Molecular and biological compatibility with host alpha-synuclein influences fibril pathogenicity

Summary

The accumulation and propagation of misfolded α-Synuclein (α-Syn) is a central feature of Parkinson’s disease (PD) and other synucleinopathies. Molecular compatibility between a fibrillar seed and its native protein state is a major determinant of amyloid self-replication. We show that cross-seeded aggregation of human (Hu) and mouse (Ms) α-Syn is bidirectionally restricted. Although fibrils formed by Hu-Ms-α-Syn chimeric mutants can overcome this inhibition in cell-free systems, sequence homology poorly predicts their efficiency in inducing α-Syn pathology in primary neurons or after intracerebral injection into wildtype mice. Chimeric α-Syn fibrils demonstrate enhanced or reduced pathogenicities compared to wildtype Hu- or Ms-α-Syn fibrils. Furthermore, α-Syn mutants induced to polymerize by fibrillar seeds inherit the functional properties of their template, suggesting transferable pathogenic and non-pathogenic states likely influence the initial engagement between exogenous α-Syn seeds with endogenous neuronal α-Syn. Thus, transmission of synucleinopathies is regulated by biological processes in addition to molecular compatibility.

Graphical abstract

Introduction

The propagation of misfolded proteins is an increasingly recognized disease mechanism in multiple neurodegenerative disorders (Jucker and Walker, 2013). Synucleinopathies, such as PD, are characterized by the intracellular accumulation of aggregated α-Syn (Lee and Trojanowski, 2006). Recent work indicates that fibrillar α-Syn seeds the pathological conversion of normal neuronal α-Syn when pathological α-Syn derived from diseased brains or recombinant α-Syn pre-formed fibrils (PFFs) enter neurons and induces the formation of fibrillary structures similar to Lewy bodies (LBs) and Lewy neurites (LNs) that typify PD (Luk et al., 2012a; Recasens et al., 2014; Volpicelli-Daley et al., 2011). Misfolded α-Syn propagation is further enhanced by the transfer of pathological α-Syn species from affected neurons to healthy ones (Desplats et al., 2009; Luk et al., 2012b) suggesting that the stereotypical spread of α-Syn pathology along neuroanatomical pathways observed in PD patients results from a templated-recruitment process reminiscent of that described for prion diseases (Braak et al., 2003; Dunning et al., 2013).

Despite its potentially central role in disease, little is known about the initial steps by which pathological α-Syn seeds corrupt native cellular α-Syn. Studies of mammalian and yeast amyloids have established that pathological seeding across species is generally inefficient. Resistance to cross-species templating likely arises at the molecular level, where sequence incompatibilities result in amyloid conformations or post-translational modifications that reduce the interface between the amyloid seed and substrate (Angers et al., 2010; Bett et al., 2012; Jones and Surewicz, 2005; Tessier and Lindquist, 2007). Even single mismatches in amino-acid sequence can greatly reduce the pathogenicity of amyloids or the susceptibility of a species as host for propagation (Santoso et al., 2000; Scott et al., 1993; Tycko, 2015). Distinct conformational strains of fibrillar α-Syn which induce dissimilar histopathological phenotypes have also been identified (Bousset et al., 2013; Guo et al., 2013; Peelaerts et al., 2015; Woerman et al., 2015), suggesting that tertiary and quaternary structure may influence disease pathogenesis in synucleinopathies.

The assembly kinetics of α-Syn indicate a nucleation-dependent recruitment process (Biere et al., 2000; Rochet and Lansbury, 2000; Wood et al., 1999). Substitution of the human sequence at either position 53 or 87 with the native mouse residue accelerates fibrillization rates substantially (Kang et al., 2011). Likewise, assembly rates for human-mouse chimeric α-Syn proteins fall between that of the two wildtype (wt) proteins and are proportional to their homology to wt mouse (Ms^wt α-Syn). Past studies have focused primarily on de novo α-Syn fibrillization in cell-free systems, although seeded fibrillization may be particularly relevant to the cell-to-cell spread likely occurring in PD. Recent studies showed that PFFs generated from Ms^wt α-Syn cause LBs/LNs within weeks after intracerebral injection in wt mice and are more potent than wt human (Hu^wt) α-Syn PFFs which induce pathology on the scale of months (Luk et al., 2012a; Masuda-Suzukake et al., 2013). We hypothesized that primary sequence differences between Ms^wt and Hu^wt α-Syn (Figure 1A; Larsen et al., 2009) account for this gap in seeding ability and designed a mutagenesis strategy targeting the 7 divergent positions between Ms^wt and Hu^wt α-Syn to identify key residues and domains that facilitate efficient templated conversion of α-Syn in vitro, in cultured neurons, and in wt mice.

Figure 1. Sequence differences between Hu^wt and Ms^wt α-Syn reduce cross-seeded fibrillization rate in vitro.

(A), Amino-acid sequences of Hu^wt and Ms^wt α-Syn are denoted in red and black, respectively. Divergent residues are highlighted. Chimeric α-Syn proteins contain orthologous residues at the indicated positions. (B), Schematic of the proposed molecular mechanism for cross-seeded α-Syn aggregation. Hu^wt monomer (black circles) readily add to Hu^wt PFFs (black wedges) through cis-templating, but recruitment of Ms^wt monomer (red circles) is slower due to imperfect trans-templating. Upon overcoming this initial step, recruitment of Ms^wt monomer is facilitated by complete sequence homology to hybrid PFFs. (C,E), Normalized initiation rates for 1% PFF seeded Ms^wt (C) and Hu^wt (E) monomer. Data represent time to reach 50% aggregation. (D,F), Normalized elongation rates (slope from 10%-90% pelletable fibrils) for 1% PFF seeded Ms^wt (D) and Hu^wt (F). All data shown as mean ± SEM (N=4). Dashed lines denote rate for unseeded reactions. One-way ANOVA (Tukey post-hoc), *p<0.05, **p<0.01, *** p<0.001 vs Hu^wt PFFs. See also Figures S1-S3.

Surprisingly, while sequence homology correctly predicted inefficient cross-seeding between Hu^wt and Ms^wt α-Syn in both cultured neurons and in wt mice, we found that in vitro seeding efficiencies of chimeric α-Syn PFFs correlated poorly with their ability to induce LBs and LNs in vivo. Finally, mutant α-Syn PFFs can confer this enhanced or reduced propagation capacity to monomers of another sequence, providing evidence that the ectopic seeding efficiencies observed arise from unique properties encoded within α-Syn PFF preparations.

Results

Sequence homology predicts efficiency of α-Syn cross-seeding in vitro

Sequence homology between a misfolded protein (i.e., a fibrillar or oligomeric seed) and its native soluble monomer (substrate) strongly influences the efficiency with which amyloids initiate templated conversion (Cobb and Surewicz, 2009). From a molecular perspective, the current hypothesis supports the concept that the rate of homologous or “self” recruitment (cis templating; Figure 1B, left path) occurs faster than that of cross-seeding, mainly due to delayed elongation caused by the reduced trans-seeding efficacy between the non-homologous seed and substrate. Once levels of converted substrate reach a critical threshold, cis-seeding of this sequence becomes the dominant kinetic factor, resulting in efficient elongation (Figure 1B, right path).

To determine whether seeding by α-Syn PFFs also follows such a relationship, we first confirmed that full-length Hu^wt, Ms^wt, and α-Syn chimeras (in which divergent residues were substituted with the orthologous amino-acids) are assembly-competent (Figures 1A, S1). Following incubation with agitation, all wt and chimeric α-Syn constructs formed fibrils averaging 12-15 nm in diameter as observed by electron microscopy (Figure S1B). Furthermore, PFFs showed comparable (>85%) levels of insoluble amyloid α-Syn after sedimentation and Thioflavin T (ThT) binding, although ThT fluorescence intensity was not proportional to fibril quantity (Figure S1A, C).

We evaluated how α-Syn sequence affects the ability of fibrils to act as pathological seeds by comparing the relative rates of Ms^wt α-Syn monomer aggregation in the presence of PFFs made from the different α-Syn constructs. Self-seeding with Ms^wt PFFs at 1% (w/w) of the monomer concentration dramatically accelerated the assembly of Ms^wt α-Syn compared to unseeded reactions (Figures 1C and S2). By contrast, initiation rates of cross-seeding reactions containing Hu^wt PFFs were similar to unseeded reactions.

Chimeric PFFs seeded Ms^wt α-Syn assembly with efficiencies between that of Hu^wt and Ms^wt PFFs. This correlated closely to the degree of homology between substrate and seeds, but did not achieve the cis-templating efficiency of Ms^wt PFFs (Figures 1C and S2). Interspecies substitutions of Ms^wt with the Hu^wt residue at Thr53 to Ala (Ms^T53A) and Asn87 to Ser (Ms^N87S) respectively reduced efficiencies by ~65% and 75% compared to Ms^wt PFFs. Fibrils combining these two substitutions (Ms^AS PFFs) seeded with ~80% lower efficiency than Ms^wt α-Syn PFFs.

Conversely, interspecies substitutions of Hu^wt with Ms^wt residues increased their seeding capabilities with conversion of Ser87 to Asn (Hu^S87N) having a greater effect than the A53T substitution (Figure 1C and S2). Simultaneous substitution at both positions (Hu^TN) conferred an initiation rate approaching 50% that of Ms^wt PFFs. Additional C-terminal substitutions, e.g. Asn103 to Gly (Hu^TNG), had relatively minor effects on the seeding efficiency of chimeric PFFs, consistent with previous studies showing that regions immediately flanking the hydrophobic NAC domain work in concert to influence overall in vitro fibrillization propensity of α-Syn (Bertoncini et al., 2005). Although seeding initiation rates varied greatly between chimeric α-Syn PFFs as seeds, elongation rates were similar between these reactions (Figures 1D, S2E). Thus, sequence discrepancies between the monomer substrate (Ms^wt) and seeding PFFs primarily affect the initial trans-templating event (i.e., initiation rate).

We reasoned that if differences at specific residues lead to inefficient templating between Hu^wt and Ms^wt α-Syn, then seeding initiation rates for wt and chimeric α-Syn PFFs should follow the same rank order when Hu^wt α-Syn monomer is used as the substrate instead of Ms^wt. As expected, cis-templating of Hu^wt α-Syn monomers by Hu^wt PFFs was significantly more efficient compared to PFFs comprised of other α-Syn constructs (Figures 1E and S3). Trans-templating of Hu^wt monomers by Ms^wt, Hu^TN and Hu^TNG PFFs proceeded inefficiently with initiation rates similar to unseeded reactions. However, trans-templating with Ms^AS PFFs, which contain human residues at 53 and 87, was nearly as efficient as Hu^wt PFFs, in contrast to the effect of converting Hu^wt at residues 53 and 87 to the Ms^wt sequence which effectively nullified seeding activity. These observations support the view that compatibility at these two residues greatly facilitates seed/substrate interaction. Therefore, similar to other amyloids, the propagation of α-Syn amyloid in vitro appears to be driven largely by individual residues rather than unique motifs.

As observed with Ms^wt α-Syn monomer as the substrate, overall elongation rates of Hu^wt α-Syn monomer in PFF-seeded reactions were comparable regardless of the α-Syn sequence encoded by the seed (Figure 1F, S3E), further supporting the hypothesis that once a critical level of cross-seeding is achieved, templating switches to a self-seeding mode that is less dependent on the composition of the seed. Taken together, while both wt and chimeric α-Syn PFFs are capable of seeding fibril formation in vitro, the initiation rates of templating are strongly determined by sequence homology between seeds and substrate.

Reduced efficiency of human-mouse α-Syn cross-seeding in neurons

To determine if the cross-seeding properties identified in vitro also apply to seeding in cells, we next compared the ability of wt and chimeric α-Syn PFFs to induce LN- and LB-like inclusions in cultured neurons. Consistent with our previous studies (Volpicelli-Daley et al., 2011; Volpicelli-Daley et al., 2014), hippocampal neurons from non-transgenic (CD-1) mice treated with 70 nM Ms^wt PFFs for 7d showed a dramatic increase in pSer129 α-Syn (pSyn; Figure 2A, B), a highly specific marker of LBs and LNs in human PD brains (Fujiwara et al., 2002). As predicted by in vitro seeding data, and in agreement with cross-species seeding being inefficient, addition of Hu^wt PFFs at the same concentration elicited lower levels of pSyn inclusions. Hu^A53T α-Syn PFFs exhibited intermediate levels of LB and LN pathology that mirrored the in vitro seeding potency of these fibrils with Ms^wt monomer. Biochemical analysis of Triton-X100 soluble and insoluble α-Syn 7d after seeding yielded results consistent with the immunocytochemistry (Figure 2C, D). Additionally, while differences in seeding efficacy persisted, all PFFs were able to elicit detectable levels of pathology when added to neurons at a higher concentration (140 nM; Figure S4), indicating that although sequence mismatches between PFF and substrate delay the onset of pathology, insoluble inclusions develop once seeding has been initiated.

Figure 2. Sequence-dependent and -independent impediments to α-Syn seeding in neurons.

(A) Immunostaining of pSyn (red) in primary hippocampal neurons treated at DIV10 for 7d with 70 nM of indicated α-Syn PFFs. Nuclei are stained with DAPI (blue). Scale bar = 50 μm. (B) Quantification of pSyn immunoreactivity in cultures [(area occupied × density)/DAPI count]. Results shown as mean ± SEM (N=4) relative to Hu^wt levels of pSyn (dashed line). (C) Primary hippocampal neurons treated with PFFs were sequentially extracted with 1% Triton X-100 and 2% SDS. Lysates were immunoblotted for total α-Syn or, pSyn, with βIII tubulin (TUJ1) as a loading control. (D) Densitometric quantification of insoluble α-Syn fractions in experiment in C. Results shown as mean ± SEM (N=3). *p < 0.05; **p < 0.01, *** p<0.001, one-way ANOVA vs Hu^wt PFF treatment (Tukey post-hoc test). See also Figure S4.

Intriguingly, Hu^S87N α-Syn gave rise to PFFs whose seeding in neurons exceeded that of Ms^wt PFFs (Figure 2A-D), even though these PFFs show a significantly slower initiation rate in Ms^wt monomer than Ms^wt PFFs in the in vitro seeded-aggregation assay (Figure 1B). Unexpectedly, Hu^TN PFFs elicited 94% and 85% less pSyn than Ms^wt PFFs at 70 nM and 140 nM, respectively, and were even less potent than Hu^wt PFFs despite sharing more homology with endogenous Ms^wt α-Syn (Figure 2C, D) and being 2.5-fold more efficient than Hu^wt PFFs in seeding Ms^wt monomer assembly in vitro (Figures 1C and S2). Thus, the combination of Thr and Asn at positions 53 and 87 appears to restrict the ability to initiate pSyn inclusions in neurons. Altogether, these observations suggest that, in addition to sequence homology, there are biological processes that further regulate seeding at the cellular level.

C-terminal modifications further influence α-Syn seeding in neurons

While only modest changes in in vitro seeding efficiency were observed for chimeras containing C-terminal substitutions beyond residue 87, the relative inactivity of Hu^TN PFFs in neurons indicates that these remaining residues contribute significantly to the seeding of Ms^wt PFFs in neurons. Besides positions 53 and 87, the primary sequences between Hu^wt and Ms^wt α-Syn differ at 5 additional residues (100, 103, 107, 121, and 122; Figure 1A). We therefore expected that one or more additional substitutions to Hu^TN at these positions would have a restorative effect on seeding when the resulting PFFs were added to neurons. Indeed, seeding activity of Hu^TN PFFs could be restored by several of the substitutions, most prominently by reintroduction of the native mouse residue at 103 (i.e., Hu^TNG; Figure S4). Although Hu^TNG and Hu^TNY PFFs show significant rescue of seeding efficacy in neuronal culture, re-introduction of mutations both closer (L100M) and further (D121G and N122S) from the hydrophobic core of α-Syn elicited only partial recovery (Figure S4C, E).

To further test the role of the C-terminus in neuronal seeding and its interplay with the residues at positions 53 and 87, we generated PFFs containing only residues 1-99 of Hu^wt or Ms^wt α-Syn (1-99Hu^wt and 1-99Ms^wt). Treatment of neurons with these PFFs showed significantly reduced levels of pSyn-immunoreactive inclusions compared to full length α-Syn (compare Hu^wt and Ms^wt with 1-99Hu^wt and 1-99Ms^wt in Figure S4C-E) indicating that the C-terminus facilitates efficient seeding in neuronal cultures.

Pathological seeding properties of PFFs are preserved in vivo

The divergence in pathological seeding activities between chimeric α-Syn PFFs in neurons versus cell-free assays prompted us to evaluate whether PFFs have similar cross-seeding properties in vivo. C57Bl6/C3H mice were injected with either Ms^wt or Hu^wt PFFs by means of a single unilateral injection into the dorsal striatum, which induces Lewy-like pathology in multiple CNS regions connected to the injection site, including piriform cortex (Pir), frontal agranular cortex (FC), amygdala (Amyg), and dopamine (DA) neurons of the substantia nigra pars compacta (SNpc) (Luk et al., 2012a; Masuda-Suzukake et al., 2013). As expected, Ms^wt PFFs induced a higher burden of α-Syn pathology compared to Hu^wt PFFs when examined by immunohistochemistry (IHC) 30 days post injection (dpi) in the Pir and FC, two regions that rapidly develop α-Syn inclusions after intrastriatal PFF inoculation. Whereas LBs were present in Pir and FC of Ms^wt PFF treated animals at 30 dpi, the same regions in Hu^wt PFF injected mice showed only traces of LN-like accumulations, reflective of the early stages of seeded pathology (Osterberg et al., 2015), and consistent with inefficient cross-seeding between Hu^wt and Ms^wt α-Syn.

We then asked whether the variations in seeding observed among chimeric PFFs in cultured neurons are preserved in vivo by examining the distribution of α-Syn pathology in mice injected with PFFs generated from Hu^S87N, Hu^TN or Hu^TNG monomer. By 30 dpi, striatal injections of Hu^S87N PFFs induced abundant α-Syn positive LNs and LBs in Pir and ipsilateral FC (Figure 3, second column). In contrast, mice injected with Hu^TN PFFs were devoid of α-Syn pathology in all areas examined by IHC for pSyn (Figure 3) and misfolded α-Syn (Syn506, data not shown). The inability of Hu^TN PFFs to induce α-Syn pathology was reversed by the reintroduction of the native mouse residue at 103 since Hu^TNG PFFs elicited comparable levels of pathology to Hu^S87N PFFs in these same regions. Thus, the induction of Lewy-like inclusions with chimeric α-Syn PFFs in mice is more congruent with data from our neuronal culture than from in vitro seeding studies and further suggests that these differences are underpinned by similar cellular processes.

Figure 3. Cross-seeding efficiencyof Hu^wt and Ms^wt α-Syn PFFs in vivo.

Wt mice were injected with either wt (Hu or Ms) or chimeric (Hu^S87N, Hu^TN, or Hu^TNG) α-Syn PFFs via a single unilateral injection into the dorsal striatum. IHC for pSyn was used to visualize LB- and LN-like inclusions. Representative images are shown for piriform cortex (Pir), frontal agranular cortex (FC) or olfactory mitral cell layer (MCL) ipsilateral to the injection site at 30 dpi. Inclusions are visible by 30 dpi in all three regions following treatment with Ms^wt, Hu^S87N, or Hu^TNG PFFs whereas Hu^wt induced mild α-Syn accumulation in processes (arrowheads). No pathology was detectable in animals that received Hu^TN PFFs. Scale bars = 30 μm. *p < 0.05; **p < 0.01, one-way ANOVA vs Ms^wt PFF treatment (Tukey post-hoc test).

Notably, the mitral cell layer (MCL) of the ipsilateral olfactory bulb, a discrete neuron population lacking direct projections to the injection site, also showed numerous pSyn inclusions after striatal injection of Ms^wt, Hu^S87N and Hu^TNG PFFs (Figure 3, bottom row). The comparable overall distribution of α-Syn pathology induced by these PFFs also point to shared neuroanatomical routes of spread consistent with trans-cellular spread of misfolded α-Syn. The MCL of mice injected with Hu^wt or Hu^TN PFFs lacked any α-Syn pathology, suggesting that α-Syn pathology spreads at a slower rate than with Ms^wt, Hu^S87N or Hu^TNG PFFs, consistent with a delayed initiation of aggregation in these mice.

Chimeric α-Syn sequences can enhance or suppress pathological seeding in vivo

To more precisely determine differences in the relative seeding by PFFs in vivo, we compared the extent of α-Syn pathology they induced over time in the amygdala, one of the first areas to develop LB- and LN-like pathology after intrastriatal PFF injections (Luk et al., 2012a). At 14 dpi, neurons containing pSyn were virtually undetectable in animals inoculated with Hu^wt PFFs (Figure 4A-C). By this time, however, Ms^wt PFF-treated animals had accumulated considerable pSyn-positive inclusions, primarily in neurites of amygdaloid nuclei. Mice injected with Hu^S87N or Hu^TNG PFFs exhibited even more abundant LNs as well as mature LB-like inclusions. Injection of Hu^A53T α-Syn PFFs (which, like Hu^S87N PFFs, contain only one murine residue) resulted in mainly LN pathology at 14 dpi, indicating they are less potent than either Hu^S87N or Hu^TNG PFFs in eliciting α-Syn pathology. The rank-ordered efficiencies with which PFFs induced pathology in vivo is therefore consistent with the seeding efficiency observed in culture.

Figure 4. Relative pathological seeding efficacy of wt and chimeric α-Syn PFFs is preserved in vivo.

(A) Wt mice were sacrificed at 14, 30, or 90 d after intrastriatal injection of wt or chimeric α-Syn PFFs. (B) Inclusion pathology in the ipsilateral amygdala revealed by anti-pSyn, Syn506, or thioflavin S (ThS). Injected α-Syn PFFs are arranged with increasingly sequence homology to the host Ms^wt α-Syn from top to bottom. Pathology is detectable in mice injected with Ms^wt, Hu^S87N, and Hu^TNG PFFs at 14 dpi. Hu^A53T PFF injected mice developed sparse pSyn immunoreactive processes (inset) but no amygdala pathology was detectable in the Hu^wt cohort until 30 dpi. Scale bars = 50 μm. (C) Quantification of α-Syn pathology at each time point. Data expressed as mean % area occupied by pSyn inclusions ± SD (N ≥4 per group).

Examination of PFF-injected brains at 30, 90, and 180 dpi revealed that α-Syn pathology continued to evolve with time in all mice treated with pathogenic PFFs (Figure 4B, C). By 30 dpi, LB-like inclusions were prominent in the amygdala of mice exposed to Hu^A53T, Hu^S87N, Hu^TNG or Ms^wt PFFs. However, only LN-like inclusions could be detected in Hu^wt PFF-injected mice, consistent with delayed initiation of seeding in these animals. The appearance of cytoplasmic inclusions occurred at later incubation periods (90 dpi), matching the previously described evolution for seeded α-Syn pathology in vivo (Luk et al., 2012a; Osterberg et al., 2015). Neuronal inclusions were also labeled with an antibody selective for misfolded α-Syn (Syn506; Waxman et al., 2008) and thioflavin S, confirming pathological α-Syn as a major constituent. However, we failed to observe any pathology in mice injected with Hu^TN PFFs through all time points up to 180 dpi, suggesting a complete lack of pathogenicity of these PFFs in vivo that was not observed with neuronal cultures.

Pathogenicity of α-Syn PFFs correlates with rate of neurodegeneration in vivo

In agreement with our previous studies, pSyn immunoreactive inclusions could be found in ipsilateral ventral SNpc neurons 30 dpi with Ms^wt PFFs (Figure 5A). Comparable levels of pSyn positive pathology were also seen in the SNpc following injections of either Hu^S87N or Hu^TNG PFFs, whereas Hu^A53T PFFs resulted in primarily LN accumulations characteristic of early stages of seeded pathology. Inclusions were barely detectable in DA neurons after Hu^wt PFFs treatment and mice inoculated with Hu^TN PFFs remained free of α-Syn inclusions in all areas examined.

Figure 5. α-Syn pathology in midbrain DA neurons following injections of different α-Syn PFFs.

(A) Representative images of pSyn immunoreactive inclusions accumulating in SNpc DA neurons at 30 or 90d after inoculation with α-Syn PFFs. PFFs are listed in order of increasing homology to Ms^wt α-Syn. Inclusions were detectable in all cohorts except Hu^TN PFFs which remained clear of pSyn accumulation up to 90 dpi (white arrowheads). Scale bars = 100 μm. (B) Quantification of the proportion of SNpc DA neurons containing pSyn-positive inclusions. Initiation of α-Syn pathology is significantly delayed in mice treated with Hu^wt PFFs compared to Ms and chimeric PFFs. (C) DA neuron number in the SNpc after treatment with indicated wt or chimeric PFFs at 30, 90, and 180 dpi. All data shown are for the hemisphere ipsilateral to the injection site and shown as mean ± SEM (N = 5-8 per group). No inclusions were detected in the contralateral SNpc and neuron numbers there were unchanged for all groups. (D) Rotarod performance of α-Syn PFF-injected mice at 30, 90, and 180 dpi with indicated PFFs. *p < 0.05, **p <0.01, ***p < 0.001, two-way ANOVA (Tukey’s HSD). See also Figure S5.

Quantification of tyrosine-hydroxylase (TH)-positive SNpc neurons harboring pSyn inclusions confirmed that by 30 dpi following injections of Ms^wt PFFs, pathological α-Syn deposits were already present in ~28% of SNpc DA neurons (Figure 5B). A comparable percentage of DA neurons were affected in mice treated with Hu^S87N or Hu^TNG PFFs (26 and 29%, respectively) indicating that these PFFs likely initiate α-Syn pathology in similar DA subpopulations. Consistent with their delayed seeding capacities, this proportion was not reached in Hu^A53T PFF-treated mice until 90 dpi, whereas less than 20% of DA neurons in Hu^wt PFF injected animals displayed pSyn pathology at 180 dpi. Thus, with the exception of Hu^TN, DA cells can ultimately be seeded by each type of PFF. Interestingly, the proportion of SNpc neurons bearing α-Syn inclusions did not exceed 38% after injections with PFFs of any α-Syn construct at any time-point up to 180 dpi, suggesting that templated α-Syn pathology follows a similar trajectory of spread for pathogenic PFFs.

Since SNpc neurons that harbor α-Syn inclusions progressively degenerate in experimental studies (Luk et al., 2012a; Paumier et al., 2015; Recasens et al., 2014), we quantified the remaining number of DA neurons in this region at each time point to determine if the rate of inclusion formation correlated with cell loss (Figures 5C and S6). The most significant reduction in DA neurons was observed in cohorts injected with the most pathogenic PFFs (i.e., Hu^S87N, Hu^TNG or Ms^wt), where SNpc neurons were reduced by 9-22% at 90 dpi. This decrease approached ~40% by 180 dpi, corresponding to the drop in the number and proportion of pSyn-bearing neurons at this time point (Figure 5B, C). Consistent with this DA neuron loss, mice injected with Hu^S87N, Hu^TNG or Ms^wt PFFs also showed progressive deterioration in rotarod performance that was significant by 180 dpi (Figure 5D). Importantly, neither Hu^wt nor Hu^TN PFFs caused any significant reduction in DA neuron number up to 180 dpi.

Wildtype and chimeric α-Syn fibrils are potent seeds in non-neuronal cells

The notable deficiency in seeding activity exhibited by Hu^TN PFFs in neuron cultures and in vivo argues that its initial engagement with cellular α-Syn may be the rate-limiting step. To investigate this possibility, we measured the seeding activity of wt and chimeric PFFs in a previously established cell-model of seeded synucleinopathy where PFFs are delivered directly by cationic liposomes to the cytoplasm of HEK293 cells transfected with either Ms^wt or Hu^wt α-Syn (Luk et al., 2009). All PFF preparations, including Hu^TN PFFs, induced a similar amount of insoluble pSyn inclusions in cells (Figure S6). Thus, when differences in PFF internalization/delivery are eliminated, Hu^TN PFFs are equally potent to other PFFs in recruiting and converting endogenous α-Syn into pathological aggregates, consistent with observations from our in vitro seeding assays.

Non-pathogenic chimeric α-Syn PFFs shows normal entry into neurons

Given that Hu^TN PFFs are highly efficient in inducing pSyn inclusions when delivered into cells, yet seed poorly when passively administered to neurons or mice, we asked if the internalization of Hu^TN PFFs by neurons is compromised. We previously showed that α-Syn PFFs are detectable in >80% of neurons after 6 h of exposure to Hu^wt PFFs (Tran et al., 2014). Hippocampal neurons derived from Snca^−/− mice were used to compare the internalization efficacy of Hu^wt, Hu^TN and Ms^wt PFFs at 6 and 24 h post transduction to eliminate confounds of detecting endogenous α-Syn. All PFF preparations were readily internalized as indicated by detectable α-Syn positive puncta juxtaposed to markers of early endosomes (EEA-1) and lysosomes (LAMP1) (Figure S6 D, E). The proportion of neurons containing internalized α-Syn PFFs was similar across these three constructs, suggesting that the lack of neuronal seeding seen with Hu^TN PFFs is not due to an overall reduction in PFF entry into neurons. Similarly, the proportion of neurons containing α-Syn PFF in LAMP1-positive structures were indistinguishable across these fibril types, suggesting that, at least at this early stage, they share a common route of internalization and subcellular processing (Figure S6 F, G). No differences were observed in the proportion of PFF-bearing neurons at 24 h post-treatment, further indicating that internalization is not responsible for the inability of Hu^TN PFFs to induce α-Syn inclusions in neurons (data not shown).

Hu^TN and chimeric PFFs display distinct protease sensitivities

Since Hu^TN differs in sequence from pathogenic chimeras such as Hu^A53T and Hu^S87N by a single amino acid, we tested whether wt and chimeric PFFs assembled de novo may encode distinct conformations. Partial digestion with various proteases revealed several unique fragmentation patterns among PFFs. Proteinase K (PK) treatment clearly differentiated Hu^wt, Hu^A53T, Hu^S87N, and Ms^wt PFFs (Figure 6A). We and others have previously shown that the five major bands following PK digestion of Hu^wt PFFs consist of 5 polypeptides: 1) aa1-140, 2) ~aa19-140, 3) ~aa32-140, 4) ~aa19-120, and 5) ~32-120 (Miake et al., 2002). Several trends emerge from the digestion patterns of PFFs as the α-Syn sequence is modified from Hu^wt to Ms^wt. Firstly, increased N-terminal cleavage occurs with perturbation of the Hu^wt sequence simultaneously at positions 53 and 87, resulting in a relative decline of full-length α-Syn (band 1) compared to bands 2-5, (compare band 1 for Hu^TN, Hu^TNG, and Ms^WT with Hu^wt, Hu ^A53T, and Hu^S87N in Figure 6A). Second, the shift from relatively equivalent amounts of bands 2-4 in Hu^wt, Hu ^A53T, and Hu^S87N PFF digests to a near total loss of band 3 in the Hu^TN, Hu^TNG chimeras indicates that these residue changes prevent further N-terminal truncation to generate band 3, while facilitating cleavage at the C-terminus to form band 4 (Figure 6A, B). Finally, PK digestion of Ms^wt PFFs shows a dramatically different pattern with rapid digestion to band 4, indicating facile degradation at both N- and C-termini (Figure 6A-D). When comparing the digestion pattern of Hu^TN and Ms^wt PFFs, in which only the C-termini vary, reduction in the stability of band 2 from Hu to Ms implies that the switch in C-terminal residues has a collective effect in exposing the N-terminus of Ms^wt PFFs for further degradation.

Figure 6. Protease treatment of wt and chimeric PFFs reveal distinct digestion patterns.

Wt and chimeric α-Syn PFFs were digested with Proteinase K (A) or Pronase E (B) at the indicated concentrations. Digestion products were resolved on a 12% Bis-tris gel and visualized with Coomassie Brilliant Blue. Major bands generated by each enzyme are indicated in red. (C-G) Quantification of specific PFF digestion fragments are shown. Results represent mean band intensity relative to undigested PFFs. Data shown as mean ± SEM from ≥3 independent digestion experiments. *p < 0.05, **p <0.01, ***p < 0.001, two-way ANOVA (Bonferroni test) for all groups vs. Hu^TN, Hu^TNG, and Ms^wt (C); Ms^wt vs. all groups (D); Hu^wt vs. all groups (E); Hu^TNG vs. all groups (F); Hu^TN vs. all groups (G). See also Figures S5-S7.

Since PK treatment did not differentiate all of the PFFs examined, particularly proseeding Hu^TNG PFFs from non-pathogenic Hu^TN PFFs, additional proteases were used. While trypsin and chymotrypsin yielded additional differences between Hu^wt, Hu^A53T, and Hu^S87N PFFs, Hu^TN and Hu^TNG PFFs continued to lack clear differences in digestion patterns despite their starkly different potency in vivo (Figure S7). However, partial digestion with Pronase E revealed distinct patterns of protein fragments for Hu^TN and Hu^TNG PFFs (Figure 6B, E-G). Hu^TN PFFs appeared to be more resistant than Hu^TNG PFFs at low concentrations (Figure 6E, F), but failed to build up the stable amounts of smaller fragments prominent in digestions of all seeding competent PFFs, including Hu^TNG (e.g., bands 5 and 7 in Figure 6B, G). Collectively, these results indicate the differences between Hu^wt and Ms^wt PFFs are largely driven by residues at 53 and 87, although terminal regions of α-Syn also play an important role in the overall structure of PFFs.

Pathological seeding properties of α-Syn PFFs are transferrable

Since our biochemical data thus far established that Hu^TN and Hu^wt PFFs have distinct tertiary structures and that Hu^TN PFFs formed de novo are non-seeding in neurons, we determined if these two properties are interdependent. To test whether the nature of α-Syn seeds dictate PFF tertiary structure and/or neuronal seeding ability, we seeded assembly of Hu^TN monomer with 20% Hu^wt PFFs to generate Hu^TN[Hu^wt] PFFs (Figure 7A). In the same manner, Hu^wt[Hu^TN] PFFs were made by seeding monomeric Hu^wt with de novo Hu^TN fibrils. Upon PK treatment, Hu^TN[Hu^wt] PFFs produced a banding pattern nearly identical to that of Hu^wt PFFs, despite being comprised primarily of Hu^TN monomer (Figure 7B). Importantly, none of the bands examined represented truncation events beyond positions ~32 or ~110, terminal areas in which these constructs share identical sequences, thus eliminating the possibility that the substitutions within the Hu^TN sequence were responsible for the observed differences. Moreover, this pattern was clearly distinct from PK digestion of de novo Hu^wt and Hu^TN PFFs mixed together just before PK treatment. Conversely, the digestion pattern of Hu^wt[Hu^TN] PFFs mirrored that of Hu^TN PFFs (Figure 7B ). Thus, under these seeding conditions, Hu^wt and Hu^TN α-Syn monomers are capable of adopting structural properties of the templating PFFs.

Figure 7. Pathogenic and non-pathogenic properties of PFFs are transferable.

(A) Schematic illustrating generation of hybrid fibrils through seeding with pathogenic and non-pathogenic PFFs. (B) Wt and chimeric α-Syn PFFs assembled de novo were digested with PK at the indicated concentrations. Digestion products in Bis-tris gels were stained with Coomassie Brilliant Blue. (C) Immunostaining for pSyn in primary hippocampal neurons treated with Hu^wt PFFs (140 nM or 28 nM to test the effect of the quantity used for seeding), Hu^TN[Hu^wt] PFFs (20% Hu^WT seeded Hu^TN PFFs), Hu^TN PFFs, and Hu^wt[Hu^TN] PFFs (20%Hu^TN seeded Hu^wt PFFs) at 140 nM. Primary hippocampal neurons (DIV10) were treated with α-Syn PFFs and 7d later cells were fixed and stained. Nuclei are stained with DAPI (blue). Scale bar = 50 μM. (D) Quantification of pSyn levels in PFF-treated neurons [(area occupied × average intensity)/DAPI]. Results shown as mean ± SEM (N=3; *p < 0.05; **p < 0.01, one-way ANOVA vs Hu^wt PFF treatment with Tukey post-hoc test). (E) DIV10 primary hippocampal neurons were treated for 7d with 140 nM PFFs and then sequentially extracted with 1% Triton X-100 followed by 2% SDS. Lysates from Triton and SDS fractions were immunoblotted with Syn9027 (total α-Syn), MJF-R13 (pSyn), and TUJ1 loading control. See also Figure S7.

To determine if Hu^TN[Hu^wt] and Hu^wt[Hu^TN] PFFs generated by cross-seeding also inherit the pathogenicity of the templating PFFs, we compared their ability to induce α-Syn inclusions in hippocampal neurons (Figure 7D-F). As predicted by its digestion pattern, Hu^TN[Hu^wt] PFFs were potent in eliciting LBs and LNs in cultured neurons (Figure 7C). Remarkably, Hu^TN[Hu^wt] PFFs treated cultures developed approximately 4-fold more insoluble pSyn pathology as measured by IF and immunoblot compared to de novo aggregated Hu^TN PFFs, indicating that the pathogenicity of Hu^wt PFFs had been successfully conferred onto Hu^TN α-Syn (Figure 7D, E). Surprisingly, the level of pSyn pathology induced by Hu^TN[Hu^wt] PFFs was nearly twice the amount in cultures treated with Hu^wt PFFs, likely due to the additional homology between Hu^TN and Ms α-Syn. Conversely, neurons exposed to Hu^wt[Hu^TN] PFFs showed a marked reduction in α-Syn inclusions compared to cultures that received de novo aggregated Hu^wt PFFs, indicating that Hu^wt had adopted the non-pathogenic state of the Hu^TN seeds used to generate this material, demonstrating that both a pathogenic and a non-pathogenic state are transferable between different α-Syn sequences.

Discussion

Recent studies have shown that misfolded α-Syn species template the conversion of native α-Syn into pathological PD-like inclusions, a process that likely underlies the spread of cellular α-Syn pathologies observed in prototypical synucleinopathies such as PD. Although fibril-seeded LN and LB formation has been observed in both neuron cultures and in vivo models, how pathological α-Syn potentiates the conversion of the endogenous normal α-Syn pool is unclear. Our data here show that, similar to other amyloids, PFF-mediated cross-seeding between homologous α-Syn sequences is significantly more efficient than that of Hu and Ms α-Syn. Using a series of Hu-Ms chimeric α-Syn PFFs, we find that at the molecular level seeding efficiency is determined by sequence homology between the α-Syn PFF seeds and soluble α-Syn monomers.

However, our findings reveal a more complex picture at the cellular level as the in vitro seeding potencies for several PFFs did not directly correlate with their ability to induce pathology. While Hu-Ms cross-seeding is markedly less effective when induced by PFFs in neurons and in vivo, we also identified chimeric α-Syn proteins that give rise to PFFs that either significantly enhanced (Hu^S87N and Hu^TNG) or suppressed (Hu^TN) induction of α-Syn pathologies compared to Ms^wt α-Syn PFFs. Among the residues examined, the two residues closest to the core hydrophobic (NAC) domain of α-Syn (53 and 87) show the strongest influence on PFF cross-seeding in vivo, as highlighted by the fact that single substitutions towards resembling either Hu^wt or Ms^wt immediately restore neuronal seeding activity to PFFs, although pathogenicity is also influenced by interactions between these two residues and the C-terminal region (particularly around aa 103-107).

Our in vivo data also showed that these altered seeding characteristics apply across multiple brain regions and neuronal subtypes. Most strikingly, Hu^TN PFFs have a lower seeding capacity than even Hu^wt PFFs in both neuron culture and in vivo despite its greater homology with Ms α-Syn and higher seeding potency in cell-free systems. Thus, in addition to molecular compatibility between seed and substrate, processes operating at the cellular level also determine the efficiency of α-Syn pathology transmission.

When cross-seeded with Hu^wt PFFs, Hu^TN monomers form pathogenic PFFs. Similarly, Hu^wt monomer assembled in the presence of non-pathogenic Hu^TN seeds display reduced potency in neurons, making it unlikely that co-purifying contaminants are responsible for the different activities of seeding or non-seeding α-Syn constructs. These data also suggest that pathogenicity can be transferred between these two sequences through in vitro templating, consistent with the notion that α-Syn fibrils can assume multiple conformations (Bousset et al., 2013; Guo et al., 2013). It is possible that specific fibril conformations have distinct biological activities, although further studies are needed to elucidate the relationship between structure and pathogenicity in neurons. Indeed, our PFF preparations may potentially contain multiple misfolded species with differing pathogenicities, and it is conceivable that alteration of the α-Syn primary sequence could shift this equilibrium.

Our findings present a number of possible, non-exclusive, scenarios as to how different α-Syn seeds affect pathological α-Syn propagation. Firstly, PFFs may differ in their ability to enter into neurons. In this regard, we did not detect differences in the internalization or subcellular compartmentalization of Hu^TN PFFs vs. other pathogenic fibrils although more sensitive methods are necessary to interrogate specific differences in subcellular localization or if various PFFs differ in their susceptibility to intracellular processing or degradation. Secondly, neuronal α-Syn is preferentially localized to presynaptic vesicles and is associated with membranes under physiological conditions where it maintains a different conformation than its unfolded state in cell-free systems (Dettmer et al., 2015). Therefore, monomeric α-Syn could be highly accessible to pathological α-Syn species of various configurations (hence sequence compatibility governs cell-free seeding rate) whereas membrane-bound α-Syn might interface with only a subset of configurations such as those found within pathogenic PFF preparations (Sidhu et al., 2016). This remains to be confirmed, although in agreement with this, Hu^TN PFFs efficiently induce α-Syn inclusions when transduced into α-Syn overexpressing non-neuronal cells, suggesting a reduced ability to interact with endogenous neuronal α-Syn. Thirdly, pathogenic and non-pathogenic PFFs may differ in their interaction with as yet unknown cofactors that could facilitate pathological α-Syn seeding in neurons (Supattapone, 2014).

It is interesting that while the species barrier for the infectious conformer of human prion protein (PrPsc), which in its normal conformation is a surface glycoprotein, is particularly sensitive to changes in the amino acid sequence, the phenomenon described here for α-Syn, an intracellular protein, shows a dramatic dependence on features beyond primary structure (Cobb and Surewicz, 2009; Collinge and Clarke, 2007). This suggests that where a seed and the endogenous protein meet is a key consideration for understanding how the various proteins that aggregate in neurodegenerative disease propagate. Further understanding of the fate of internalized seeding-incompetent PFFs comprised of Hu^TN α-Syn should help elucidate the nature of this initial interaction and thus provide important mechanistic insights into the process of templated α-Syn seeding in neurons.

Lastly, our data showing that similar neuroanatomical regions develop α-Syn pathology regardless of the cell-free seeding efficiency of the injected α-Syn PFFs suggests that the sequences described here alter their ability to trigger the initial misfolding of cellular α-Syn, and is in line with a model whereby α-Syn pathology is induced through seeding in trans but that subsequent propagation proceeds through cycles of seeding in cis involving primarily endogenous α-Syn. Consistent with this, we observed that DA neuron degeneration was most rapid following injection of Ms^wt and chimeric α-Syn PFFs with greater neuronal seeding rates and, consequently, more quickly developing α-Syn pathology. This observation suggests a direct relationship between the onset of α-Syn pathology burden and neuronal toxicity. In summary, our investigation of cross-seeding between Hu and Ms α-Syn has revealed important insights into the pathogenicity of a protein central to neurodegeneration.

Experimental Procedures

Animals

Wt (C57BL6/C3H) mice were obtained from the Jackson Laboratories (Bar Harbor, ME). Primary neuronal cultures were prepared from timed-pregnant (CD-1) mice (Charles River Laboratories; Wilmington, MA) and Snca^−/− mice (Abeliovich et al., 2000) that were maintained on a C57BL6 background. All housing, breeding, and procedures were performed according to the NIH Guide for the Care and Use of Experimental Animals and approved by the University of Pennsylvania Institutional Animal Care and Use Committee (IACUC).

Recombinant α-Syn Purification and In Vitro Fibrillization

Constructs for chimeric α-Syn proteins were generated through PCR site-directed mutagenesis with the PRK172/Hu^wt α-Syn or PRK172/Ms^wt α-Syn plasmid as template. Wt and chimeric α-Syn constructs were expressed in E.coli BL21 (DE3) RIL cells and purified as previously described (Giasson et al., 2001; Luk et al., 2012a; Luk et al., 2009; Volpicelli-Daley et al., 2014). Fibrillization was conducted by diluting recombinant α-Syn to 360 μM (5 mg/mL) in sterile Dulbecco’s PBS (pH 7.0; Mediatech Inc.) with constant agitation (1,000 rpm, at 37°C) for 7 days in an Eppendorf Thermomixer C^™ with ThermoTop^™. Successful fibrillization was verified by sedimentation, TEM analysis, and ThT-binding. Further details are described in Supplemental Experimental Procedures.

Kinetic analysis of seeded α-Syn in vitro fibrillization

Efficacy of seeding for a particular construct was assayed by adding 1% PFFs (25 μg/mL) in to a fibrillization reaction containing 99% freshly thawed and spun Ms^wt or Hu^wt monomer (2.5 mg/mL in the final reaction). All seeded fibrillization reactions were performed at a final concentration of 2.5 mg/mL (180 μM) and final volume of 750 μL in order to facilitate differentiation of aggregation rates. Sedimentation was performed at the time points indicated. For each α-Syn construct, >2 independently prepared monomer batches were tested for each fibrillization paradigm, and >2 series of seeded fibrillization reactions were run for each monomer batch. Relative quantities of protein in each fraction were assessed by densitometery using a Licor Odyssey scanner at 680 nm. Graphing was performed with Graphpad Prism 4 and the curve-fitting feature of the software was used to generate the elongation rate (Hill slope) and initiation rate (EC50) data for Figures 1, S2 & S3.

Primary Neuron Cultures and Fibril Transduction

Primary neuron cultures were prepared from CD1 or Snca^−/− mice on embryonic day (E) 15–E17 as previously described (Volpicelli-Daley et al., 2014). Dissociated hippocampal neurons were plated onto poly-D-lysine coated coverslips in 24-well plates (100,000 cells/coverslip). Transduction was performed at 10DIV using α-Syn PFFs diluted to 2.4 μM in PBS (without Mg²⁺/Ca²⁺) and sonicated with a Biorupter (Diagenode; Denville, NJ) at high power for 10 cycles (30s on, 30s off at 10°C ). At 4 d post-transduction, 30% of the media was gently aspirated from each well and replaced with 200 μL of fresh media. For biochemical studies, PFFs (140 nM) were added to neurons in a 6-well plate containing 1 million cells per well. Transduced neurons were harvested for immunocytochemistry or sequential extraction at 7 days post transduction (17 DIV).

Proteolytic digestion of PFFs

Proteinase K (P-2308), trypsin (T-9201), chymotrypsin (C-4129), and Pronase E (P-5147) were obtained from Sigma and stock solutions prepared in water. Chymotrypsin was dissolved in 1 mM HCl. Ten micrograms of PFFs generated from de novo assembly reactions were digested with the indicated concentrations of protease in Dulbecco’s PBS. Digestions were performed in a final volume of 20 μL and incubated at 37°C for either 30 min (Proteinase K, trypsin) or 40 min (Pronase E, chymotrypsin). Digestions were stopped with 1 mM PMSF or 100 μM aprotinin. Reaction samples were then boiled with SDS-sample buffer for 5 min and resolved on NuPAGE Novex 12% Bis-Tris gels (Invitrogen) for Proteinase K, chymotrypsin, and Pronase E reactions or Novex 16% Tricine gels for trypsin digestions. All experiments were preformed 3-5 times with at least 2 independently prepared batches of PFFs. Representative images from Coomassie stained gels are shown in Figure 6 and Figure S7.

Stereotaxic injection of PFFs

Prior to intrastriatal injection, PFFs were first diluted in sterile PBS and sonicated as described above. Mice (2-3 months old) were anesthetized with ketamine hydrochloride (100 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.). A single injection (5 μg PFFs) was stereotaxically injected into the striatum (co-ordinates: +0.2 mm relative to Bregma, +2.0 mm from midline, +2.6 mm beneath the dura). Injections were made with 10 μL syringes (Hamilton, NV) at a rate of 0.1 μL per min (2.5 μL total) with the needle in place for ≥ 5 min at each target. Animals were monitored regularly following recovery from surgery, and sacrificed at various pre-determined time points (14, 30, 90, or 180 dpi) by overdose with ketamine/xylazine. For histological studies the brain and spinal cord were removed after transcardial perfusion with PBS and underwent overnight postfixation in either neutral buffered formalin (Fisher Scientific) or 70% ethanol (in 150 mM NaCl, pH 7.4), before being processed and embedded in paraffin.

Statistical analysis

Data from in vitro cross-seeding reactions, neuronal culture experiments and immunoblot quantification were compared with GraphPad Prism software (v.4), using one-way analysis of variance (ANOVA). Post-hoc testing was performed using the Tukey method. One-way ANOVA was also used to compare in vivo pSyn pathology within select regions of interest at 30dpi. Comparison of histopathological and behavioral data in the timed-cohort series was performed using STATA software (v.11). Means from each group were analyzed by two-way ANOVA with α-Syn PFF treatment and incubation time as factors. Post-hoc pairwise comparisons were performed using Tukey’s HSD method. Results from protease digestion experiments were analyzed by two-way ANOVA (α-Syn PFF and protease concentration as factors). Bonferonni correction was applied to pairwise comparisons.