Similar neuronal imprint and absence of cross-seeded partner fibrils in α-synuclein aggregates from MSA and Parkinson’s disease brains

Aggregated alpha-synuclein (α-syn) is a principal constituent of Lewy bodies (LBs) and glial cytoplasmic inclusions (GCIs) observed respectively inside neurons in Parkinson’s disease (PD) and oligodendrocytes in multiple system atrophy (MSA). Yet, the cellular origin, the pathophysiological role, and the mechanism of formation of these inclusions bodies (IBs) remain to be elucidated. It has recently been proposed that α-syn IBs eventually cause the demise of the host cell by virtue of the cumulative sequestration of partner proteins and organelles. In particular, the hypothesis of a local cross-seeding of other fibrillization-prone proteins like tau or TDP-43 has also been put forward. We submitted sarkosyl-insoluble extracts of post-mortem brain tissue from PD, MSA and control subjects to a comparative proteomic analysis to address these points. Our studies indicate that: i) α-syn is by far the most enriched protein in PD and MSA extracts compared to controls; ii) PD and MSA extracts share a striking overlap of their sarkosyl-insoluble proteomes, consisting of a vast majority of mitochondrial and neuronal synaptic proteins, and (iii) other fibrillization-prone protein candidates possibly cross-seeded by α-syn are neither found in PD nor MSA extracts. Thus, our results (i) support the idea that pre-assembled building blocks originating in neurons serve to the formation of GCIs in MSA, (ii) show no sign of amyloid cross-seeding in either synucleinopathy, and (iii) point to the sequestration of mitochondria and of neuronal synaptic components in both LBs and GCIs.

Besides their histopathological characterization, the origins, the mechanisms of formation, and the pathophysiological role of IBs remain unelucidated. In agreement with early studies showing that IB-harboring neurons show no apparent signs of apoptosis 7,8 and that there is no correlation between the extent of LB pathology and the extent of neuronal depletion in PD [9][10][11] , it appears likely that the α-syn fibrils accumulated in the IBs are not toxic by themselves, and that their compaction represents the result of a neutralization process rather. Instead, the concomitant and cumulative incorporation of mitochondria and proteins essential for cell function into the growing IBs could sporadically be detrimental to the host cell 12 . At the same time, amyloidogenic partner proteins like tau or TDP-43 could also, in some instances, be locally cross-seeded and cause the cell's demise [13][14][15] . In both cases, the catabolism of IBs released by the dying host cells in the extracellular space could then cause a local leak of small IB fragments and α-syn fibrils from the IB mass, fragments that in turn would be capable of being taken up by the neighboring cells in which they would trigger a de novo aggregation of endogenous α-syn, the formation of novel IBs, and so on. Indeed, in mice, the intracerebral injection of IBs extracted from human DLB brains leads to the development and to the spread of a synucleinopathy with neuropathological and cytological features that are strikingly identical to the one observed after injections of pure recombinant α-syn fibrils 16,17 . Thus, if this IB-dependent intercellular spread mechanism holds, it is tempting to speculate that the composition of IBs present in a host cell could reflect the history of the spread, i.e., show proteins of the donor cell transferred together with the IB fragment seeds and eventually associated with the α-syn fibrils mass in the neo-formed IB.
Such a theoretical possibility of retracing the spread history of IBs could probably help to solve the enigma concerning GCIs in MSA. These inclusions feature a high load of fibrillar α-syn, which contrasts with the very low physiological abundance of the protein in mature oligodendrocytes 18,19 . To date, the origin of these oligodendroglial IBs remains to be unraveled.
However, these candidates' relevance in the disease pathogenesis is limited by the experimental procedures' ability to isolate IBs from the surrounding subcellular structures and proteins.
To this end, we used a different approach in the present study. We purified and isolated aggregated α-syn and its associated insoluble proteomes from PD and MSA brains using their sarkosyl-insolubility. We used Sarkospin, a previously developed procedure for purifying pathological protein aggregates by sedimentation 17,28,29 . By adapting Sarkospin to α-syn, we sought to specifically isolate the aggregated forms of the protein from their physiological monomeric and oligomeric counterparts. The latter separation allowed us to scrutinize the accompanying insoluble proteomes associated with each disease and shed light on the cellular origin and the components of IBs.

Extraction of pathological alpha-synuclein from synucleinopathy brains and separation from its regular counterparts by an adapted Sarkospin procedure
Intending to extract and identify the insoluble proteomes associated with aggregated α-syn for each synucleinopathy, we adapted the Sarkospin procedure to the purification of α-syn pathological assemblies found in human post-mortem tissue samples of PD and MSA subjects. We previously developed this method to extract aggregated TDP-43 29 . It was adapted by using ultracentrifugation of the sample mixed within a sucrose cushion after its solubilization in sarkosyl at 37°C with simultaneous nuclease treatment (Fig. 1a). The resulting supernatant contains sarkosyl-soluble material. The insoluble entities are collected in a dry pellet that is compatible with a further identification or quantification of its protein content by immunoblotting or mass spectrometry. Above all, the extraction of pathological assemblies through insolubility allows a direct comparison to control samples by applying the same procedure to healthy brain samples. We thus proceeded with parallel Sarkospin extractions, sedimentations, and filter-trap assays of sarkosyl-insoluble aggregates (Fig. 1b, c) from the brains of human control, PD, and MSA subjects (n=3 independent subject brains per control or disease group). The results indicate that the adapted Sarkospin procedure efficiently separated pathological aggregated α-syn from its soluble counterparts. As observed in Fig. 1b, high molecular weight α-syn species that are filter-trapped on nitrocellulose membrane are specifically present in the pellet 7 fractions of PD and MSA samples (MJFR1, nc). In contrast, small mono-and oligomeric forms of the protein that get to pvdf membranes are observable mainly in all groups' supernatants (MJFR1, pvdf).
Another marker of pathological forms of the protein, namely its S129-phosphorylation (EP1536Y positive signal), indicates that large insoluble pS129-α-syn-containing aggregates are isolated in the pellet fractions for synucleinopathy samples (EP1536Y, nc) as opposed to the physiological EP1536Y-positive soluble entities found in all supernatants (EP1536Y, pvdf). Also, α-syn aggregates purified in synucleinopathy sample pellets have proved to be peculiarly resistant to denaturation (SDS, heat) and proteolysis (Proteinase K) while supernatants comprised only sensitive monomeric bands (Sup. Fig. 2a, b). Notably, actin, yet physiologically forming large assemblies, is found only in supernatant Sarkospin fractions and witnesses the procedure's efficient solubilization (Fig. 1a, b).
By quantification on filter-trap immunoblots for n=9 brains, of Sarkospin supernatant and pellet fractions content in the proteins mentioned above, it appeared that over 80% of the nc-trapped α-syn in brain extracts from PD and MSA patients was pelleted, and less than 5% in brain extracts from control subjects (Fig. 1c, MJFR1 nc). In PD and MSA samples, S129-phosphorylated α-syn coincided with the insoluble α-syn pool since 90% of the pS129-α-syn signal (Fig. 1c, EP1536Y nc) was found in the pellets. In clear contrast, this figure was less than 5% for the control subjects. Noteworthy, a small subset of insoluble aggregates passes to pvdf membranes, but the majority (60-70% for PD and MSA, 99% for CTL) of α-syn present on this support is from supernatants.
Collectively, these results indicate that the novel Sarkospin procedure efficiently extracts pathological α-syn aggregates from human brain samples and physically separates these entities from physiological mono-and oligomeric α-syn forms present in healthy brain samples. Thus, we attained a purification method compatible with a subsequent proteomic analysis to extract and identify insoluble proteins from synucleinopathy subject brains and their direct comparison with parallel Sarkospin extractions from control patient brains.

Sarkospin pellet fractions
Proteomic analysis of Sarkospin pellets extracted from control, PD, and MSA samples (n=3 independent subject brains per group) identified a total of 1022 proteins that resisted sarkosyl solubilization ( Fig. 2a,  Interestingly many of the most significantly enriched candidates were not identified in previous studies and had a null bibliographic LB and GCI score (smallest-sized red and blue bubbles close to α-syn on Fig. 2a and 2b, respectively). Besides, many proteins previously identified (i.e., with high bibliographic LB or GCI indices) are found in the extracts. Still, they are not significantly enriched in synucleinopathy pellets compared to control patient samples with no synucleinopathy (large dark bubbles on Fig. 2a, b). This could be explained by the absence in previous studies of a comparator group with equivalent fractions from healthy brains allowing to rule out an unselective enrichment.
Indeed, for proteins physiologically highly expressed in the brain, their high abundance could simply lead to an increased detection probability whatever the IB purification method, irrespective of the IB pathology. We challenged this possibility for all the 1022 proteins present in the insoluble proteome.  Therefore, the present Sarkospin procedure, coupled with a subsequent comparative proteomic analysis, is an adequate method for unraveling insoluble proteomes and unbiasedly refining the list of candidates by considering their fold enrichment with regards to control samples. This approach defines the insoluble proteomes specific for each synucleinopathy and identifies LB and GCI components.

Major overlap between PD and MSA insoluble proteomes
A total of 206 unique proteins were gated previously as significantly enriched in at least one of the synucleinopathy types compared to controls ( Fig. 2a and 2b, red and blue, Appendix D). Among this group of proteins, 84 are significantly enriched in PD as well as in MSA compared to controls with FC>1.5 and p<0.05 (Appendix E). Inversely, among the 206 disease-associated proteins, only seven proteins were found to be selectively enriched in one synucleinopathy compared to the other, with three PD-specific and four MSA-specific (Fig. 4, red and blue, Appendix F-G). This highlights the extreme narrowness of the PD-and MSA-specific insoluble proteomes and indicates that the most substantial part of synucleinopathy-associated insoluble proteomes is shared between these two pathologies.
In other words, these data show the tremendous overlap of PD and MSA-associated insoluble proteomes. More specifically, in the most enriched proteins for both diseases or even for MSA specifically, a consistent proportion of candidates have been identified to play a role in PD pathogenesis (Appendix C, E, G). Namely, for example, NipSnap-1 (average FC=7 p=8.0x10 -3 ) was shown to be implicated in Parkin-related mitophagy 30 ; Septin-5 (average FC=4.6 p=2.5x10 -3 ) was associated with dopamine-dependent neurotoxicity in early-onset PD also linked to Parkin 31 ; even Protein phosphatase methylesterase 1 which we find to be MSA-specific (Fig. 4, a) has previously been also implicated in PD pathogenesis as enriched in PD subject substantia nigra 32 and playing a role in α-syn phosphorylation through PP2A activity regulation 33 .
Notably, from the four MSA-specific candidates we identify, none shows a glial-specific expression or a glial tropism (Fig. 4, a-d, Appendix G). Besides protein phosphatase methylesterase Altogether, while the absence of oligodendroglial makers in the MSA-specific insoluble proteome is unexpected, the massive overlap of PD and MSA insoluble proteomes could suggest a common neuronal origin of these insoluble entities in both PD and MSA. To put this possibility under scrutiny, we compared the lists of 130 and 160 proteins enriched respectively in PD and MSA pellets, to the published insoluble proteome found in an experimental neuronal synucleinopathy modelized in mouse neuronal cultures using preformed α-syn fibrils (PFFs) as seeds 12 (Sup. Fig. 3). Strikingly, 38 and 58 insoluble proteins are shared between these PFF-treated neurons and PD and MSA respectively, with several -such as NipSnap-1 for PD, and Arginase-1 for MSA, showing a substantial enrichment in both proteomics studies (Sup. Fig. 3a and   3B). This striking overlap with yet completely independent studies and samples types validates several candidates, and brings a strong support to the notion that GCIs in MSA are made of α-syn amyloid building blocks preassembled in neurons To confirm the enrichment of these candidates in PD, MSA or PFF-treated neurons insoluble proteomes, we ran an independent Sarkospin assay and subjected the fractions to the dot blot analysis used for the validation of Sarkospin (Fig. 1), with antibodies directed to ten proteins of interest, after validating their specificity on western blots (Sup. Fig. 4  First of all, a clear-cut result is that no cross-seeded amyloidogenic partner proteins are associated with fibrillar α-syn in PD and MSA.

16
The bubble chart representation of GO analysis in Fig. 5 compellingly shows that the most significantly enriched and importantly represented GO clusters for synucleinopathies indicate a neuronal imprint, with a remarkable synaptic and mitochondrial membrane trend (Fig. 5). Indeed, all biological processes, molecular functions, and cellular components reaching maximal fold change and statistical significance are related to synaptic and mitochondrial localization or pathways. there is an initial florid neuronal synucleinopathy that develops to progressively decrease with time and leave the place to the progressive appearance of oligodendrocytes harboring GCI-like inclusions 34 . This is also in line with the notion that overexpression of α-syn in oligodendrocytes fails to cause the buildup of GCIs 28 . Finally, our observations do not support the idea that cross-seeding processes involving amyloidogenic proteins other than α-syn occur in PD and MSA.

DISCUSSION
The development of a novel procedure by adaptation of our formerly published Sarkospin method 28,29 (Fig. 1) allowed the extraction and purification of pathological aggregates from synucleinopathy subject brain tissues. The data presented here confirm that α-syn is the only amyloid and by far the most enriched protein present in PD and MSA sarkosyl-insoluble fractions of brain samples 6 (Fig. 2). The presence of this neuronal protein in oligodendrocyte inclusions is the core point of the enigma we sought to address in the present study 35 : what are the origin and the reason for the formation of LBs and GCIs?
Characterizing the proteomes of LBs and GCIs has been repeatedly attempted in previous studies in order to understand which proteins are involved in their formation and their packing.
Besides α-syn, mass spectrometry analysis of these IB-enriched fractions identified hundreds of proteins that potentially play a key role in the pathogenesis or the origin and mechanisms of formation of LBs and GCIs. These proteins include structural and cytoskeletal elements (neurofilaments, tubulins, tubulin polymerization promoting protein, TPPP/p25), α-synuclein-binding proteins (14-3-3, synphillin-1), components of the ubiquitin-proteasome system, α-ß-crystallin, heat shock proteins, or DJ-1. However, these candidates' relevance to the molecular pathology of IBs can be limited by the ability of the experimental procedures to isolate IBs with sufficient purity and yield from the surrounding structures and protein contaminants. Recently, a purification method including a partial proteolysis step allowed the refinement of this list of proteins and demonstrated a high burden of synaptic vesicle-related proteins of the IBs 25 . These candidates are implicated in clathrin-mediated endocytosis (clathrin, AP-2 complex, dynamin), retrograde transport (dynein, dynactin, spectrin), or synaptic vesicle fusion (synaptosomal-associated protein 25, vesicle-associated membrane protein 2, syntaxin-1).
Using extracts from control, PD, and MSA brains, the comparative proteomic analysis we performed identified proteins enriched in PD and MSA brain samples (Fig. 2), with novel undescribed candidates, and refinement by excluding a list of previously identified proteins as probable false positives and/or protein not tightly associated with the sarkosyl-insoluble α-syn amyloids of IBs (Fig.   2) 3,5,6,20,21,[23][24][25]27 .
The protein enrichment in the respective PD or MSA insoluble proteome showed a tremendous overlap between them, with only few proteins being significantly different (Fig. 4). Lastly, the latter results associated with a gene ontology analysis revealed compellingly that both insoluble proteomes are composed of the vast majority of neuronal proteins, with a significant synaptic overrepresentation (Fig. 5), as well as numerous mitochondrial proteins. These data are in total agreement with previous proteomic studies 25 and microscopic analysis 4 which revealed a burden of synaptic-vesicle-related proteins, and membranous structures, such as vesicles or dysmorphic mitochondria. Noteworthy, our study independently validated many synaptic proteins found in the previously mentioned study 25 .
These results strengthen the hypothesis of the neuronal origin of GCIs. It is tempting to speculate that in MSA α-syn could primarily get aggregated in neurons, then released as preformed inclusions bodies, eaten-up by surrounding oligodendrocytes, and eventually matured/stored as inert GCIs 34 .
The fact that oligodendrocyte-specific proteins are not enriched in MSA samples by our Sarkospin procedure is in line with the idea that the oligodendrocyte contribution is late, with glial proteins only secondarily and loosely associated with the GCIs and not properly with the amyloid core of the inclusion, previously pre-assembled in neurons.
Of particular importance, the list of proteins identified in the present study is restricted to the production of peptides identifiable by mass spectrometry, deriving from the presence of tryptic cleavage sites. Suppose one considers the high representation of neurodegenerative disease-associated proteins in partially trypsin-resistant proteins. In that case, it is possible that their enrichment here was underestimated, or even that some escaped identification. However, the conditions used here are known to reveal TDP-43 and tau 29 , and the absence of detection of cross-seeded amyloid partners does not thus derive from an insufficient proteolytic cleavage.
Extending the present study to other brain regions and at different pathology stages would be of particular interest. In pafticular, using amygdala or putamen extracts for focusing on LBs or GCIs protein compositions, respectively, should help refining the present lists and identifying less represented candidates. Also, Sarkospin extraction and purification of synucleinopathy subject brain tissue yet devoid of any inclusion 36 , and the insoluble proteomes' comparison to the ones we defined here is of high importance for understanding the formation and composition of IBs. The Sarkospin procedure we developed here allows testing the seeding ability 37 or the pathogenicity of aggregated proteins 38 , an asset to understand the role of the inclusions in the disease process. Also, facing these insoluble proteomes' components to lists of identified interactants of mono-oligomeric physiological α-syn, or α-syn amyloid binders would help understand why the presence of these proteins in the inclusions and to decrypt their mechanisms of pathological association.

Human brain samples
The samples were obtained from brains collected in a Brain Donation Program of the Brain Bank "GIE NeuroCEB" (Neuro-CEB BB-0033-00011). The consents were signed by the patients themselves or their next of kin in their name, in accordance with the French Bioethical Laws. The Brain Bank GIE NeuroCEB has been declared at the Ministry of Higher Education and Research and has received approval to distribute samples (agreement AC-2013-1887). Human cortices (cingulate gyrus) were dissected from freshly frozen post-mortem brain samples from n=3 control, sporadic PD or MSA subjects respectively. The choice of cingulate gyrus samples was based on a prior biochemical analysis of different brain regions from the CTL, PD and MSA subjects in this series by immunoblotting against pS129 phosphorylated (Sup. Fig. 1a) and aggregated (Sup. Fig. 1b) α-syn.
On the basis of these data, we found that the cingulate gyrus was the most comparable region in terms of total amyloid α-syn load for these PD and MSA subjects. In addition, the pathological records regarding the contralateral hemispheres indicated the presence of substantial amounts of LBs and GCIs in this area for the PD and MSA subjects, respectively. We thus proceeded with a comparative proteomic analysis of the insoluble proteomes using this brain region.

Sarkospin procedure
For the extraction and purification of aggregates from brains samples, the Sarkospin procedure was adapted from previously published protocols 28,29 . Brain tissue samples were homogenized at   Table).

Analysis of the protein contents of Sarkospin fractions by filter trap
Immunoreactivity was whether visualized by chemiluminescence or infrared using Clarity ECL and Chemidoc (Biorad) or Odissey systems (Li-Cor) respectively.

Proteinase K treatments and western blot
For PK resistance assays, equal volumes of solubilized homogenates, Sarkospin supernatants or pellets fractions were treated or not with 1 µg.ml -1 Proteinase K (Sigma) from 0 to 60 minutes at 37°C. At the end of the indicated time, samples were added Laemmli 1x prior to denaturation at 95 °C for 5 min, and loaded on Mini-Protean TGX 12% gels (Biorad) followed by SDS-PAGE electrophoresis. Gels were whether stained for total protein amount (silver stain, Biorad) or transferred on nitrocellulose 0.2 µm membranes with Trans-Blot Turbo transfer system (Biorad) using the Mixed molecular weight program. Membranes were fixed with PFA, and proteins were immunolabelled as described for filter trap.

Mass spectrometry analysis of Sarkospin pellets
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD024998.

Sample preparation and protein digestion
Protein to +7 charged ions were selected for fragmentation. Others settings were as follows: no sheath nor auxiliary gas flow, heated capillary temperature, 275 °C; normalized HCD collision energy of 30% and an isolation width of 1.6 m/z. Monoisotopic precursor selection (MIPS) was set to Peptide and an intensity threshold was set to 5 x 10 3 .

ACKNOWLEDGMENTS
The project was conducted using financial support from the Region Nouvelle Aquitaine, the

COMPETING INTERESTS
Authors declare no competing interests.

SUPPLEMENTARY FIGURES
Sup. Fig. 1. Comparison of pathological alpha-synuclein load in different brain regions of synucleinopathy subjects by immunoblotting. Brain homogenates from different brain regions (gyrus, amygdala, putamen, midbrain and cerebellum) of the three control (black), PD (red) and MSA (blue) human subjects of the study were subjected to dot blot and subsequent immunolabelling and quantification of pS129-positive (a, EP1536Y/actin) and aggregated (b, LB509/actin) α-synuclein load. The relative amounts (A.U.) are plotted for each subject brain region, and the respective p-values of Tuckey corrected two-way ANOVAs are represented above each couple of comparisons.