amount of time. Following the incubation, samples were taken out of the wells and each well received 10 ml PBS-B for a 60 minute blocking step, followed by three PBS-B-W washes. The detection antibody was added at a dilution of 2 nM (DesAb) or 5 nM (mAb) and incubated for 30 min (DesAbs) or 10 min (mAb) (mAb). The wells were then lled with 3 ml of PBS and sealed with a plasma-cleaned cover slip aer being washed with PBST ve times in total.
TIRF imaging. Imaging was carried out using a home-built total internal reection uorescence (TIRF) microscope consisting of a Ti-E Eclipse inverted microscope (Nikon) with an oilimmersion objective (UPlanSApo, 100Â, 1.49 NA; Olympus) and a perfect focus system. AF647 was excited with a 638 nm laser (iBeam-Smart, Toptica) and coupled into the sample by a dichroic mirror (Di01-R405/488/561/635, Semrock). Emission was collected by the objective lens and separated from excitation light using a dichroic mirror (Di01-R405/488/561/635; Semrock) and passed through appropriate emission lters (FF01-692/40-25; Semrock). The emitted uorescence was then focused onto an air-cooled EMCCD camera (Photometrics Evolve, EVO-512-M-FW-16-AC-110). Image stacks of 50 frames were acquired with an exposure time of 50 ms. To minimize any bias associated with the ROI selection, an automated script (Micro-Manager) was used to collect images in a grid. Images were analyzed using a custom-written ImageJ script analyzing the maxima in each image.
dSTORM imaging. For single molecular super-resolution imaging, the sample slides were prepared as described above and aer diffraction-limited imaging was completed the imaging solution was removed. Three additional sheets of 50well PDMS gaskets (Sigma, GBL103250-10EA) were then connected to the rst sheet in order to retain a larger volume in each well. Then, mercaptoethylamine (MEA) (50 mM) was pipetted into each well along with PBS-Tris (50 mM), glucose (0.5 mM), glucose oxidase (1.3 mM), catalase (2.2 mM), and dSTORM buffer (50 mM). Before imaging, MEA was immediately added to the buffer. By stimulating the sample with the 638 nm laser at 150 mW laser power while also running the 405 nm laser at 20 mW power, image stacks of 8000 frames were captured at 33 frames per second with an exposure duration of 30 ms. The resulting image had 105.4 nm for each pixel. Superresolution images were reconstructed using a custom-written Python script with the Fiji package Thunderstorm.