Blog

By Derya Emin, Yu P. Zhang, Evgeniia Lobanova, Alyssa Miller, Xuecong Li, Zengjie Xia, Helen Dakin, Dimitrios I. Sideris, Jeff Y. L. Lam, Rohan T. Ranasinghe, Antonina Kouli, Yanyan Zhao, Suman De, Tuomas P. J. Knowles, Michele Vendruscolo, Francesco S. Ruggeri, Franklin I. Aigbirhio, Caroline H. Williams-Gray, and David Klenerman

Excerpt from the article published in Nature Communications 13, 5512, 20 September 2022. DOI: https://doi.org/10.1038/s41467-022-33252-6

Editor’s Highlights

• Since α-synuclein aggregates are highly heterogeneous, it is difficult to determine how the physical characteristics such as size and morphology determine their neurotoxic properties. 
• Oligomers only account for a small fraction of the total intercellular α-synuclein but are of great interest as they are known to be involved in several different pathways leading to cell death and neurodegeneration. 
• Some of these mechanisms include disruption of the cell membrane, synaptic loss, dysfunctions of the endoplasmic reticulum and mitochondria, seeding capacity, and neuroinflammation. 
• The small soluble non-fibrillar α-synuclein aggregates smaller than 200 nm in length are highly toxic species causing lipid membrane permeabilization and inflammation.
• These small soluble aggregates are the critical species driving toxicity in the PD brain.

Abstract

Soluble α-synuclein aggregates varying in size, structure, and morphology have been closely linked to neuronal death in Parkinson’s disease. However, the heterogeneity of different co-existing aggregate species makes it hard to isolate and study their individual toxic properties. Here, we show a reliable non-perturbative method to separate a heterogeneous mixture of protein aggregates by size. We find that aggregates of wild-type α-synuclein smaller than 200 nm in length, formed during an in vitro aggregation reaction, cause inflammation and permeabilization of single-liposome membranes and that larger aggregates are less toxic. Studying soluble aggregates extracted from post-mortem human brains also reveals that these aggregates are similar in size and structure to the smaller aggregates formed in aggregation reactions in the test tube. Furthermore, we find that the soluble aggregates present in Parkinson’s disease brains are smaller, largely less than 100 nm, and more inflammatory compared to the larger aggregates present in control brains. This study suggests that the small non-fibrillar α-synuclein aggregates are the critical species driving neuroinflammation and disease progression.

Introduction

Parkinson’s disease (PD) is the second most common neurodegenerative disorder after Alzheimer’s disease, affecting up to 3% of the population over the age of 65¹. PD is initially characterized by the irreversible damage to dopaminergic neurons in the substantia nigra that correlates with motor symptoms including bradykinesia, rigidity, and tremor; but more widespread pathology in limbic and cortical regions evolves over the disease course and is linked to a range of non-motor features^2,3,4. Post-mortem brain analysis has shown the presence of proteinaceous inclusions, often referred to as Lewy bodies, which primarily contain aggregated α-synuclein⁵. Although, a recent paper identified disintegrated membranes and organelles as the main constituents of Lewy bodies rather than α-synuclein⁶. Lewy bodies are found within the substantia nigra as well as in other subcortical and cortical regions to a variable extent⁷. Their presence in limbic and cortical regions correlates with the development of PD-associated dementia⁸. During an aggregation reaction, α-synuclein monomers convert intracellularly into various sizes, structures, and morphologies over time^9,10,11,12. All these different α-synuclein species coexist, making it challenging to identify the toxic species in complex human biofluids and tissues¹³. Previous studies have shown the small soluble prefibrillar aggregates of α-synuclein, often referred to as oligomers, rather than mature fibrils, are the most cytotoxic aggregate species in vitro^11,14,15,16. There is no rigorous definition of oligomers but can include any soluble aggregate of an intermediate size between monomers and insoluble fibrils¹⁷.

Oligomers only account for a small fraction of the total intercellular α-synuclein but are of great interest as they are known to be involved in several different pathways leading to cell death and neurodegeneration. Some of these mechanisms include disruption of the cell membrane^14,18,19, synaptic loss^20,21,22,23, dysfunctions of the endoplasmic reticulum and mitochondria^24,25, seeding capacity^14,26,27,28, and neuroinflammation^22,29,30,31. However, other work has shown that while oligomers cause cell death to dopaminergic neurons, fibrils, which are capable of seeding aggregates, are more toxic and lead to sustained progressive loss of neurons^32,33. Many of these experiments are performed using acute doses of aggregates so it is unclear which mechanisms of toxicity will be most important at physiological concentrations of aggregates where the distribution of aggregate sizes might differ. Therefore, despite various efforts, key questions regarding how toxicity changes with size and which distinct species form in humans remain unanswered. So far, the generation and stabilization of toxic in vitro oligomers has been challenging. Previously published work has either made use of trapped oligomers^16,34, the addition of stabilizers or cross-linkers^{14,35,36,37,38,39}, high concentrations or mutated versions of α-synuclein^11,32. Despite these major advances, a crucial unanswered question is whether these lab-generated oligomers have different properties to the actual aggregates found in the brain. In this work, we have used a centrifugation-based method to isolate different species of α-synuclein. A whole range of differently structured aggregates are extracted and can be compared to each other. This method is minimally perturbative and does not need the addition of external reagents or the use of mutants, enabling the study of toxicity as a function of size. Previous publications on tau and Aβ sucrose gradient fractionation have given some insight into the structure of the potentially toxic oligomers^40,41.

While many previous studies characterizing α-synuclein aggregates by size have been conducted in vitro, more research remains to be done on identifying the soluble aggregates present in the human brain. Endogenous levels of α-synuclein aggregates were measured in the substantia nigra in control and PD post-mortem tissue by Je et al. at the single-molecule level⁴². They reported that the oligomeric species accounted for 56% of the α-synuclein species in the PD samples but only 23% in non-PD control samples. The next step is the study of the toxicity of soluble aggregates from PD brain tissue.

In this study, we focus on identifying the α-synuclein aggregate-induced toxicity depending on aggregate size and morphology. Therefore, we use a density centrifugation method to separate an aggregation mixture by size and characterize the structure of differently sized aggregates using high-resolution imaging techniques. Later, we follow their toxic properties using two well-established assays: one based on the ability of the aggregates to disrupt cell membranes, and the second based on the aggregate-induced inflammatory response of BV2 microglial-like cells. We also compare these fractionated aggregates to extracted soluble aggregates from human post-mortem PD and control brains, in terms of size and toxicity, to determine the aggregate characteristics that are most closely linked to the disease state.

Results

Structural characterization of α-synuclein fractions

Producing a distinct aggregate species with specific size and morphology has been challenging due to the heterogeneity inherent in the aggregation process. To generate a more homogeneous aggregate population, we utilized a sucrose gradient approach, which is based on the principle that proteins can be separated by mass by accumulating in different sucrose concentrations during centrifugation. Therefore, we mixed α-synuclein from different time points taken from an aggregation reaction at 37 ˚C during constant shaking and loaded this mixture onto a discontinuous gradient with 10, 20, 30, 40 and 50% sucrose concentrations (Fig. 1A). Using transmission electron microscopy (TEM), we identified that most of the small structured α-synuclein aggregates could be found in 20% fraction, with predominantly spherical species present, which started to get more elongated in the 30% fraction. The 40 and 50% fractions consisted mainly of fibrillar species with varying lengths. The abundance of aggregates in the 10% fractions was very low, pointing out that this fraction might largely consist of α-synuclein monomer. (Fig. 1B).

α-synuclein aggregates could be detected in the 20–50% sucrose fractions using total internal reflection microscopy (TIRFM) with the amyloid-specific fluorescent dye thioflavin T (ThT) (Fig. 2). Aggregates in the 10% fraction could not be detected using ThT, pointing to the fact that there is minimal β-sheet containing species in this fraction (Supplementary Fig. 1). This method allowed us to determine and compare the fluorescence intensity of each aggregate (Fig. 2 and Supplementary Table 1). The data show a gradual increase in the median intensity from low to high sucrose concentrations (Fig. 2A–D) accompanied by an apparent shift in size as the small elliptical species predominantly present in the 20 and 30% fractions become larger and turn into elongated fibrils detectable in the 40% and 50% fraction. Using the Kolmogorov-Smirnov test, the p-values show that the corresponding intensity from each fraction is significantly different from the previous one confirming an increase in size.

Length distribution of α-synuclein sucrose fractions

We used the fluorescent dye thioflavin X (ThX), a derivative of the widely used dye thioflavin T with enhanced photophysical properties, to study small structural details through transient amyloid binding⁴³. By taking advantage of its transient nature, we could perform single-molecule super-resolution microscopy of α-synuclein aggregates with a precision of 20 nm⁴⁴. Similarly to ThT, the 10% fraction could not be super-resolved using ThX due the minimal presence of β-sheet aggregates. We observe a clear trend indicating a gradual increase in length (Fig. 3A, B). The 20% fraction has an average length of 190 ± 30 nm, the 30% fraction of 240 ± 30 nm, the 40% of 290 ± 30 nm and the 50% of 390 ± 90 nm. The differences in the length distribution are statistically significant. The super-resolution representative images (Fig. 3D–G) also corroborate a clear size change.

Morphology is another parameter that can be analysed to understand aggregate structure. Using our super-resolution imaging, we investigated the eccentricity of the aggregates in the different fractions (Fig. 3C). The eccentricity of a cluster was determined by fitting an ellipse to the cluster and determining the focal distance of the ellipse divided by the maximum distance of the major axis. Values close to 1 indicate a fibrillar morphology, whereas a perfect circle has a value of 0. The values of the eccentricity of aggregates is typically between 0.3 and 1. We observed a change in eccentricity in the different sucrose fractions, although it was not statistically significant, the aggregates became more fibrillar in structure with an increasing percentage of sucrose. The 40% and 50% fractions had higher eccentricities, reaching 0.93 in the 50% fraction, while the 20% and 30% fractions had eccentricities around 0.84.

Analysis of structural changes using AFM and antibodies

In order to analyze and unravel the 3D structure of the aggregates present in each sucrose fraction, we used high-resolution and phase-controlled AFM imaging, which is able to resolve structures with angstroms resolution including monomers (Fig. 4A–E)^45,46,47. Samples for AFM analysis were deposited on a negative mica surface which is ideal to study soluble aggregates and oligomers. However, mica has less affinity to larger aggregates although some larger aggregates are visible (Supplementary Fig. 2)^47,48,49. We found a similar trend to a previously published report characterizing Aβ with sucrose gradient⁴⁰. There is a net increase in the height and the cross-sectional diameter of the aggregates present in each fraction, with the 50% fraction containing the largest aggregates (Fig. 4F, G). Compared to the higher density fractions, the 10 and 20% fractions have predominantly monomeric protein (up to 0.5 nm in height)⁴⁷. The 20 and 30% both showed the presence of protofilaments with a height of 0.6–0.8 nm. In the 40 and 50% fractions, larger mature amyloid aggregates with a height of up to 6 nm were identified (Supplementary Fig. 2). While these data were in qualitative agreement with the super-resolution imaging of aggregates, it also showed that AFM imaging was more sensitive towards smaller aggregates and monomers present in the fractions that were less accessible to ThX super-resolution imaging.

Antibody-based imaging is a powerful tool to study the surface accessibility of specific aggregates. Given that the super-resolution imaging showed an increase in length from the 20 to 50% sucrose fractions to gain further structural insight, we characterized aggregate structure with two different α-synuclein antibodies (Fig. 5). We used, MJF, a conformation-specific antibody recognizing aggregated/filamentous species, and SC, a sequence-specific antibody targeting amino acids 121–125, using methods adapted from a previously published study⁴². The number of α-synuclein aggregates detected with the SC antibody, which targets a C-terminus epitope, was greater in the lower sucrose fractions and smallest in the 40-50% fractions, indicating that the C-terminus is more accessible in the smaller aggregates and enclosed in fibrils, which is in agreement with previously published NMR data⁵⁰. The MJF antibody did not detect any aggregated α-synuclein in the 10% fraction, confirming a lack of β-sheet-rich structures in this fraction. A modest number of aggregates was instead detected using this antibody in the 20–40% fractions, with maximum quantity in the 50% fraction, which consists mainly of fibrils based on the SR data. These results are in agreement with our previous data (Figs. 2 and 3). To confirm that the number of detected species is due to the specific binding of the detection antibody to the aggregate, we compared the binding of an isotype antibody control (IgG) to the fractions (Supplementary Fig. 3). The number of detected spots decreased for the case of the labelled isotype control by a factor of ~20 for SC and about 30 for MJF, confirming that the signal is due to specific antibody-antigen interactions.

The cytotoxic properties of aggregates correlates with their size

To understand how the sizes of the individual aggregates influence their cytotoxic properties, each sucrose fraction was diluted to a total monomer concentration of 500 nM and added to BV2 mouse microglia cells for 24 h. Previous work shows that α-synuclein aggregates are stable once formed, even upon dilution to picomolar concentrations⁹, and we have previously found no change in the number of aggregates under these conditions^29,51. Afterwards, the release of tumour necrosis factor α (TNFα), a pro-inflammatory cytokine, strongly implicated in the progression of PD⁵², was quantified after 24 h upon incubation with the fractions (Fig. 6A, C). The TNFα response was greatest upon stimulation with the 20% fraction. The 20% fraction accounts for sub-diffraction small species with an average length of 190 nm as determined by super-resolution microscopy (Fig. 3A). The larger fibrillar species present in the 40 and 50% fractions caused the lowest neuroinflammatory response, while the species in the 30% fraction produced an intermediate response. The aggregates in the 10% fraction are smaller than the species in the 20% according to the AFM analysis (Fig. 4F, G) but produced a similar inflammatory response to the 20% fraction. Hence, overall the data suggest that smaller non-fibrillar aggregates are more inflammatory than larger fibrillar aggregates. The fractions were then added to liposomes and their ability to disrupt lipid membranes and cause calcium influx was assessed (Fig. 6B, D). In this experiment, the aggregates present in the 10 and 20% fractions caused more membrane disruption than the larger species present in the 30–50% fractions. Overall, these results support the hypothesis that small soluble aggregates rather than fibrils cause greater toxicity.

Smaller aggregates of α-synuclein are found in PD brains and are more inflammatory

Soluble aggregates were extracted from post-mortem human brain tissue according to a previously published protocol (Supplementary Fig. 4)⁵³. Briefly, brain pieces were cut to 300 mg and incubated in the buffer for 1.5 h. Afterwards, the upper 90% was centrifuged. Following centrifugation, the remaining upper 90% was centrifuged again. In the last step, the extracted aggregates were dialysed for 72 h in order to remove any pro-inflammatory cytokines. We focused on the amygdala as the region of interest as recent clinicopathological work has demonstrated that both α-synuclein pathology and neuroinflammation in this region are correlated and associated with the development of PD dementia⁸. We compared soluble protein extracts from the amygdala from three PD cases and three similarly aged controls without neurological disease during life (Table 1). Immunohistochemically analysis of formalin-fixed sections confirmed Lewy body pathology in the amygdala in all three PD cases and none of the controls (Supplementary Fig. 5). The aggregate length was determined by aptamer paint (AD-PAINT) (Fig. 7), a super-resolution method that uses an aptamer for aggregate recognition, which can detect both α-synuclein and Aβ aggregates⁵⁴. Most of the aggregates were smaller than 100 nm, with control samples having a higher proportion of larger aggregates (Fig. 7A–D). The length distribution of the aggregates differed between PD and control brain extracts with a median aggregate length of 55 nm for PD and 68 nm for controls (Fig. 7E). Length distributions plotted for individual brain samples are shown in Supplementary Fig. 6. We plotted the relative differences between PD and controls for the normalized and cumulative histograms of the length distributions (Fig. 7B, D). Negative values indicate that more species of that particular length are present in the control group. There were about 3% more aggregates smaller than 50 nm in the PD samples, whereas aggregates larger than 100 nm in length are 1% more abundant in the controls (Fig. 7B). According to the cumulative length distributions (Fig. 7D), there was the biggest difference (~13%) in the number of aggregates smaller than 74 nm in PD versus controls. Representative super-resolution images are shown for PD (Fig. 7G) and controls (Fig. 7H). The PD and control aggregates had the same eccentricity of 0.78 indicating that they are predominantly non-fibrillar (Fig. 7D). Overall, these data show that smaller non-fibrillar aggregates <100 nm are present in the PD brain samples over the control samples.

Diagnosis	Sex	Age (years)	Disease duration (years)	Postmortem interval (hours)	Cause of death
Control	M	72	–	55	Myocardial infarction
Control	M	70	–	12	Carcinoma of the oesophagus
Control	M	87	–	61	Ruptured aortic aneurysm
Parkinson’s disease	M	86	18	6	Parkinson’s disease
Parkinson’s disease	M	64	19	108	Congestive cardiac failure
Parkinson’s disease	M	75	12	51	Parkinson’s disease

To independently confirm these results, we performed high-resolution AFM phase-controlled imaging on the smaller soluble aggregates from the brain samples (Supplementary Fig. 7). In this case, due to the increased complexity of the samples it was not possible to resolve monomers, so the AFM and super-resolution data are in closer agreement. The results confirm that PD soluble aggregates had slightly smaller median height (PD 4.2 ± 0.05 nm versus control 4.4 ± 0.05 nm, p < 0.001) and smaller cross-sectional diameter than control ones (PD 20 ± 1 nm versus control 22 ± 1 nm, p < 0.001). Consequently, the AFM data confirmed the super-resolution results that a higher proportion of smaller non-fibrillar aggregates are present in the PD samples compared to the controls.

Since the aptamer cannot distinguish between α-synuclein and Aβ, we further used an antibody-based approach in order to understand the composition of the extracted aggregates. We used the SiMPull method described above⁴² to compare aggregates derived from the PD and control brain samples. We chose the 6E10 antibody to detect Aβ and SC for detecting the toxic α-synuclein aggregates. The SC antibody was chosen since it detected the most aggregates in the 10 and 20% sucrose fractions, which exhibited the highest toxic response in our biological assays (Figs. 5, 6). We compared the fluorescence intensity for α-synuclein and Aβ between PD and control samples (Fig. 8A and Supplementary Fig. 8B). By using the same pair of antibodies for capturing and detection, we avoid the detection of monomeric aggregates since they only have one accessible epitope. In the case of 6E10 (Supplementary Fig. 8), there was no significant difference in the number of spots or the intensity between control and PD brain-derived aggregates, suggesting that the differences in aggregate length seen with AD-PAINT are not driven by Aβ. When using the SC antibody (Fig. 8A–C), the detected α-synuclein aggregates derived from PD brains showed no difference in total number of detected aggregates but were lower in intensity than those from control brain (mean PD 35863 vs HC 40477 AU). Aggregate brightness correlates with size with smaller aggregates appearing less bright. The maximum cumulative difference in intensity was 6% between PD and controls, suggesting that the differences in α-synuclein aggregates represent the size discrepancy between PD and controls observed using AD-PAINT imaging. To confirm that the control group consists of longer aggregates, direct STORM-SiMPull was performed with the aggregate-specific MJF antibody (Supplementary Fig. 9A), which showed the highest binding towards fibrils using the α-synuclein sucrose fractions (Fig. 5B). The cumulative frequency distribution shows that the PD group have more aggregates smaller than 80 nm and the controls having longer aggregates exceeding 100 nm in length. Taking the ratio between detected spots using the MJF and the SC antibody (Supplementary Fig. 9B) revealed that the PD brain extracts contain less fibrillar α-synuclein aggregates, detected using the MJF antibody, than the control group. Hence PD brain extracts contain smaller aggregates than the control brain extracts, in agreement with the AFM and super-resolution imaging data (Fig. 7 and Supplementary Fig. 7).

Aggregates extracted from PD patients, which comprise smaller α-synuclein aggregates compared to controls (Figs. 7, 8 and Supplementary Figs. 7, 9), also produce a greater inflammatory response when cultured with BV2 mouse microglial cells over a time course of 96 h (Fig. 8E). The data were normalized to the respective protein concentration. Since the total quantity of aggregates determined by AD-PAINT (Fig. 8D) is similar, this difference in inflammatory capacity is presumed to be related to the difference in α-synuclein aggregate size, and not driven by the number of aggregates. The BV2 cells were also cultured with lipopolysaccharide (LPS) as a positive control, which caused a high TNFα response and with buffer only as a negative control (B) (Supplementary Fig. 10). The buffer did not show any significant difference compared to untreated cells (UNT). The membrane permeabilization assay was not sensitive enough to detect responses from the soaked brain-derived samples, possibly due to the lower abundance of aggregates in comparison to the recombinant α-synuclein preparations.

Discussion

Since α-synuclein aggregates are highly heterogeneous, it is difficult to determine how the physical characteristics such as size and morphology determine their neurotoxic properties. Different hypotheses have been suggested about which exact aggregate forms lead to cell death. While initial studies suggested that amyloid fibrils are the primary neurotoxic species involved in neuronal death due to their abundance in Lewy bodies^55,56, the focus has shifted in recent years towards the soluble α-synuclein oligomers, which form on the pathway of fibril formation. The challenges in characterizing these oligomeric species are that they are highly heterogeneous and less abundant compared to the monomer species^11,14,15,16. In contrast to previous studies that have used mutant forms of α-synuclein or protocols to enrich oligomers in certain states^{11,14,16,32,34,35,36,37,38,39}, we have used here wild-type α-synuclein and studied the wide range of species of different size and structure formed during an aggregation reaction.

Using different ultrasensitive biophysical methods, we were able to characterize differently sized α-synuclein aggregates. To separate the aggregates in size, we chose a centrifugation-based technique, which is shown to be non-perturbative and reproducible^40,41. By using the sucrose density gradient separation, we were able to isolate five different fractions, which all exhibit different species lengths and eccentricities as shown with ThX super-resolution microscopy, ThT imaging, TEM and AFM and SiMPull. Each of these methods has its advantages and limitations. ThT and ThX are both β-sheet-specific dyes, which makes it difficult to study species lacking β-sheet structures. We have seen that the content of β-sheet containing aggregates correlates with higher sucrose densities confirming that the aggregates contain higher amounts of β-sheet structures with increasing fractions. The resolution of the ThX super-resolution images is about 30 nm and the images obtained are of the aggregates in solution⁴⁴.

TEM and AFM are label-free high-resolution methods, which rely on the drying of the sample on grids or mica. Both methods revealed the presence of small, pre-fibrillar spherical aggregates in the 10 and 20% fractions. The 30% fraction contained elongated aggregates. However, TEM only returns 2D projections of the aggregates, while AFM allows analysis of their 3D morphology and volume. According to the AFM analysis, the 20 and 30% fractions contained α-synuclein protofilaments, whereas the 40 and 50% consisted mainly of larger species. The size of the fractionated proteins increases as a function of sucrose gradient as confirmed independently by single-molecule fluorescence and AFM. It must be noted that these techniques are complementary and should not be expected to return identical values. Average values in diameter of the fractionated aggregates measured by these two techniques are different due to their intrinsic sensitivity and the use of different surfaces for deposition. The affinity to small and large aggregates varies, which causes higher or lower average values of the diameter of sampled species compared to the distribution of the populations in solution. When measuring average of particle sizes in number, AFM is largely biased towards small particles as the technique is sensitive towards monomers and small oligomers. When measuring on negative mica surfaces at a scale of 4 × 4 µm, it is also less sensitive in number to larger aggregates (>100 nm). Conversely, single-molecule fluorescence is insensitive to single monomers and small non-β-sheet aggregates which are not ThX active and capable of measuring the size of very large aggregates at a scale of 200 × 200 µm skewing the average to larger values.

With ThX super-resolution, we measured the length and eccentricity from the SR images but lacked structural insight. Then, to gain further structural information, we used an antibody-based imaging approach. The MJF antibody, which binds to aggregated/filamentous α-synuclein, detected roughly 20 times more aggregates in the 50% high-density fractions compared to the lower-density fractions. In contrast, the SC antibody, which targets amino acid residues in the C-terminus of α-synuclein, detected 10 times more aggregates in the lower sucrose density fractions, which have no large fibrillary species present. This is in agreement with previous structural studies of α-synuclein aggregates, indicating that the C-terminal domain is not accessible in fibrils⁵⁰.

After assessing their structural properties in great detail, we used two different biological assays as a measure of toxicity, normalizing to total protein concentration to allow direct comparison between the fractions. We first studied their ability to disrupt lipid membranes, and we observed that the smaller α-synuclein aggregates predominantly found in the 10% and 20% sucrose fractions were more potent in causing membrane destabilisation and permeabilization. We then confirmed that the small aggregates in these same fractions lead to an increased release of the pro-inflammatory cytokine TNFα from BV2 microglial cells. Our finding that smaller α-synuclein aggregates lead to a greater inflammatory response is in agreement with previously published work, which has shown that small α-synuclein oligomers are more likely to induce TNFα production in BV2 cells compared to large α-synuclein fibrils²⁷. There has been extensive study on the size-dependent properties of Aβ⁴⁰, revealing that the smaller aggregates are more prone to cause membrane permeabilization and inflammation, in agreement with our findings for α-synuclein and suggesting that this might be a general property of aggregates.

Taking all of these results together, we isolated five differently sized α-synuclein species and observed that with increasing sucrose density, the length, eccentricity, and abundance of β-sheet containing aggregates increased and that the aggregates with accessible C-terminus decreased. We correlated aggregate size with toxicity and found that small aggregates, <200 nm in size, were the most toxic species in our biological assays. This also allows us to predict the toxicity of samples, whose concentration is too low to perform toxicity measurements, based on the size of the aggregates present.

So far, soluble Aβ aggregates isolated from AD brains were shown to cause long-term potentiation deficits but comparable studies using PD post-mortem tissue were missing⁵⁷. In this initial work, we have extracted soluble aggregates from post-mortem brain tissue from three PD patients and three controls and analysed the length, composition and toxic properties of the extracted aggregates. Insoluble α-synuclein aggregates are known to be present in PD brains but not in control brains⁵⁸, but these are not extracted in the soaking process. Soaking leads to the extraction of only soluble aggregates which were present at similar concentrations in the PD and control brains. This avoids any potential artefacts due to homogenisation of large insoluble aggregates. Previous work has shown that soaking is a minimally perturbative method to preferentially extract the toxic aggregates from post-mortem brain, producing samples containing aggregates of the size and structure found in vivo⁵³. Interestingly, the soluble aggregates from the post-mortem brain were present in similar sizes and morphology to those formed by aggregating α-synuclein in a test-tube, but there were differences in the size distribution of the aggregates from PD patients and controls. This finding is different to what Je et al.⁴² reported, possibly because they looked at both insoluble and soluble species. We found that in the PD brain extracts, the aggregates identified by AD-PAINT (both Aβ and α-synuclein) were shorter than the control extracts, and according to the SiMPull analysis, the α-synuclein aggregates are dimmer in the PD samples, which correlates with aggregate size. The PD samples were also more inflammatory compared to the control samples, suggesting that this difference is due to the presence of smaller α-synuclein aggregates in the PD samples. This is exactly what would be predicted based on the experiments using the different fractions of recombinant aggregates separated using the sucrose gradient, i.e. aggregates from controls, which are larger on average, would be less inflammatory. This result is similar to the result of our recent study on a Braak stage III AD brains, where we found that the hippocampus contained a higher proportion of smaller aggregates than the visual association cortex and that these hippocampal aggregates caused an increased inflammatory response⁵⁹. There was closer agreement in the aggregate sizes measured by super-resolution imaging and AFM than for the sucrose gradient factions. This is because no fibrils were present and the more complex substrate background did not allow the AFM to detect monomeric and small oligomeric species. Both methods show that the largest difference in aggregate size for PD compared to controls is in the range of 0–100 nm. This refines the size range of the toxic species obtained from the sucrose gradient suggesting it is the aggregates smaller than 100 nm that are most toxic.

The concentration of the aggregates in the soaked brain samples is low and probably close to the situation in vivo where low levels of aggregate-induced toxicity of long times lead to the development of disease. This means that sensitive assays are needed to measure the toxicity in these samples. Our highly sensitive membrane permeabilization assay was not sensitive enough to detect changes produced by the soaked brain samples. Therefore, we used an inflammatory assay but this assay still did not have a sufficient dynamic range to allow us to perform immunodepletions to establish if the inflammatory response was caused by Aβ or α-synuclein aggregates, although changes in the inflammatory response correlated with changes in the α-synuclein aggregates. A more sensitive inflammatory assay would be needed to make these experiments feasible. In addition, it would be beneficial to develop sensitive toxicity assays for neurons that work at the low levels of aggregates present in these samples and explore if differences in the α-synuclein aggregate heterogeneity or toxicity correlate with patient symptoms or pathology.

Based on our data, we propose that neurons have mechanisms to form and secrete larger, less toxic soluble aggregates as a route of aggregate removal, thereby preventing neurotoxicity. These mechanisms presumably fail during the development of PD, leading to the release of more small toxic aggregates as well as the formation of insoluble aggregates, Lewy bodies, in some neurons. Future work will explore the variation in the number and size of aggregates in different brain regions and be performed on a larger number of brain samples. Understanding what factors control the size and structure of the aggregates produced in the brain is clearly important to understand what cellular processes fail and lead to the development of PD, and may also be related to the patient symptoms.

In summary, we have shown that the small soluble non-fibrillar α-synuclein aggregates smaller than 200 nm in length are highly toxic species causing lipid membrane permeabilization and inflammation. Furthermore, in human post-mortem samples from the amygdala, a critical brain region where Lewy body pathology and neuroinflammation are observed in PD, the soluble aggregates are significantly smaller and more inflammatory than controls. Hence, our study provides evidence that these small soluble aggregates are the critical species driving toxicity in the PD brain.