Seeded assembly in vitro does not replicate the structures of α-synuclein filaments from multiple system atrophy

The propagation of conformational strains by templated seeding is central to the prion concept. Seeded assembly of α-synuclein into filaments is believed to underlie the prion-like spreading of protein inclusions in a number of human neurodegenerative diseases, including Parkinson's disease, dementia with Lewy bodies (DLB) and multiple system atrophy (MSA). We previously determined the atomic structures of α-synuclein filaments from the putamen of five individuals with MSA. Here, we used filament preparations from three of these brains for the in vitro seeded assembly of recombinant human α-synuclein. We find that the structures of the seeded assemblies differ from those of the seeds, suggesting that additional, as yet unknown, factors play a role in the propagation of pathology. Identification of these factors will be essential for understanding the prion-like spreading of α-synuclein proteinopathies.


Introduction
. Some SNCA mutations and gene multiplications also cause DLB. Abundant α-66 synuclein inclusions are present in all cases of inherited disease. Sequence variation in the 67 regulatory region of SNCA is associated with increased α-synuclein expression and a 68 heightened risk of developing sporadic PD, which accounts for over 90% of cases of this 69 disease (Nalls et al., 2014). Expressed α-synuclein, wild-type or mutant, assembles into 70 filaments in vitro (Conway et al., 1998). Moreover, expression of human mutant α-synuclein 71 in animal models causes its aggregation and neurodegeneration (Giasson et al., 2002). 72 Experimental evidence has shown that assembled α-synuclein from MSA behaves like a 73 prion (Holec & Woerman, 2020). Intracerebral or peripheral injection of MSA brain extracts  Vilar et al., 2008). In some of these studies, filaments were also amplified by using seeds 87 from human brain and recombinant human protein as substrate. 88 We recently showed that the structures of α-synuclein filaments from MSA consist of type 89 I and type II filaments, each with two different protofilaments (Schweighauser et al., 2020). Cryo-EM 2D class averages of MSA case 5 purified filaments before sonication (c) and after 140 sonication (d). 141

Cryo-EM imaging of seeded a-synuclein filaments 142
We used cryo-EM to image the filaments formed following incubation of recombinant a-143 synuclein with seeds from each MSA case. Visual inspection of micrographs of filaments 144 from experiments that used seeds from MSA cases 1 and 2 indicated the presence of two 145 main filament types, which we called type 1 and type 2. Type 1 filaments have an average 146 crossover distance of 800 Å and widths of 60-130 Å; type 2 filaments have a crossover 147 distance of 900 Å and widths of 80-130 Å. We also observed straight filaments with no 148 observable twist. It is unclear if they correspond to filaments of types 1 or 2 that untwisted 149 because of sample preparation artefacts, such as interactions with the air-water interface, or if 150 they represent additional filament types. Due to the lack of twist, we were unable to solve the 151 structures of these filaments. 152 Two-dimensional classification readily separated type 1 and type 2 filaments for further 153 processing and indicated that both types are 2-fold symmetric along their helical axis (Fig.  154 1d). Further 3D classification revealed that type 1 and type 2 filaments occurred in two 155 variants in the data set of filaments that formed with seeds from MSA case 1. They are 156 characterised by small differences in protofilament folds. We called the predominant 157 protofilament 'fold A' and the minor protofilament 'fold B'. We could not identify 158 protofilaments with fold B when seeds from MSA case 2 were used. Reconstructions of filaments containing protofilaments of fold B were less well defined than 165 those of filaments with two protofilaments of fold A. Assembly with seeds from MSA case 5 166 resulted almost exclusively in the formation of a different type of filament, which we called 167 type 3. Type 3 filaments were thinner, more bendy and longer than filaments of types 1 and 168 2. Type 3 filaments have a crossover of 900 Å and widths of 55-65 Å. We

Cryo-EM structures of type 1 and type 2 a-synuclein filaments 258
Most type 1 and type 2 filaments that formed with seeds from MSA case 1, and all the 259 filaments that formed with seeds from MSA case 2, consisted of two protofilaments of fold A 260 that were related by C2 symmetry. Filaments of types 1 and 2 differed in their inter-261 protofilament packing (Figure 2). In type 1 filaments, two salt bridges between E46 and K58 262 held the protofilaments together, by creating a large solvent-filled channel. The inter-263 protofilament interface in type 2 filaments was formed by two salt bridges between K45 and 264 E46 of each protofilament. The smeared reconstructed densities at the points furthest away 265 from the helical axis suggest that the inter-protofilament interface of type 2 filaments is more 266 flexible than that of type 1 filaments. Protofilament fold A consists of 8 β-sheets: β1-6 form 267 a roughly Z-shaped hairpin-like structure, with glycines or KTK motifs between the β-268 sheets at the bends; β7-8 fold back against β4, leaving a small triangular cavity between β5, 269 β6 and β7. This fold is unlike any of those of the MSA type I and type II protofilaments. It is 270 almost identical to the protofilament fold that was reported for in vitro aggregated When using seeds from MSA cases 1 and 2, which contain a mixture of type I and type II 405 filaments, and recombinant human a-synuclein as substrate, we observed the formation of 406 type 1 and type 2 filaments. When using seeds from MSA case 5, with only type II filaments, 407 we observed the formation of filaments of type 3. These observations suggest that in seeded 408 assemblies, type I filaments overshadow type II MSA filaments, despite the observation that 409 Type 3 filaments, which assembled from MSA type II seeds, fit the model of structural 429 equivalence between seeds and seeded assemblies better than type 1 and type 2 filaments, 430 because their structure overlaps almost completely with that of type IIB protofilaments from 431 the putamen of patients with MSA. We previously attributed additional cryo-EM densities at 432 the inter-protofilament interfaces of type I and type II MSA filaments to negatively charged, 433 non-proteinaceous molecules. It is possible that the absence of these molecules in the seeded 434 assembly experiments led to the formation of a structure that represents only half of the seed 435 structures. These findings indicate that protofilament IIB, but not IIA, can form from 436 recombinant a-synuclein through seeded assembly without added cofactor. 437 Abundant GCIs in oligodendrocytes are the major neuropathological hallmark of MSA 438 (Papp et al., 1989). Thus, differences in the cellular milieu between oligodendrocytes and 439 other brain cells may play a role in the seeded aggregation of MSA filaments. 440 Oligodendrocytes have been shown to transform misfolded a-synuclein into a GCI-like strain 441 (Peng et al., 2018). 442 Besides the possible incorporation of other molecules in α-synuclein filaments from human 443 brain, it is also conceivable that recombinant α-synuclein is not able to form MSA filaments. The filament preparations used in this study have been described (Schweighauser et al., 516 2020). Briefly, frozen putamen from MSA cases 1, 2 and 5 was homogenised in 20 % vol 517 (w/v) extraction buffer (10 mM Tris-HCl, pH 7.5, 0.8 M NaCl, 1 mM EGTA, 10% sucrose, 2 518 % sarkosyl, pH 7.5) and incubated for 30 min at 37 ºC. The homogenates were centrifuged 519 for 10 min at 10,000g at room temperature, followed by a 20 min spin of the resulting 520 supernatants at 100,000g. The pellets were resuspended in 500 μl/g extraction buffer and 521 centrifuged at 3,000g for 5 min to remove large contaminants. The supernatants were diluted 522 in 50 mM Tris-HCl, pH 7.5, containing 150 mM NaCl, 10% sucrose and 0.2% sarkosyl, and 523 centrifuged at 166,000g for 30 min. Sarkosyl-insoluble pellets were resuspended in 50 μl/g 524 tissue and filament concentrations estimated by negative-stain EM. Prior to seeded assembly 525 experiments, pellets were centrifuged at 2,000 g for 5 min, the resulting supernatants were 526 diluted 10-fold, and sonicated in an Eppendorf tube using a VialTweeter (Hielscher) at a 527 cumulative power of 100 W. Sonication did not alter the structure of the seeds, as suggested 528 by negative-stain EM (Figure 1 -figure supplement 1), and as confirmed by cryo-EM 2D 529 class averages of the seeds before and after sonication (Figure 1 -figure supplement 2). 530 531 Seeded assembly 532 Purified recombinant α-synuclein was centrifuged at 20,000 x g for 1 hr to remove potential 533 aggregates. 70 μM recombinant α-synuclein was incubated with 2 μM MSA seeds (as 534 assessed by negative-stain EM) in 100 mM PIPES pH 6.5, 500 mM NaCl, 0.05% NaN3, and 535 5 μM thioflavin-T, in a final volume of 200 µl per experiment. Controls used buffer without 536 seeds. Seeded assembly proceeded for 120 h at 37 ºC in a FLUOstar Omega (BMG Labtech) 537 microplate reader where the samples were alternatingly shaken for 1 minute at 400 rpm, and 538 left to rest for 1 minute, during which fluorescence was measured. 539 For cryo-EM, seeded assembly conditions were identical, but no thioflavin-T was added to 540 the buffer and the samples were shaken continuously for 72 hrs. Seeded assembly 541 experiments for cryo-EM were also performed in PBS buffer, supplemented with 1 mM 542 pyrophosphate and 0.05% NaN3. The resulting filaments were pelleted, resuspended in 200 μl 543 and sonicated as described above, and then used as seeds (2 µM) for a second-generation 544 seeded assembly experiment with recombinant α-synuclein (70 µM) in the same PBS buffer. 545 546

Cryo-EM grid preparation and imaging 547
Prior to freeze plunging, filaments were pelleted for 45 min at 100,000x g and resuspended at 548 100 μM a-synuclein in 50 mM Tris, pH 7.5, 50 mM NaCl. Four μl of sample was applied to 549 glow-discharged 1.2/1.3 holey carbon coated gold grids (Quantifoil AU R1.2/1.3, 300 mesh) 550 for 30s, blotted with filter paper for 3.5 s and plunge-frozen in liquid ethane using an FEI 551 Vitrobot Mark IV. Filaments were imaged on a Thermo Fischer Titan Krios microscope 552 operating at 300 kV equipped with a Gatan K2 Summit direct detector in counting mode and 553 a GIF Quantum energy filter (Gatan) with a slit width of 20 eV to remove inelastically 554 scattered electrons. Acquisition details are given in Tables 1 and 2.  555 556

Helical reconstruction 557
Filaments were reconstructed in RELION-3.1 (Zivanov et al., 2020) using helical 558 reconstruction (He & Scheres, 2017). Movie frames were corrected for beam-induced 559 motions and dose-weighted in RELION using its own motion-correction implementation 560 (Zivanov et al., 2018). Non-dose-weighted micrographs were used for CTF estimation with 561 CTFFIND-4.1 (Rohou & Grigorieff, 2015). Filaments were picked manually, ignoring those 562 without a clear twist. Initially, particle segments were extracted using a box size of 550 pixels 563 and an interbox distance of 14 Å and downscaled to 225 pixels for 2D classification. For 564 filaments formed from the seeds of MSA cases 1 and 2, filament types 1 and 2 were 565 separated at this initial 2D classification stage. Crossover-distances were obtained by manual 566 measurements in the micrographs and used to calculate initial estimates for the helical twist 567 of the different filament types: -1.0º for type 1; -0.8º for type 2; and -1.5° for type 3, 568 assuming a helical rise of 4.75 Å. De novo 3D initial models were then constructed from 2D 569 class averages representing one whole cross-over of the different filament types using the 570 relion_helix_inimodel2d program (Scheres, 2020). Subsequently, segments were 571 re-extracted without down-sampling in boxes of 256×256 pixels for use in 3D auto-572 refinements and classifications. Several rounds of refinements were performed, while 573 progressively increasing the resolution of the starting model from 10 Å to 4.5 Å and 574 switching on optimisation of the helical rise and helical twist once β-strands were separated 575 in the starting model. For filaments from seeds of MSA case 1, additional 3D classifications 576 focussed classifications on exterior regions of the filament were used to distinguish the 577 presence of minority polymorphs (with protofilament fold B as described in the main text). 578 Final reconstructions were obtained after Bayesian polishing and CTF refinement, followed 579 by 3D auto-refinement, a 3D classification step without alignment to select the segments 580 contributing to the best classes, a final round of 3D auto-refinement and standard RELION 581 post-processing with a soft solvent mask that extended to 20 % of the box height. 582 583

Atomic modelling 584
Atomic models of the filaments were built de novo in Coot (Emsley & Cowtan, 2004) using 585 the maps of the data set for MSA case 2 for type 1 and type 2 filaments with protofilament 586 fold A, and maps of the data set for MSA case 2 for type 1 and type 2 filaments with 587 protofilament fold B. For protofilament fold A, the atomic model with PDB-ID 6UFR of 588 E46K a-synuclein (Boyer et al., 2020) was used as guide. For type 3 filaments, the atomic 589 model with PDB-ID 6PEO (Boyer et al., 2019) of H50Q a-synuclein was used. Models 590 comprising 6 β-sheet rungs were refined in real-space using ISOLDE (Croll, 2018), with 591 interactive flexible molecular dynamics to obtain optimal β-sheet packing chemistry. The 592 resulting models were validated with MolProbity (Chen et al., 2010). Details about the 593 atomic models are described in Table 1. 594 The schematics in Figure 3 -figure supplement 3e-f and Figure 5 -figure supplement 1b  595 were made with T.Nakane's atoms2svg.py script, which is publicly available from 596 https://doi.org/10.5281/zenodo.4090924. 597 598