Induction of de novo α-synuclein fibrillization in aneuronal model for Parkinson’s diseaseMohamed-Bilal Faresa, Bohumil Macoa, Abid Oueslatia,1, Edward Rockensteinb, Natalia Ninkinac, Vladimir L. Buchmanc,Eliezer Masliahb, and Hilal A. Lashuela,d,2

aBrain Mind Institute, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, CH-1015, Switzerland; bDepartment of Neurosciences, University ofCalifornia San Diego, La Jolla, CA 92093; cSchool of Biosciences, Cardiff University, Cardiff, CF10 3AX, United Kingdom; and dQatar Biomedical ResearchInstitute, Hamad Bin Khalifa University, Doha, 5825, Qatar

Edited by Thomas C. Südhof, Stanford University School of Medicine, Stanford, CA, and approved January 4, 2016 (received for review July 25, 2015)

Lewy bodies (LBs) are intraneuronal inclusions consisting primarilyof fibrillized human α-synuclein (hα-Syn) protein, which representthe major pathological hallmark of Parkinson’s disease (PD). Al-though doubling hα-Syn expression provokes LB pathology in humans,hα-Syn overexpression does not trigger the formation of fibrillarLB-like inclusions in mice. We hypothesized that interactions betweenexogenous hα-Syn and endogenous mouse synuclein homologs couldbe attenuating hα-Syn fibrillization in mice, and therefore, we system-atically assessed hα-Syn aggregation propensity in neurons derivedfrom α-Syn–KO, β-Syn–KO, γ-Syn–KO, and triple-KO mice lacking ex-pression of all three synuclein homologs. Herein, we show that hα-Synforms hyperphosphorylated (at S129) and ubiquitin-positive LB-like in-clusions in primary neurons of α-Syn–KO, β-Syn–KO, and triple-KOmice, as well as in transgenic α-Syn–KO mouse brains in vivo. Impor-tantly, correlative light and electron microscopy, immunogold labeling,and thioflavin-S binding established their fibrillar ultrastructure, andfluorescence recovery after photobleaching/photoconversion experi-ments showed that these inclusions grow in size and incorporatesoluble proteins. We further investigated whether the presence ofhomologous α-Syn species would interfere with the seeding andspreading of α-Syn pathology. Our results are in line with increasingevidence demonstrating that the spreading of α-Syn pathology is mostprominent when the injected preformed fibrils and host-expressedα-Syn monomers are from the same species. These findings provideinsights that will help advance the development of neuronal andin vivo models for understanding mechanisms underlying hα-Synintraneuronal fibrillization and its contribution to PD pathogenesis,and for screening pharmacologic and genetic modulators of α-Synfibrillization in neurons.

Parkinson’s disease | alpha-synuclein | aggregation

The aggregation of proteins into fibrillar structures is a keyhallmark of many neurodegenerative disorders. In Parkin-

son’s disease (PD), α-synuclein (α-Syn), a predominantly pre-synaptic protein involved in the regulation of neurotransmitterrelease, abnormally fibrilizes and forms intraneuronal inclusionstermed “Lewy bodies” (LBs) (1, 2). So far, the mechanisms un-derlying LB formation remain poorly understood, and the impactof LB presence on neuronal viability remains controversial, inpart due to the lack of animal models recapitulating α-Synfibrillization into LB-like inclusions in the brain.Because patients with familial history of parkinsonism were

found carrying either multiplications or point mutations of theα-Syn gene SNCA (2), most animal models of PD have beengenerated by overexpressing WT human α-Syn (hα-Syn) or mu-tant forms linked to familial PD (3). Strikingly, although rodentmodels expressing hα-Syn do not recapitulate the formation offibrillar LBs within dopaminergic neurons, hα-Syn overexpressionin Drosophila led to dramatic neuronal loss accompanied withLB-like structures comprising fibrillar α-Syn (4). Drosophila, un-like rodents, do not express an endogenous α-Syn homolog,implying that hα-Syn fibril formation could be more favorablein models lacking endogenous α-Syn expression. Subsequent ex-

periments in mice supported this suggestion, as hα-Syn transgenic(Tg) mice lacking endogenous mouse α-Syn (mα-Syn) expressionexhibited exacerbated pathology compared with WT counterparts(5), and manifested fibrillar/granular accumulations in the olfac-tory bulb (6). Interestingly, in vitro test-tube experiments alsoshowed that small amounts of mα-Syn directly inhibit the fibrilli-zation of purified hα-Syn protein in solution (7).Despite these interesting observations, it remained unclear

whether endogenously expressed synuclein homologs directlyinhibit hα-Syn aggregation in neurons, and whether ectopichα-Syn expression in their absence would allow de novo hα-Synfibrillization events. Therefore, we systematically evaluated thepropensity of hα-Syn to aggregate on synuclein-KO backgrounds.Using a battery of biochemical and imaging techniques, we dem-onstrate that in cultured primary neurons and brains of mα-Syn–KO mice, overexpressed hα-Syn aggregates readily into inclusionsthat are similar to LBs in terms of solubility, immunoreactivity, andamyloidogenicity, and represent a bona fide fibrillization process asrevealed by serial-section transmission electron microscopy (ssTEM),live imaging, and response to pharmacological aggregation inhibitors.Similarly, primary neurons lacking expression of β-Syn (β-Syn−/−) orof all three homologs (α-, β-, and γ-Syn) also manifest enhanced hα-Syn aggregation, thereby suggesting that endogenous synucleinhomologs may represent natural regulators of abnormal α-Syn ag-gregation. To dissect potential mechanisms underlying this phe-nomenon, we performed immunoprecipitation and surface plasmon

Significance

Although it has been established for over 100 years that Lewybodies (LBs) represent the major pathological hallmark of Parkin-son’s disease (PD), we still do not know why these fibrillar intra-neuronal inclusions of α-synuclein (α-Syn) protein form, or howthey contribute to disease progression. One of the major causesunderlying this gap in knowledge is the lack of animal models thatreproduce the formation of fibrillar LB-like inclusions. In this study,we show that the absence of human α-Syn (hα-Syn) fibrillizationinto LBs in mice can be attributed to interactions between hα-Synand its endogenously expressed mouse α-Syn homologue. More-over, we provide well-characterized primary neuronal and in vivomodels that recapitulate the main molecular feature of PD, bonafide α-Syn fibrillization.

Author contributions: M.-B.F., B.M., A.O., E.R., E.M., and H.A.L. designed research; M.-B.F.,B.M., A.O., and E.R. performed research; E.R., N.N., V.L.B., and E.M. contributed newreagents/analytic tools; M.-B.F., B.M., A.O., E.R., N.N., V.L.B., E.M., and H.A.L. analyzeddata; and M.-B.F., N.N., V.L.B., and H.A.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Present address: Centre de Recherche du Centre Hospitalier de Québec, Axe Neurosci-ence et Département de Médecine Moléculaire de l’Université Laval, Québec, CanadaG1V 4G2.

2To whom correspondence should be addressed. Email: hilal.lashuel@epfl.ch.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1512876113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1512876113 PNAS Early Edition | 1 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

resonance (SPR) experiments, which showed that mα-Syn interactspreferentially with aggregated hα-Syn preformed fibrils (PFFs)rather than with monomers. Importantly, in vitro aggregationexperiments and in vivo assessment of cross-seeding pro-pensities of hα-Syn and mα-Syn showed that the observed cross-species interactions attenuate seeding and spreading of aggre-gates. These findings possibly explain why current rodent PDmodels expressing hα-Syn do not exhibit pronounced hα-Synfibrillization, and provide models that reproduce a criticalpathological feature of the disease: the de novo formation offibrillar hα-Syn aggregates.

ResultsInduction of hα-Syn Inclusion Formation in SNCA−/− Primary Neurons.Several studies have reported that overexpressed hα-Syn inmouse primary neurons exhibits diffuse localization without formingdiscreet inclusions (8–11). To assess whether the absence of mα-Synwould affect this distribution, we examined the localization oftransiently expressed hα-Syn in primary neurons derived fromC57Bl6/J/ola/hsd mice (SNCA−/−) lacking expression of mα-Syn(Fig. S1 A and B) (12). Interestingly, although hα-Syn exhibited thepreviously reported diffuse distribution in WT neurons, a significantproportion of the SNCA−/− neurons developed spheroid-like in-clusions that were consistently observed in close association with

nuclear and cytoplasmic membranes (Fig. 1A and Movie S1). Asimilar phenotype was observed in primary neurons derived fromanother SNCA−/−mouse strain (B6,129 × 1-Sncatm1Rosl/J) generatedby targeted deletion of two exons of the mα-Syn gene (Fig. S1 Cand D) (13), thereby establishing that the formation of inclusionsis directly linked to the specific loss of mα-Syn expression. In linewith this finding, re-expression of mα-Syn in SNCA−/− neuronssignificantly reduced the formation of intraneuronal hα-Syn in-clusions, and restored the diffuse distribution of hα-Syn in mostneurons (Fig. 1B).To investigate whether the total abolishment of mα-Syn ex-

pression is necessary to promote hα-Syn inclusion formation, weassessed whether decreasing mα-Syn levels in WT neurons viashRNA-mediated silencing is sufficient to promote this process.Transient expression of three different vectors encoding shRNAhairpin loops showed efficient silencing of mα-Syn in transfectedneurons as compared to neurons transfected with scrambledshRNA sequences (Fig. S1E), and biochemical analysis of lenti-virally infected neurons showed ∼50% reduction in mα-Syn ex-pression (Fig. S1F). Notably however, mα-Syn silencing did notpromote the formation of hα-Syn inclusions in WT neurons (Fig.S1G), suggesting that even low levels of mα-Syn are sufficient toattenuate inclusion formation.

Fig. 1. The absence of mα-Syn promotes hα-Syn aggregation in primary neurons. (A) hα-Syn exhibits diffuse distribution in WT neurons and forms inclusions(>1 μm) in cell bodies and neurites in a significant proportion of SNCA−/− neurons. β3-Tubulin staining was performed to reveal neurons, and merged imagesshow orthogonal Z-stack projections (Orth. Proj.). (B) Significantly fewer inclusions (>1 μm) are observed in SNCA−/− neurons with restored mα-Syn expression.(C) Significantly less monomeric hα-Syn is detected in nonionic detergent-soluble (Det. Sol.) fractions of lentivirally infected SNCA−/− neurons than in WTcounterparts (n = 4 independent experiments). In contrast, nonionic detergent-insoluble (Det. Insol.) fractions of infected SNCA−/− neurons show moremonomeric and HMW hα-Syn species, comparable to the detergent-insoluble fractions of WT neurons treated with hα-Syn PFFs. Actin was used to control for equalprotein loading. In B and C, hα-Syn and mα-Syn were differentially revealed using the Syn-211 and D37A6 antibodies, respectively. In A–C, the Mann–Whitney test wasapplied to obtain P values. InA and B, 25 neurons per conditionwere quantified (n= 3 independent experiments). (D) Nine different anti–α-Syn antibodies (epitopes inTable S1) detect inclusions in SNCA−/− neurons transfected with hα-Syn. (E) hα-Syn inclusions in transfected SNCA−/− neurons are ThS-positive, phosphorylated at S129,and ubiquitinated. Colocalization was confirmed by assessing regression coefficients (R2). (Scale bars: 10 μm in A and B and 5 μm in D and E.)

2 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1512876113 Fares et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

hα-Syn Inclusions Reproduce Key LB Features and Exhibit FibrillarUltrastructure. Several cellular models manifesting α-Syn accumu-lation into inclusions in human cell lines have been reported, uponα-Syn overexpression alone (14–16), coexpression with synphilin-1(17), exposure to proteasome inhibitors (9), or application ofoxidative/nitrative insults (18, 19). Nevertheless, inclusions observedunder these conditions typically do not reproduce all key LBfeatures, including decreased solubility, hyperphosphorylation atSerine-129 (pS129) (20), ubiquitination (21), thioflavin-S (ThS)binding, and fibrillar ultrastructural organization (1). Therefore,we assessed whether the hα-Syn inclusions observed in SNCA−/−

neurons fulfill these pathological features.To determine whether the formation of inclusions is linked to

decreased hα-Syn solubility, WT or SNCA−/− neurons were infectedwith lentiviruses encoding hα-Syn and were then fractionated intononionic detergent-soluble and nonionic detergent-insolublefractions. Using this assay, lentivirally infected SNCA−/− neuronsshowed significantly less monomeric hα-Syn in detergent-solublefractions compared to WT counterparts (Fig. 1C). Interestingly,this effect was not caused by decreased total hα-Syn expression inSNCA−/− neurons, because similar total hα-Syn RNA and pro-tein levels were observed in unfractionated WT and SNCA−/−

neurons (Fig. S2 A and B). In contrast, the decrease in solublemonomeric hα-Syn was concomitant with the appearance ofhigh-molecular-weight (HMW) hα-Syn species in detergent-insoluble fractions of lentivirally infected SNCA−/− neurons (Fig.1C). To validate our fractionation protocol, we treated WTneurons with insoluble hα-SynAlexa633 PFFs (characterized in Fig.S3), which showed a strong signal almost exclusively within thedetergent-insoluble fractions (Fig. 1C), as previously reported(22). Taken together, these results suggest that abolishing mα-Synexpression enhances the aggregation propensity of hα-Syn in pri-mary neurons, as reflected by decreased solubility and the forma-tion of insoluble HMW hα-Syn aggregates.

One possibility is that the inclusions in SNCA−/− neuronsrepresent aggregated hα-Syn that is contained within vesiclesof the endo-lysosomal pathway. Therefore, we systematicallyassessed the colocalization of inclusions with specific markers forearly endosomes [Ras-related protein (Rab) 5 and Ras homologgene family member B (RhoB)], late endosomes (Rab7), lyso-somes [lysosomal-associated membrane protein 1 (Lamp1)],autophagosomes [microtubule-associated proteins 1A/1B lightchain 3 (LC3)], and multilamellar bodies [CD63 antigen(CD63)]. As shown in Fig. S4A, although some hα-Syn inclu-sions showed weak partial colocalization with cotransfectedmarkers of late endosomes (Rab7-GFP) and lysosomes(Lamp1-RFP) indicating potential degradation by this path-way, most of the inclusions showed no colocalization with anyof the cotransfected markers.Next, we assessed whether the inclusions formed in SNCA−/−

neurons are readily detectable using common LB probes, namely,antibodies against α-Syn (1), pS129–α-Syn (20), and ubiquitin(21), as well as thioflavin-S (23). As shown in Fig. 1D, immu-nostaining using nine different anti–α-Syn antibodies with epi-topes spanning most of the α-Syn sequence all revealed brightspheroid intraneuronal inclusions, thereby establishing that thesestructures indeed comprise full-length hα-Syn. Moreover, dualimmunofluorescence analysis showed that the inclusions arealso hyperphosphorylated at S129 and are ubiquitinated, andcostaining with thioflavin-S indicated that the inclusions couldcomprise cross–β-sheet fibrillar content (Fig. 1E).To establish whether the hα-Syn inclusions comprise the

characteristic fibrillar ultrastructure observed in LBs, we per-formed correlative light and electron microscopy (CLEM) whichallows ultrastructural examination of neurons previously identi-fied by confocal microscopy as containing fluorescent inclusions(Fig. 2 A–C). Seventy-two hours posttransfection, hα-Syn in-clusions were detected in close association with nuclear and

Fig. 2. Ultrastructure of hα-Syn inclusions in SNCA−/− neurons. (A) For CLEM analysis, a stained neuron was imaged by confocal microscopy and then wasreprocessed and imaged by TEM, resulting in a high-resolution 3D acquisition of the neuron. For illustration, the nucleus is rendered in blue, the cytosol inpurple, and inclusions in yellow. (Scale bars: 5 μm.) (B) To correlate structures observed by ssTEM and fluorescence, single-plane images were superimposed.High-resolution images of inclusions (red boxes 1 and 2) are shown in the lower panel, with arrowheads indicating filamentous structures. (Scale bars: 5 μm inthe upper panel and 2 μm in the lower panel.) (C) Same analysis as in B on different confocal and ssTEM planes from the same neuron. (D) Most filamentsobserved by ssTEM (n = 473) have a width ranging from 7–13 nm. The frequency histogram and normal curve (in red) are shown, as well as values of mean ±SD. (E) Immunogold labeling of SNCA−/− neurons expressing hα-Syn with the ab-6176 anti-synuclein antibody establishes that filaments are positive forsynuclein. Different magnifications of the inclusion (blue arrowhead) and filaments (green arrowheads) are shown. (Scale bar: 1 μm in the left panel and125 nm in the middle and right panels.) (F) A 3D model of the inclusion (blue arrowhead in E) reveals a mesh of intertwining filamentous structures (in purple)with gold particles reacting with the inner and outer portions. The nucleus is rendered in blue, the inclusion in yellow, and immunogold particles in green.Different views of the inclusion are shown.

Fares et al. PNAS Early Edition | 3 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

cytoplasmic membranes or near mitochondria but were rarely foundsurrounded by a uniform lipid bilayer, thereby further ruling outthe possibility that the inclusions are vesicular in nature. Re-markably, high-magnification ssTEM of the correlated inclusionsrevealed intertwining whirls of filamentous structures in mostanalyzed inclusions. Individual measurements of these structuresshowed that the filaments have an average diameter rangingfrom 7–13 nm (Fig. 2D), which is similar to that of fibrils withingenuine LBs (24), and thereby suggests that hα-Syn could beaggregating into fibrils within these inclusions.To rule out the possibility that these flexible filamentous structures

could represent membranes that have been entrapped within inclu-sions, we performed an additional CLEM experiment in which wetreated neurons exhibiting de novo aggregates (composed ofMyc–hα-Syn) with hα-SynAlexa633 PFFs for 24 h and then,within adjacent neurons, compared the ultrastructure and diameterof de novo filaments (comprising only Myc–hα-Syn), internalizedPFFs, mixtures of both, and intracellular membranes (Fig. S4 B andC). As shown in Fig. S4 D and E, although intracellular membranesshowed a broad distribution of diameters (7–12 nm with a meanaverage of 8.7 nm), internalized PFFs and mixtures of de novoformed filaments and PFFs showed a much more defined widthdistribution of 10–13 nm and a higher mean diameter of ∼11 nm.Interestingly, although filaments in de novo inclusions exhibiteda broader width distribution compared to PFFs and mixtures(7–13 nm), with some of the filaments having a low diameter thatcould classify them as membranous, the great majority (∼75%)were in the range of 10–13 nm, within the narrow distribution ofPFFs.To determine whether hα-Syn is the constituent of the ob-

served fibrillar structures, we sought to assess their reactivitywith antibodies against α-Syn by immunogold labeling. First, weprobed the ability of four different antibodies that have differentepitopes against the N terminus, C terminus, and non-Abetacomponent (NAC) region of α-Syn to detect hα-Syn PFFs. No-tably, we found that polyclonal anti–PAN-Syn antibodies such asab-6176 and FL-140 detect PFFs much more readily thanmonoclonal hα-Syn–specific antibodies (Fig. S5A). Therefore, weassessed the reactivity of the fibrillar structures observed withinhα-Syn–expressing SNCA−/− neurons to the ab-6176 antibody.Importantly, 3D reconstruction of ssTEM images clearly showedimmunogold particles reacting with outer and inner portionsof the filaments detected within inclusions (Fig. 2 E and F), therebyconfirming the presence of synuclein within these structures.Taken together, these findings demonstrate that the inclusions

observed in SNCA−/− neurons exhibit similarities to LBs in terms ofsolubility and immunoreactivity, and demonstrate signs of bona fidefibrillization events. However, the filaments do not have an orderlyarrangement with a dense core, as typically observed in brainstemLBs, but are embedded among an electron-translucent medium,which is probably composed of sequestered soluble α-Syn protein.Because the filaments are not organized as in authentic LBs, theymay recapitulate early stages of LB formation that may needmore time or additional factors to mature and remodel into moreorganized LB-like structures.

hα-Syn Inclusions in SNCA−/− Neurons Grow in Size and IncorporateSoluble α-Syn. We then investigated mechanisms underlying hα-Syn inclusion growth over time using fluorescence microscopy. InSNCA−/− neurons transiently expressing hα-Syn, inclusions weredetected 24, 48, and 72 h posttransfection, and the size of in-clusions appeared to increase over time as assessed by quantifying3D-rendered inclusion volumes (Fig. 3A). To investigate the ki-netics of inclusion growth in real time, we performed fluorescencelive-imaging experiments using hα-Syn that is N-terminally tagged tomEOS2, a fluorescent protein that is photoconvertible from brightgreen (506 nm) to orange-red (584 nm) when excited at near-UVwavelengths (25). As shown in Fig. 3B, somatic and neuritic

mEOS2–α-Syn inclusions were detected 24 h posttransfectionin SNCA−/− neurons and continued to grow in volume over72 h. Moreover, the detected inclusions were reactive with antibodiesagainst total α-Syn, ubiquitin, and pS129–α-Syn (Fig. 3C), suggestingthat the fusion protein reproduces the aggregation properties ofuntagged hα-Syn and could be used to investigate the kinetics ofinclusion formation in neurons.Short-interval (∼1 h) confocal live imaging followed by 3D

surface rendering showed that the volume of individual inclu-sions increased over time, with very few reversible fusion eventsnoted (Fig. 3D). Moreover, analysis of the speed of movingparticles showed that most detected inclusions are immobile(Fig. 3D), thereby ruling out the possibility of inclusion growthbeing caused mainly by the fusion of small, motile aggregates.Notably, the percent increase in volume was variable amongdifferent inclusions, with some doubling their volume within50 min and others showing modest changes. This observationsuggests that the growth rate of inclusions is not linear and constantover time, and is not necessarily similar for the whole population ofinclusions within the same neuron.To determine whether mEOS2–α-Syn inclusions incorporate

soluble hα-Syn over time, we performed fluorescence recoveryafter photobleaching (FRAP) and photoconversion live-imagingexperiments (Fig. 3E), which allow assessing the rates of proteindiffusion into (DIN) and out of (DOUT) inclusions respectively,and determining the relative proportion of immobile proteinwithin inclusions (immobile fraction, IF). In FRAP experimentswe photobleached individual inclusions and monitored the re-covered fluorescence from diffusing soluble cytosolic protein. Asshown in Fig. 3F and Movie S2, FRAP measurements showedthe recovery of fluorescent signal within photobleached inclu-sions, and monoexponential fitting of FRAP plots from ∼90different inclusions allowed the estimation of DIN and IF. Theaverage value for DIN was very low (∼0.03 μm2/s), consistent withprevious results for immobilized α-Syn–tetracysteine inclusionsin SH-SY5Y cells (0.03–0.04 μm2/s) (26), suggesting the presenceof a structure that impedes free diffusion of soluble hα-Syn. Inline with this notion, ∼32% of the protein was estimated to befound within the IF, thereby establishing that inclusions com-prise immobilized proteins that are unable to equilibrate with thecytoplasmic pool, likely because they are bound or sequestered ina dense compact structure. The presence of high amounts ofsoluble mobile protein (∼68%) within inclusions is in line withour ssTEM data showing filamentous structures embeddedwithin a bulk of electron-translucent milieu, which is probablysequestered soluble hα-Syn protein.To assess the DOUT, we used the photoconvertible property of

mEOS2, where we photoconverted individual inclusions by near-UV irradiation and monitored the decrease in photoconvertedfluorescence intensity within inclusions. The decay in fluores-cence would reflect outward diffusion of proteins from the in-clusions to the cytosol. As expected, irradiation of inclusionswith a 405-nm laser resulted in rapid photoconversion of 488-nmmEOS2–α-Syn fluorescence to 568 nm (Fig. 3G and Movie S3).Strikingly, although exponential fitting of decay curves from ∼50different neurons revealed a slightly lower estimation of the IFthan obtained from FRAP experiments (16 ±10%), the esti-mated DOUT (0.0022 ± 0.0007 μm2/s) was almost 16-fold lowerthan the DIN (Fig. 3H). This result suggests that, within the sametime period, significantly more protein enters into inclusionsthan diffuses out. As such, our findings together suggest thatthe hα-Syn aggregates in SNCA−/− neurons incorporate solubleprotein into stable structures, most likely the fibrils observedby ssTEM.

Inclusion Formation in SNCA−/− Neurons Is an Aggregation-DrivenProcess. To establish whether the formation of hα-Syn inclusionsis an aggregation-driven process, we assessed whether altering

4 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1512876113 Fares et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

the propensity of hα-Syn for aggregation affects the formationand/or growth of neuronal inclusions. Two modified hα-Synproteins, a C-terminally truncated variant (at position 120) thathas been shown to aggregate more readily than full-lengthα-Syn in vitro and in vivo (27), and a variant with the S87Esubstitution that inhibits hα-Syn aggregation in similar models(28, 29), were expressed in WT or SNCA−/− neurons. As shownin Fig. 4A, although similar proportions of SNCA−/− neuronsexpressing full-length (α-SynFL) or truncated (α-SynΔ120) hα-Syndeveloped inclusions, the S87E aggregation-deficient variantexhibited mostly diffuse distribution. This effect was not caused by adifference in protein expression levels among the three proteins,because similar mean fluorescence levels per neuron were detectedacross conditions (Fig. S5B). As expression of the α-SynΔ120 mutantdid not affect the number of neurons forming inclusions, we in-vestigated whether the kinetics of mEOS2–α-Syn inclusion growthare affected by this modification (Fig. 4B). Surface reconstruction ofinclusions showed that at 24 h posttransfection mEOS2–α-SynΔ120formed significantly larger inclusions than mEOS2–α-SynWT. In con-trast, the few inclusions formed by the mEOS2–α-SynS87E mutantwere much smaller compared to mEOS2–α-SynWT inclusions at 48 h.Furthermore, we investigated whether treatment with a phar-

macological α-Syn aggregation inhibitor, tolcapone (30), wouldaffect the formation of inclusions in this model. Remarkably,tolcapone treatment reduced the formation of inclusions by ∼60%and restored the diffuse distribution of hα-Syn in SNCA−/− neurons(Fig. 4C). Together, these results demonstrate that the formation

of hα-Syn inclusions in SNCA−/− neurons is an aggregation-driven process, as inclusion formation was significantly inhibitedboth by the introduction of a single amino acid substitution knownto inhibit hα-Syn’s aggregation propensity, and by treatment with aknown aggregation inhibitor.

Endogenous β-Syn KO Is a Natural hα-Syn Aggregation Inhibitor inNeurons. Our results suggest that homologous members of thesynuclein family could act as natural physiological inhibitors ofabnormal α-Syn aggregation. To test this hypothesis, we com-pared hα-Syn’s propensity for aggregation in primary neuronsderived from mice lacking expression of β-Syn (β-Syn KO) (31),γ-Syn (γ-Syn KO) (32), or all three synucleins (triple KO) (33).Loss of α- and β-Syn in corresponding KO cultures was validatedby immunocytochemistry (Fig. 5A); γ-Syn expression was notdetectable in any of the cultures, probably because of its lowexpression levels in hippocampal neurons, as previously reported(34). Remarkably, when we assessed the distribution of exoge-nously expressed hα-Syn across the different cultures (Fig. 5B),we found that β-Syn–KO and triple-KO neurons similarly de-velop α-Syn inclusions, unlike WT controls and γ-Syn–KO neu-rons that mostly exhibit diffuse hα-Syn distribution. Importantly,the inclusions formed in β-Syn–KO and triple-KO neurons werealso hyperphosphorylated at S129, thus exhibiting a key patho-logical LB-like phenotype (Fig. 5C). These findings suggest that,unlike endogenous γ-Syn, the other two homologs, α-Synand β-Syn, have similar inhibitory effects on ectopic hα-Syn

Fig. 3. hα-Syn inclusions in SNCA−/− neurons grow and incorporate soluble hα-Syn. (A) Immunofluorescence and 3D rendering of inclusions reveals asignificant increase in inclusion volume at 48 and 72 h posttransfection in 25 MAP2+ neurons per condition. (B) Similarly, mEOS2–α-Syn inclusions (n =50 per condition) show a significant increase in inclusion volume 72 h posttransfection. (C) mEOS2–α-Syn inclusions are immunopositive for total α-Syn (Syn-1),pS129–α-Syn (WAKO), and ubiquitin. (D) mEOS2–Syn inclusions are immobile and grow in volume during ∼1 h of live imaging. (Scale bars: 5 μm in A–C and 2 μm inD.) (E) FRAP and photoconversion experiments allowmonitoring rates of Din and Dout of inclusions, respectively. (F and G, Upper) Fluorescence before, during, andafter inclusion photobleaching (F) or photoconversion (G) is shown. Higher magnification images of yellow boxes are shown, with crossed circles denoting areas ofphotobleaching (F) or photoconversion (G). (Lower) Average plots (± SD) of normalized FRAP or photoconversion recovery curves (n = 87 or 63 inclusions, re-spectively). (Scale bars: 2 μm at lower magnification and 1 μm at higher magnification.) Ch, channel. (H) Values for immobile fractions and diffusion coefficientsobtained from photoconversion curves (n = 50 inclusions) are significantly lower than those obtained by FRAP. In A, B, and H, the Mann–Whitney test was appliedto obtain P values.

Fares et al. PNAS Early Edition | 5 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

aggregation. Our observation that similar percentages ofβ-Syn–KO and triple-KO neurons develop inclusions (∼35%)rules out the presence of a synergistic effect for the deletion ofboth proteins.

The Absence of mα-Syn Promotes Specific Aggregation of hα-Syn inVivo. To test whether the presence of mα-Syn affects the aggregationpropensity of hα-Syn in the brain, we crossed mice lacking endoge-nous mα-Syn (SNCA−/−) (13) with mice carrying two copies of hα-Syn (hα-Syn+/+) (Fig. S6A) (35) to generate (hα-Syn+/+SNCA−/−)mice. Biochemical fractionation of brain homogenates intocytosolic (soluble) and particulate (insoluble) protein fractions wasperformed to assess whether these mice exhibit changes in hα-Synaggregation and solubility, compared to hα-Syn+/+ controls. In linewith our previous data, less monomeric hα-Syn was detected in cy-tosolic fractions derived from hα-Syn+/+SNCA−/− mice than in cy-tosolic fractions derived from hα-Syn+/+ counterparts (Fig. 6A). Thisdecrease in monomeric hα-Syn was concomitant with the appear-ance of HMW hα-Syn species within particulate fractions, which

were detected using three different antibodies against hα-Syn,thereby establishing their specificity (Fig. 6A and Fig. S6B).Moreover, the decrease in soluble hα-Syn was not the result ofdecreased total hα-Syn expression in hα-Syn+/+SNCA−/− micecompared to hα-Syn+/+ counterparts, as similar total hα-SynmRNA and protein levels were observed in unfractionatedbrains of both genotypes (Fig. S6 C and D). These findingssuggest that hα-Syn aggregates readily and becomes lesssoluble in the absence of mα-Syn in Tg mouse brains.To evaluate further the extent of hα-Syn aggregation in mouse

brains, we performed immunohistochemical analyses on sectionstreated with proteinase K, which allows selective visualization ofhighly structured proteinase K resistant aggregates. Consistentwith previous studies (35), hα-Syn+/+ mice showed abundantintraneuronal α-Syn immunoreactive inclusions in the cortex,hippocampus, and striatum, some of which were proteinase Kresistant (Fig. 6B). Strikingly, neurons from hα-Syn+/+SNCA−/−

mice showed higher immunoreactivity to anti–α-Syn antibodies andexhibited a significant increase in proteinase K-resistant inclu-sions in the cortex and hippocampus (Fig. 6B). Moreover, doubleimmunolabeling showed that these inclusions are phosphory-lated at S129 and exhibit strong reactivity for the synapticvesicle protein synaptophysin (Fig. 6 C and D), both of whichare markers of LB pathology (20, 36). Taken together, thesefindings demonstrate that the aggregation propensity of hα-Synis enhanced in the absence of mα-Syn in vivo, as reflected bydecreased hα-Syn solubility and accumulation into proteinaseK-resistant inclusions.To assess whether abolishing mα-Syn expression would pro-

mote the aggregation of human β-Syn (hβ-Syn), the considerablyless aggregation-prone homolog of hα-Syn, we crossed Tg miceexpressing hβ-Syn (37) with SNCA−/− mice to generate hβ-Syn+/+

SNCA−/− mice. Biochemical fractionation revealed that the ab-sence of mα-Syn did not affect the levels of soluble hβ-Syn orcause the appearance of insoluble hβ-Syn species (Fig. S6E). In

Fig. 4. hα-Syn inclusion formation in SNCA−/− neurons reflects an aggregation-driven process. (A) Significantly fewer SNCA− /− neurons expressingα-SynS87E develop inclusions (>1 μm) than their α-SynFL and SynΔ120

counterparts (mean ± SD, n = 3 independent experiments with 25 neuronsper condition). (B) Live imaging and 3D rendering of inclusions (50 neuronsper condition) show that mEOS2–α-SynΔ120 forms significantly larger inclu-sions 24 h posttransfection than mEOS2–α-SynWT, and that the mEOS2–α-SynS87E

mutant forms smaller inclusions at 48 h in SNCA−/− neurons. (C) Significantlyfewer tolcapone-treated SNCA−/− neurons expressing hα-Syn (25 neurons percondition) develop inclusions (>1 μm) than DMSO-treated controls (mean ± SD,n = 3 independent experiments). In A the one-way ANOVA and Scheffé post hocanalysis were applied, and in B and C the Mann–Whitney test was applied toobtain P values. Orth. Proj., orthogonal projection. (Scale bars: 5 μm.)

Fig. 5. Endogenous β-Syn naturally acts as an inhibitor of hα-Syn aggre-gation in primary neurons. (A) Immunocytochemistry validates the loss ofα- and β-Syn expression in α-Syn–KO, β-Syn–KO, and triple Syn–KO neuronscompared to WT neurons. (B) Significantly more β-Syn–KO neurons andtriple-KO neurons (one-way ANOVA and Scheffé post hoc analysis) developinclusions (>1 μm in diameter) compared to non-Tg neurons (mean ± SD, n =3 independent experiments, 20 neurons per condition). (C) Inclusions formedin β-Syn–KO and triple Syn–KO neurons are pS129–α-Syn–positive. Colocali-zation was confirmed by assessing regression coefficients (R2). (Scale bars:20 μm in A and 5 μm in B and C.)

6 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1512876113 Fares et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

addition, immunohistochemical analysis showed that hβ-Syn waslocalized mostly within neuronal terminals and did not form anyinclusions in the brains of hβ-Syn+/+ mice, as previously reported(37), or in the brains of hβ-Syn+/+SNCA−/− mice (Fig. S6F).These results show that α-Syn ablation does not promote theaggregation of hβ-Syn in vivo.

mα-Syn Interacts with Aggregated hα-Syn Species and AttenuatesSeeding and Spreading. To investigate the mechanism by which mα-Syn affects hα-Syn aggregation, we first assessed whether overex-pressed hα-Syn interacts directly with endogenous mα-Syn in WTprimary neurons. Notably, we did not detect any interaction be-tween the two proteins at the monomer level by coimmunopre-cipitation (Fig. S7A). To investigate whether mα-Syn interacts

with multimeric/aggregated hα-Syn species rather than hα-Synmonomers, we assessed the interaction of mα-Syn monomers withequimolar amounts of hα-Syn monomers or hα-Syn PFFs (charac-terized in Fig. S8). Intriguingly, mα-Syn was coimmunoprecipitatedonly when it was mixed with hα-Syn PFFs and not with hα-Synmonomers (Fig. 7A), indicating that monomeric mα-Syn might in-teract preferentially with hα-Syn PFFs in vitro. To confirm thesefindings and to obtain a quantitative assessment of the interactionbetween mα-Syn monomers and hα-Syn PFFs by an independentreadout, we performed SPR measurements. Briefly, saturatingamounts of either hα-Syn monomers or hα-Syn PFFs were immo-bilized by amine coupling to separate CM5 chips, and then increasingconcentrations of mα-Syn monomers were flushed onto the chips. Asshown in Fig. 7B, and in line with our immunoprecipitation experi-ments, no interaction was noted between injected mα-Syn monomersand immobilized hα-Syn monomers, even at the highest testedconcentration for mα-Syn (10 μM). In contrast, mα-Syn mono-mers interacted readily with sonicated hα-Syn PFFs in a dose-de-pendent manner, even at concentrations as low as 0.1 μM.To investigate whether endogenous mα-Syn similarly interacts

with aggregated hα-Syn species in primary neurons, we treated WTneurons with sonicated hα-Syn PFFs and assessed interactionby immunoprecipitation and immunofluorescence analyses. Impor-tantly, and in line with our in vitro results, mα-Syn coimmunopre-cipitated with pulled-down hα-Syn PFFs (Fig. S7B). Moreover,confocal imaging revealed that internalized hα-Syn PFFs colocalizewith endogenous mα-Syn (Fig. S7C), which redistributes from adiffuse somatic localization into defined puncta as previously de-scribed (22). Together, these results show that mα-Syn interactspreferentially with aggregated hα-Syn PFFs in vitro and in neurons.Next, we explored whether such an interaction between α-Syn

PFFs and α-Syn monomers from different species would seed or,alternatively, attenuate progressive monomer aggregation. To do so,we assessed the aggregation kinetics of hα-Syn or mα-Syn mono-mers in the presence of either hα-Syn PFFs or mα-Syn PFFs(characterized in Fig. S8). Strikingly, assessment of thioflavin-T(ThT) binding and the formation of fibrillar structures by TEMshowed that the seeding efficiency is significantly higher when hα-Syn monomers and PFFs are of the same species (Fig. 7C and Fig.S9 A and B). For instance, whereas the lag phase was eliminated inthe mixture of hα-Syn monomers and hα-Syn PFFs, resulting in theappearance of fibrillar structures at 8 h of incubation, the aggre-gation profile of the mixture of hα-Syn monomers and mα-SynPFFs was similar to that of hα-Syn alone, in which only oligomericspecies were formed at 8 h of incubation. Interestingly, the mixtureof mα-Syn monomers with mα-Syn PFFs showed the most dramaticelimination of the lag phase, with fibrils formed after only 2 h ofincubation. Here again, although the addition of hα-Syn PFFs tomα-Syn monomers resulted in higher ThT plateau values than seenwith mα-Syn monomers alone, it failed to eliminate the lag phaseand resulted in the formation of oligomeric species at 2 h. Takentogether, these findings strongly show that the aggregation kineticsof α-Syn are significantly accelerated when the seeds and mono-mers are from the same species, and that the presence of mixedspecies of monomers and PFFs results in reduced seeding capacityand slower aggregation kinetics.To evaluate whether the seeding and spreading potential of α-Syn

PFFs is similarly affected by the presence of PFFs and endogenousmonomers of the same species in the brain, we injected mα-SynPFFs or hα-Syn PFFs into the striatum of WT (non-Tg) or SNCA−/−

mice (as controls) and then compared the seeding and spreadingpropensity in different brain regions 1 month postinjection. Strik-ingly, we observed the most pronounced spreading of α-Syn pa-thology only when mα-Syn PFFs were injected into WT miceexpressing endogenous mα-Syn (Fig. 7D); in these mice mα-Synaccumulated into somatic inclusions that were ubiquitinatedand hyperphosphorylated at S129 in the cortex and amygdala(Fig. S9C). In contrast, hα-Syn PFFs induced significantly less

Fig. 6. The absence of mα-Syn promotes hα-Syn aggregation in transgenicmice. (A) Significantly less monomeric hα-Syn is detected in cytosolic frac-tions from hα-Syn+/+SNCA−/− mice than in hα-Syn+/+ counterparts (n = 3 in-dependent experiments), concomitant with the appearance of HMW hα-Synspecies in particulate fractions. Loss of mα-Syn expression in SNCA−/− micewas verified using the mα-Syn–specific antibody D37A6, and actin was usedto control for equal protein loading. The asterisk indicates a nonspecificband in cytosolic fractions. (B) Significantly more proteinase K (PK)-resistanthα-Syn inclusions/0.1 mm2 are detected in the neocortex and hippocampus(Hipp.) of hα-Syn+/+SNCA−/− mice than in hα-Syn+/+ mice (n = 3 mice percondition). (Scale bars: 100 μm in overviews and 15 μm in insets.) In A and Bthe Mann–Whitney test was applied to obtain P values. (C and D) Double-positive hα-Syn inclusions for total α-Syn and pS129 α-Syn (C) or synapto-physin (D) are detected in the cortex, striatum, and hippocampus ofhα-Syn+/+SNCA−/− mice. Orthogonal (Orth.) Z-stack projections verify theintracellular nature of inclusions. (Scale bars: 5 μm.)

Fares et al. PNAS Early Edition | 7 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

pathology at these regions in WT mice at the same time point,and SNCA−/− mice showed very limited overall pathology followinginjection with either hα-Syn PFFs or mα-Syn PFFs. Thesefindings are consistent with the significantly reduced seedingefficiency of hα-Syn PFFs in the presence of monomeric mα-Syn in vitro; this reduced seeding efficiency may allow moretime for natural inhibitory or clearance mechanisms to interferewith the seeding and spreading of exogenous hα-Syn PFFs inmouse brains.

DiscussionThe aggregation of α-Syn into LBs has been directly linked to thepathogenesis of PD (2). The mechanism of formation and theprecise effect of LBs on neuronal physiology and viability re-mains poorly understood, mostly because of the lack of animaland cellular models exhibiting de novo α-Syn fibrillization intoLB-like structures. Although doubling α-Syn expression is suffi-cient to cause PD in humans, attempts to reproduce molecularPD pathology based on α-Syn overexpression in various animalmodels have not been successful. This lack of success could implythe presence of cellular factors that inhibit hα-Syn aggregation inthese models. In this study, we sought to systematically examinewhether endogenous α-Syn homologs attenuate hα-Syn aggregationin mouse neurons. We found that a substantial fraction of SNCA−/−

neurons expressing hα-Syn develop inclusions that exhibit key LB-like features, including S129 hyperphosphorylation, ubiquitination,and ThS reactivity. Importantly, this phenomenon reflected de novohα-Syn fibrillization, as revealed by ssTEM following both CLEMand immunogold-labeling experiments. To the best of our knowl-edge, none of the previously published cellular models that rely

solely on the overexpression of hα-Syn have reported inlcusions thatexhibit these features altogether.Multiple lines of evidence from many of our experiments further

converged to establish that the formation of inclusions in SNCA−/−

neurons is an aggregation-driven process. Overexpression of theaggregation-incompetent mutant S87E or treatment with thepharmacological α-Syn aggregation inhibitor tolcapone significantlyreduced the formation of inclusions. In contrast, overexpression ofthe aggregation-prone truncated hα-Syn variant (1-120) resulted inthe formation of larger inclusions than seen with the WT protein.In all cases in which we observed increased inclusion formation, wealso observed decreased hα-Syn solubility and the appearance ofHMW aggregates, concomitant with the loss of monomers. More-over, time-lapse live imaging coupled with FRAP and photo-conversion experiments showed in real time that the inclusions growin size and readily incorporate soluble hα-Syn protein. Interestinglyhowever, when we explored the toxic potential of these inclusions,we found that the aggregates formed do not provoke increasedtoxicity (SI Discussion and Fig. S10). This finding is consistent withprevious studies suggesting that intracellular hα-Syn aggregatesmay be protective in cellular models (38, 39). Further studies areunderway to determine whether these aggregates exhibit a similarrole under conditions of cellular stress, such as treatment withmediators of oxidative stress or inhibitors of protein degradation,or other molecules known to contribute to PD pathogenesis.Having established that SNCA−/− primary neurons are more

permissive of hα-Syn aggregation in culture, we sought to validatethis finding in vivo. Two previous independent studies had suggestedthat the expression of hα-SynΔ120 or hα-SynA53T in Tg mice lackingmα-Syn expression leads to the formation of fibrillar inclusionswithin the substantia nigra or spinal cord (5, 6). Therefore, we

Fig. 7. mα-Syn interacts with aggregated hα-Syn species and attenuates seeding and spreading. (A) Following immunoprecipitation (IP) of hα-Syn, mα-Synmonomers (M) were detected only in the sample comprising the mixture of mα-Syn M with hα-Syn PFFs, and not with hα-Syn M. The asterisks indicate theheavy and light chains of the antibody used for IP. (B) A dose-dependent response (interaction) was noted by SPR between immobilized hα-Syn PFFs andinjected mα-Syn M, but not between the injected mα-Syn M and immobilized hα-Syn M. (C) Mixtures of α-Syn PFFs and monomers of the same species showsignificantly higher ThT binding at early time points than seen with α-Syn monomers alone or with PFFs of different species (n = 3 independent experiments).Mean ± SD values are shown. *P < 0.05, **P < 0.01, ***P < 0.001 (one-way ANOVA and Scheffé post hoc analysis). (D) Significantly more pS129–α-Syn in-clusions (arrows) are observed in the cingulate cortex and amygdala of WT mice injected with mα-Syn PFFs (n = 5 mice per condition) than in counterpartsinjected with hα-Syn PFFs or in SNCA−/− mice injected with hα-Syn/mα-Syn PFFs, which show diffuse staining (arrowheads). The Mann–Whitney test wasapplied to obtain P values. (Scale bar: 100 μm.)

8 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1512876113 Fares et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

directly compared the aggregation propensity of hα-Syn in brains ofmice expressing or lacking endogenous mα-Syn. Our data showedthat hα-Syn exhibits enhanced aggregation propensity in the absenceof mα-Syn, as reflected by decreased solubility, enhanced detectionof HMW oligomeric species, and the formation of abundant pro-teinase K-resistant and pS129 hyperphosphorylated inclusions. Torule out the possibility that knocking out mα-Syn promotes non-specific aggregation, we conducted similar studies on hβ-Syn, theα-Syn homolog that is significantly less prone to aggregate. Asexpected, knocking out mα-Syn did not affect the levels, distribution,or staining pattern of hβ-Syn in Tg mice.Our observation that the reintroduction of mα-Syn in SNCA−/−

neurons significantly attenuates the formation of hα-Syn inclusionssuggests that the mouse homolog might directly inhibit hα-Syn ag-gregation. Intriguingly, although we could not detect a prominentinteraction between hα-Syn and mα-Syn at the monomeric level inneurons, we consistently observed that monomeric mα-Syn prefer-entially interacts with sonicated PFFs in vitro and in neurons.Moreover, we found that this interaction does not promote pro-gressive seeding and spreading of pathology in vitro or in mousebrains in vivo. As such, endogenous mα-Syn might directly inhibithα-Syn fibrillization in WT primary neurons by stabilizing oligo-meric species and capping the terminals of early fibrillar structuresformed by hα-Syn. This hypothesis is in line with the early study byRochet et al. (7) showing that mα-Syn inhibits the fibrillization ofhα-Syn by stabilizing oligomers on the pathway to amyloid forma-tion in vitro. Moreover, it is consistent with literature reportsdemonstrating that the propagation and spreading of α-Syn pa-thology is prominent when the injected PFFs and host-expressedmonomers are from the same species. In most of these studies,seeding and spreading of α-Syn pathology was observed when mα-Syn PFFs were injected into non-Tg mice expressing mα-Syn (40–42) or when hα-Syn PFFs were injected into Tg mice expressinghuman α-Syn protein (43–48).An attractive general implication of our findings is that the

attenuation of aggregation by homologous proteins could representa generic biological phenomenon. This proposition is consistentwith multiple studies reporting decreased aggregation kinetics in thepresence of homologous amyloidogenic proteins underlying differ-ent diseases. For instance, the key peptides involved in the forma-tion of amyloid plaques in Alzheimer’s disease, Aβ40 and Aβ42,were shown to inhibit each other’s aggregation reciprocally in vitroin a concentration-dependent manner (49). Moreover, mutant he-moglobin β-chain polymerization in sickle cell anemia was shown tobe attenuated upon increasing ratios of its γ-chain homolog (50).Likewise, in transthyretin (TTR) amyloid diseases, murine/humanTTR homologs form heterotetramers that impair amyloid forma-tion (51, 52), and aggregation in vivo was achieved mostly in Tgmice expressing TTR mutants on an endogenous TTR-KO back-ground (52). Finally, the aggregation kinetics of the islet amyloidpolypeptide (IAPP) in bulk solution was reduced dramatically bythe presence of its rat homolog (53, 54).In light of these findings, we investigated whether this phe-

nomenon could be extended to the other endogenous hα-Syn ho-mologs, β-Syn and γ-Syn. Interestingly, we found that β-Syn–KOcultures also developed hα-Syn inclusions that were hyper-phosphorylated at S129. In contrast, the distribution of hα-Synwas not affected when expressed in γ-Syn–KO neurons, be-cause the levels of γ-Syn in WT hippocampal neurons werealready very low and therefore could not be reduced drasti-cally in γ-Syn–KO neurons. These results are also in line withprevious reports showing that β-Syn inhibits α-Syn aggregationin vitro (55, 56) as well as in vivo; double-Tg mice coex-pressing both hα-Syn and hβ-Syn exhibit ameliorated motordeficits and lower α-Syn accumulation compared to Tg miceexpressing hα-Syn alone (37, 57). However, our results fur-ther suggest that the total levels or ratio of both endogenousα- and β-Syn could be critical in protecting against hα-Syn

aggregation. Once this level drops below a certain threshold (ineither single or triple KOs), the aggregation of hα-Syn into LB-likeinclusions becomes progressive. Supporting this proposition isthe observation that intermediate reduction of mα-Syn levels byshRNA did not promote the formation of hα-Syn inclusions inWT neurons.In summary, we have shown that two endogenous members of

the synuclein family, mα-Syn and mβ-Syn, attenuate the aggrega-tion of overexpressed hα-Syn in mouse neurons. Taken togetherwith the multitude of studies that have reported the inhibition ofaggregation of other amyloidogenic proteins by their homologues(such as Aβ, TTR, hemoglobin, and IAPP), this observation couldreflect a more generic biological phenomenon in which endoge-nous homologs act as natural inhibitors of abnormal aggregation.This phenomenon could explain why most current rodent PDmodels relying on hα-Syn overexpression in WT mice do not showprominent hα-Syn fibrillization. As such, understanding the in-teractions between homologous proteins and the effect of suchinteractions on protein aggregation and amyloid formation couldfacilitate the generation of novel cellular and in vivo models ofprotein misfolding. The ability to assess the dynamics of inclusionformation quantitatively in our neuronal model provides uniqueopportunities to elucidate molecular and cellular determinantsthat influence the mechanisms of α-Syn fibrillization in livingneurons, and to identify pathways that modulate this process andcontribute to neurodegeneration. Further studies using this modelcould make it possible to identify novel drugs for the treatment ofPD based on modulating α-Syn aggregation and interneuronalspreading, or enhancing the degradation and clearance of toxicα-Syn aggregates.

Materials and MethodsFull experimental details are provided in SI Materials and Methods.

Approval of Animal Experimentation. All animal experiments were performedin accordance with guidelines and regulations of the Ecole PolytechniqueFédérale de Lausanne (EPFL), Cardiff University, and University of CaliforniaSan Diego, and were approved by the Swiss Federal Veterinary Office.

Primary Neuron Culture Preparation, Transfection, Infection, and TolcaponeTreatment. Primary hippocampal and cortical neuronal cultures were pre-pared from postnatal day 0 C57BL/6JRccHsd (WT; Harlan Laboratories),C57BL/6JOlaHsd (α-Syn KO; Harlan Laboratories) (12), B6,129 × 1-Sncatm1Rosl/J(α-Syn–KO; Jackson Laboratories) (13), β-Syn–KO (31), γ-Syn–KO (58), orα-β-γ-Syn–KO (33) mice as previously described (10). Briefly, dissected tissueswere dissociated with papain and triturated using a glass pipette. Aftercentrifugation at 400 × g for 2 min, cells were plated in minimum essentialmedium (MEM)/10% (vol/vol) horse serum onto poly-L-lysine (Sigma)–coatedcoverslips [12 mm glass (VWR International) for immunocytochemistry or13 mm plastic (Thermanox, Nunc) for immunogold staining and TEM] at 1.5 ×105 cells/mL, or at 3 × 105 cells/mL in 35-mm confocal dishes (World Precision)for live imaging, or in 10-cm dishes (BD Biosciences) for biochemical analysis.Medium was changed after 4 h to Neurobasal/ B27 medium, and neuronswere treated with Arabinosylcytosine (AraC; Sigma) after 6 d in vitro (DIV) tostop glial division. At 7 DIV, neurons were either transiently transfected forup to 3 d using Lipofectamine 2000 (Invitrogen) according to the manu-facturer’s instructions or were infected with lentiviruses at a multiplicity ofinfection (MOI) of 10 for up to 7 d. When indicated, at 24 h posttransfection,neurons were treated with 20 μM tolcapone (Sigma) for an additional24 h before immunocytochemistry.

ACKNOWLEDGMENTS. We thank Nathalie Jordan and Celine Vocat for helpin preparing primary cultures, plasmid cloning, and production of recombi-nant proteins; Mr. Samuel Kilchenmann for expert help in conducting andanalyzing surface plasmon resonance (SPR) experiments; Dr. Anne-LaureMahul and Dr. Issa Issac for helping with blinded quantification of someexperiments; Mr. Anass Chiki for help with protein characterization; Mr. MarounBou Sleiman for help with statistical analysis; Prof. Virginia Lee for kindly pro-viding the Syn-208 antibody; and Mrs. Dania Al Labban and Dr. Juan Reyes fortheir critical evaluation of themanuscript. This work was supported by the EcolePolytechnique Fédérale de Lausanne (EPFL), the Swiss National Science Foun-dation, and Wellcome Trust Grant 075615/Z/04/z (to V.L.B.).

Fares et al. PNAS Early Edition | 9 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020

1. Spillantini MG, et al. (1997) Alpha-synuclein in Lewy bodies. Nature 388(6645):839–840.2. Lashuel HA, Overk CR, Oueslati A, Masliah E (2013) The many faces of α-synuclein:

From structure and toxicity to therapeutic target. Nat Rev Neurosci 14(1):38–48.3. Dawson TM, Ko HS, Dawson VL (2010) Genetic animal models of Parkinson’s disease.

Neuron 66(5):646–661.4. Feany MB, Bender WW (2000) A Drosophila model of Parkinson’s disease. Nature

404(6776):394–398.5. Cabin DE, et al. (2005) Exacerbated synucleinopathy in mice expressing A53T SNCA on

a Snca null background. Neurobiol Aging 26(1):25–35.6. Tofaris GK, et al. (2006) Pathological changes in dopaminergic nerve cells of the

substantia nigra and olfactory bulb in mice transgenic for truncated human alpha-synuclein(1-120): Implications for Lewy body disorders. J Neurosci 26(15):3942–3950.

7. Rochet JC, Conway KA, Lansbury PT, Jr (2000) Inhibition of fibrillization and accu-mulation of prefibrillar oligomers in mixtures of human and mouse alpha-synuclein.Biochemistry 39(35):10619–10626.

8. Petrucelli L, et al. (2002) Parkin protects against the toxicity associated with mutantalpha-synuclein: Proteasome dysfunction selectively affects catecholaminergic neu-rons. Neuron 36(6):1007–1019.

9. McLean PJ, Kawamata H, Hyman BT (2001) Alpha-synuclein-enhanced green fluo-rescent protein fusion proteins form proteasome sensitive inclusions in primaryneurons. Neuroscience 104(3):901–912.

10. Fares MB, et al. (2014) The novel Parkinson’s disease linked mutation G51D attenuatesin vitro aggregation and membrane binding of α-synuclein, and enhances its secre-tion and nuclear localization in cells. Hum Mol Genet 23(17):4491–4509.

11. Khalaf O, et al. (2014) The H50Q mutation enhances α-synuclein aggregation, se-cretion, and toxicity. J Biol Chem 289(32):21856–21876.

12. Specht CG, Schoepfer R (2001) Deletion of the alpha-synuclein locus in a sub-population of C57BL/6J inbred mice. BMC Neurosci 2:11–19.

13. Abeliovich A, et al. (2000) Mice lacking alpha-synuclein display functional deficits inthe nigrostriatal dopamine system. Neuron 25(1):239–252.

14. Tabrizi SJ, et al. (2000) Expression of mutant alpha-synuclein causes increased sus-ceptibility to dopamine toxicity. Hum Mol Genet 9(18):2683–2689.

15. Opazo F, Krenz A, Heermann S, Schulz JB, Falkenburger BH (2008) Accumulation andclearance of alpha-synuclein aggregates demonstrated by time-lapse imaging.J Neurochem 106(2):529–540.

16. Tofaris GK, Layfield R, Spillantini MG (2001) alpha-synuclein metabolism and aggregation islinked to ubiquitin-independent degradation by the proteasome. FEBS Lett 509(1):22–26.

17. Engelender S, et al. (1999) Synphilin-1 associates with alpha-synuclein and promotesthe formation of cytosolic inclusions. Nat Genet 22(1):110–114.

18. Lee HJ, Shin SY, Choi C, Lee YH, Lee SJ (2002) Formation and removal of alpha-syn-uclein aggregates in cells exposed to mitochondrial inhibitors. J Biol Chem 277(7):5411–5417.

19. Paxinou E, et al. (2001) Induction of alpha-synuclein aggregation by intracellular ni-trative insult. J Neurosci 21(20):8053–8061.

20. Fujiwara H, et al. (2002) alpha-Synuclein is phosphorylated in synucleinopathy lesions.Nat Cell Biol 4(2):160–164.

21. Hasegawa M, et al. (2002) Phosphorylated alpha-synuclein is ubiquitinated in alpha-synucleinopathy lesions. J Biol Chem 277(50):49071–49076.

22. Volpicelli-Daley LA, et al. (2011) Exogenous α-synuclein fibrils induce Lewy body pa-thology leading to synaptic dysfunction and neuron death. Neuron 72(1):57–71.

23. Irwin DJ, Lee VM, Trojanowski JQ (2013) Parkinson’s disease dementia: Convergenceof α-synuclein, tau and amyloid-β pathologies. Nat Rev Neurosci 14(9):626–636.

24. Spillantini MG, Crowther RA, Jakes R, HasegawaM, Goedert M (1998) alpha-Synucleinin filamentous inclusions of Lewy bodies from Parkinson’s disease and dementia withlewy bodies. Proc Natl Acad Sci USA 95(11):6469–6473.

25. McKinney SA, Murphy CS, Hazelwood KL, Davidson MW, Looger LL (2009) A brightand photostable photoconvertible fluorescent protein. Nat Methods 6(2):131–133.

26. Roberti MJ, Jovin TM, Jares-Erijman E (2011) Confocal fluorescence anisotropy andFRAP imaging of α-synuclein amyloid aggregates in living cells. PLoS One 6(8):e23338.

27. Oueslati A, Fournier M, Lashuel HA (2010) Role of post-translational modifications inmodulating the structure, function and toxicity of alpha-synuclein: Implications forParkinson’s disease pathogenesis and therapies. Prog Brain Res 183:115–145.

28. Paleologou KE, et al. (2010) Phosphorylation at S87 is enhanced in synucleinopathies,inhibits alpha-synuclein oligomerization, and influences synuclein-membrane inter-actions. J Neurosci 30(9):3184–3198.

29. Oueslati A, Paleologou KE, Schneider BL, Aebischer P, Lashuel HA (2012) Mimickingphosphorylation at serine 87 inhibits the aggregation of human α-synuclein and protectsagainst its toxicity in a rat model of Parkinson’s disease. J Neurosci 32(5):1536–1544.

30. Di Giovanni S, et al. (2010) Entacapone and tolcapone, two catechol O-methyl-transferase inhibitors, block fibril formation of alpha-synuclein and beta-amyloid andprotect against amyloid-induced toxicity. J Biol Chem 285(20):14941–14954.

31. Chandra S, et al. (2004) Double-knockout mice for alpha- and beta-synucleins: Effecton synaptic functions. Proc Natl Acad Sci USA 101(41):14966–14971.

32. Buchman VL, et al. (1998) Persyn, a member of the synuclein family, has a distinctpattern of expression in the developing nervous system. J Neurosci 18(22):9335–9341.

33. Anwar S, et al. (2011) Functional alterations to the nigrostriatal system in mice lackingall three members of the synuclein family. J Neurosci 31(20):7264–7274.

34. Li J, Henning Jensen P, Dahlström A (2002) Differential localization of alpha-, beta-and gamma-synucleins in the rat CNS. Neuroscience 113(2):463–478.

35. Masliah E, et al. (2000) Dopaminergic loss and inclusion body formation in alpha-synuclein mice: Implications for neurodegenerative disorders. Science 287(5456):1265–1269.

36. Nishimura M, et al. (1994) Synaptophysin and chromogranin A immunoreactivities ofLewy bodies in Parkinson’s disease brains. Brain Res 634(2):339–344.

37. Hashimoto M, Rockenstein E, Mante M, Mallory M, Masliah E (2001) beta-Synucleininhibits alpha-synuclein aggregation: A possible role as an anti-parkinsonian factor.Neuron 32(2):213–223.

38. Ko LW, Ko HH, Lin WL, Kulathingal JG, Yen SH (2008) Aggregates assembled fromoverexpression of wild-type alpha-synuclein are not toxic to human neuronal cells.J Neuropathol Exp Neurol 67(11):1084–1096.

39. Tanaka M, et al. (2004) Aggresomes formed by alpha-synuclein and synphilin-1 arecytoprotective. J Biol Chem 279(6):4625–4631.

40. Luk KC, et al. (2012) Pathological α-synuclein transmission initiates Parkinson-likeneurodegeneration in nontransgenic mice. Science 338(6109):949–953.

41. Masuda-Suzukake M, et al. (2014) Pathological alpha-synuclein propagates throughneural networks. Acta Neuropathol Commun 2:88–99.

42. Paumier KL, et al. (2015) Intrastriatal injection of pre-formed mouse α-synuclein fibrilsinto rats triggers α-synuclein pathology and bilateral nigrostriatal degeneration.Neurobiol Dis 82:185–199.

43. Bétemps D, et al. (2014) Alpha-synuclein spreading in M83 mice brain revealed bydetection of pathological α-synuclein by enhanced ELISA. Acta Neuropathol Commun2:29–42.

44. Luk KC, et al. (2012) Intracerebral inoculation of pathological α-synuclein initiates a rapidlyprogressive neurodegenerative α-synucleinopathy in mice. J Exp Med 209(5):975–986.

45. Sacino AN, et al. (2014) Brain injection of α-synuclein induces multiple proteino-pathies, gliosis, and a neuronal injury marker. J Neurosci 34(37):12368–12378.

46. Peelaerts W, et al. (2015) α-Synuclein strains cause distinct synucleinopathies afterlocal and systemic administration. Nature 522(7556):340–344.

47. Sacino AN, et al. (2014) Intramuscular injection of α-synuclein induces CNS α-synucleinpathology and a rapid-onset motor phenotype in transgenic mice. Proc Natl Acad SciUSA 111(29):10732–10737.

48. Sacino AN, et al. (2013) Induction of CNS α-synuclein pathology by fibrillar and non-amyloidogenic recombinant α-synuclein. Acta Neuropathol Commun 1(1):38–51.

49. Hasegawa K, Yamaguchi I, Omata S, Gejyo F, Naiki H (1999) Interaction between Abeta(1-42) and A beta(1-40) in Alzheimer’s beta-amyloid fibril formation in vitro.Biochemistry 38(47):15514–15521.

50. Eaton WA, Hofrichter J (1995) The biophysics of sickle cell hydroxyurea therapy.Science 268(5214):1142–1143.

51. Hammarström P, Schneider F, Kelly JW (2001) Trans-suppression of misfolding in anamyloid disease. Science 293(5539):2459–2462.

52. Sousa MM, et al. (2002) Evidence for early cytotoxic aggregates in transgenic mice forhuman transthyretin Leu55Pro. Am J Pathol 161(5):1935–1948.

53. Cao P, Meng F, Abedini A, Raleigh DP (2010) The ability of rodent islet amyloidpolypeptide to inhibit amyloid formation by human islet amyloid polypeptide hasimportant implications for the mechanism of amyloid formation and the design ofinhibitors. Biochemistry 49(5):872–881.

54. Sellin D, Yan LM, Kapurniotu A, Winter R (2010) Suppression of IAPP fibrillation atanionic lipid membranes via IAPP-derived amyloid inhibitors and insulin. BiophysChem 150(1-3):73–79.

55. Uversky VN, et al. (2002) Biophysical properties of the synucleins and their pro-pensities to fibrillate: Inhibition of alpha-synuclein assembly by beta- and gamma-synucleins. J Biol Chem 277(14):11970–11978.

56. Park JY, Lansbury PT, Jr (2003) Beta-synuclein inhibits formation of alpha-synucleinprotofibrils: A possible therapeutic strategy against Parkinson’s disease. Biochemistry42(13):3696–3700.

57. Fan Y, et al. (2006) Beta-synuclein modulates alpha-synuclein neurotoxicity by re-ducing alpha-synuclein protein expression. Hum Mol Genet 15(20):3002–3011.

58. Robertson DC, et al. (2004) Developmental loss and resistance to MPTP toxicity ofdopaminergic neurones in substantia nigra pars compacta of gamma-synuclein, al-pha-synuclein and double alpha/gamma-synuclein null mutant mice. J Neurochem89(5):1126–1136.

59. Cookson MR (2009) alpha-Synuclein and neuronal cell death. Mol Neurodegener 4:9–22.

60. Fauvet B, et al. (2012) Characterization of semisynthetic and naturally Nα-acetylatedα-synuclein in vitro and in intact cells: Implications for aggregation and cellularproperties of α-synuclein. J Biol Chem 287(34):28243–28262.

61. Tenreiro S, et al. (2014) Phosphorylation modulates clearance of alpha-synuclein in-clusions in a yeast model of Parkinson’s disease. PLoS Genet 10(5):e1004302.

62. Hoi H, et al. (2010) A monomeric photoconvertible fluorescent protein for imaging ofdynamic protein localization. J Mol Biol 401(5):776–791.

63. Klivenyi P, et al. (2006) Mice lacking alpha-synuclein are resistant to mitochondrialtoxins. Neurobiol Dis 21(3):541–548.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1512876113 Fares et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

3, 2

020