α-Syn overexpression using AAVs: filling of the neuronal cytoplasm and pS129 positivity without assembly of fibrils and without inclusions.
The overexpression of hSyn is commonly used to model αSP in mouse neurons, both in vitro and in vivo. It is supposed to facilitate the emergence of α-Syn aggregation by bringing its concentration closer to the nucleation threshold. It also allows studying the human protein in a rodent context, tentatively increasing the translational relevance of the model. Several studies reported the induction of α-Syn aggregation by simple overexpression, others reported that aggregation must be induced by an additional seeding event such as PFFs treatment (37). In both cases, the discrimination between the adverse effects due to the increased α-Syn expression levels and those due to the proper amyloid aggregation are difficult to determine and relies on the capability of specifically identifying amyloid α-Syn fibrils in situ.
pS129 is a very specific marker of α-Syn inclusions in human neuropathology(30)(39) (40). Similarly, when α-Syn PFFs are used as seeds and directly injected into the brain of wt mice, pS129 antibodies specifically allow the visualization of the α-Syn inclusions that develop and spread into the mouse brain(9)(10) (41). Four months after an injection of PFFs at the level of the SN, several types of pS129-positive (pSyn) inclusions formed that were reminiscent of those observed in patients, both locally in the SN and at distance in the Striatum of the same hemisphere (Fig. 1A-E). This experimental αSP phenotype can also be observed after the intracerebral injection in mice of α-Syn fibrils extracted from DLB patient brain samples using detergent, with the formation of typical pS129-positive α-Syn inclusions (10). Note, however, that in agreement with the latter publication, we found that 4 months after injection, the αSP seeded by PFFs was not associated with any specific loss of TH-positive neurons in the SNpc. Indeed, the loss observed, which was around 28% compared to the non-injected side, was not different in sham-injected controls which received an injection of saline (Fig. S1).
For the sake of comparison, we injected AAV particles carrying the cDNA of hSyn under the control of a synapsin promoter (AAV-syn) to induce an overexpression of hSyn in the SN neurons (Fig. S1). In these AAV-injected animals, hSyn is readily detectable at 4 months post-injection in the cell bodies of the dopaminergic neurons populating the SN as well as in their striatal projections/terminals (Fig. S1). In these animals, bona fide intracellular inclusions cannot be identified inside the somata of the dopaminergic neurons (the cell bodies become globally/diffusely pS129-positive), and most strikingly, distant striatal pS129-positive inclusions are absent (Fig. 1A-E). Such lack of long-distance spread as well as the diffuse perikaryal morphology of the pS129 signal is reminiscent of previous in vivo experiments in which α-Syn overexpression resulted in S129 phosphorylation without fibrillar aggregation (37).
Note that 4 months after the intracerebral injection of AAV-syn, on top of the absence of αSP inclusions, no specific loss of TH-positive neurons was detected in the SNpc. Compared to the non-injected side, the loss observed was identical to sham-injected controls (Fig. S1).
In the last experiment, we mixed AAV-syn particles with PFFs and injected them at the level of the SN. In these “seeded plus AAV-infected” animals, hSyn was readily detectable in the cell bodies of the dopaminergic neurons of the SN as well as in their striatal projections/terminals (Fig. S1). While bona fide intracellular inclusions were still not identified inside the somata of the dopaminergic neurons (here also the cell bodies were globally/diffusely pS129-positive), distant striatal pS129 positive inclusions were present like in the PFF-only condition (Fig. 1A-E). Induction of an inclusion pathology with long-distance spread by adding PFFs to AAV-syn particles clearly indicates that the processes triggered by AAV-syn and PFFs are completely different (compare “AAV-Syn” and “AAV-syn + PFFs” in Fig. 1C), yet they are both associated with the emergence of a pS129-positive signal at the injection site (compare “AAV-Syn” and “PFFs” in Fig. 1B). This indicates that S129 phosphorylation is a misleading marker in conditions of neuronal α-Syn overexpression in vivo.
Note that here also, 4 months after the intracerebral injection of the AAV-syn plus PFFs mix, no specific loss of TH-positive neurons was detected in the SNpc. The loss observed compared to the non-injected side was identical to sham-injected controls (Fig. S1).
In parallel to these in vivo experiments, we evaluated the relationships between α-Syn overexpression, phosphorylation, and amyloid aggregation in primary cultures of cortical neurons infected or not with AAV-syn and seeded or not with PFFs. In line with the in vivo experiment, pS129 was not detected in non-infected control neurons and only appeared upon seeding with PFFs. Like in αSP pathology, the seeded structures which are positive for pS129 are bona fide cytological inclusions (LNs, perinuclear and intranuclear inclusions (Fig. 2A,B). In addition, pS129 staining was fully resistant to permeabilization of the plasma membrane with digitonin prior to fixation, indicating that it corresponds to large insoluble α-Syn assemblies – i.e., fibrils – unable to diffuse out of the neurons through the digitonin pores (Fig. 2C,D).
In contrast, in non-seeded conditions, AAV-syn-infected neurons already showed a diffuse pS129 signal with no identifiable inclusion (Fig. 2A,B). At major variance from the previous experiment, this diffuse staining was completely lost upon membrane permeabilization, indicating that S129 phosphorylation concerned in this case soluble α-Syn species that can freely diffuse out of the neurons and not insoluble fibrils (Fig. 2C,D). However, we observe that if the AAV-syn-infected neurons were also seeded with PFFs, pS129 staining became fully resistant to permeabilization revealing that fibril assembly could still be seeded at the expense of the soluble pS129 α-Syn species that were formed in reaction to the overexpression overflow (Fig. 2C,D).
These data show that in conditions of α-Syn overexpression, pS129 cannot be considered a reliable surrogate marker of neuronal α-Syn fibrillization and aggregation. In addition, even if the pS129 signal integral is increased by seeding (Fig. 2A-D), phosphorylation is not discriminant per se: phosphorylated fibrils cannot be distinguished from the phosphorylated soluble species “artifactually” emerging in these neurons.
To identify aggregated α-Syn more specifically in these experiments we tested the routine conformational antibody LB509. LB509 has a double selectivity, detecting hSyn mostly under its aggregated form. We found in our conditions that this conformational preference was very relative - that is, in hSyn overexpressing conditions (AAV-syn) LB509 detects non-aggregated α-Syn in neurons, particularly inside the synapses. This staining is completely lost upon plasma membrane permeabilization prior to fixation confirming its non-fibrillar nature. When such AAV-syn infected neurons were seeded with PFFs, aggregation could only be inferred from morphological changes and not be quantified by measuring an increase of the LB509 signal integral in the intact cells (Fig. 2E). The fibrillar status of the assemblies seeded by PFFs and recognized by LB509 could, however, be revealed by their resistance to plasma membrane permeabilization (Fig. 2E).
Interestingly, by comparing the signal distribution of pS129 and LB509 by double immunofluorescence in the latter permeabilized neurons, we observe that the signals are not completely overlapping (Fig. 2F). This indicates that pS129-positive amyloid α-Syn fibrils coexist with unphosphorylated fibrils, and that some pS129-positive fibrils are not discovered by LB509. This suggests a conformation and phosphorylation heterogeneity of the C-terminus (the target of both antibodies) belonging to the monomers engaged into the amyloid fibril core.
In conclusion, in experimental models introducing and manipulating the concentration of hSyn monomers in neurons, pS129 cannot be considered a reliable marker of aggregation because overexpressed α-Syn is phosphorylated under its soluble form. Appropriate tools are needed for identifying amyloid assemblies, which are less affected by the detection of the basal overexpression and targeting the proper amyloid conversion.
Analysis of a set of commercial antibodies reveals overlooked yet interesting features.
Using recombinant hSyn we produced and selected 3 different strains of PFFs which we characterized in previous studies: iso1, iso3, and 1B (14). For this study, we tested a panel of “routine” commercial antibodies on these different PFFs to investigate their possible ability to differentially detect the fibrils strains and the monomeric protein (Fig. 3A).
As expected, MJFR1 (42) and EP1646Y, respectively targeting the C- and the N-terminus of the protein, did not show any preference for the monomeric or the fibril α-Syn assemblies. MJFR14 which has been described as preferentially binding to fibrils (43) did not show any conformation specificity in our experimental settings. In contrast, LB509 (44) which targets the C-terminus (45) recognized both α-Syn forms, but with a clear preference for fibrils. A similar conformation-dependent preference was observed for Syn505 which binds to the 1–12 end of the N-terminus (46), and SynF1 which is targeted to the C-terminus (47) appeared most selective of amyloid fibrils vs. monomers.
While these results are globally in line with the expectations (excepted for MJFR14), we identified two particularly striking – yet overlooked – properties for 2 antibodies widely used in the literature: (i) Syn1 (Clone 42) which targets the NAC region of the protein(48) is often referred to as a “pan-α-Syn” antibody held capable of recognizing all forms of the protein. In fact, we found that it selectively recognized the monomeric protein and that it did not bind to the fibrils. This confirms repeated previous observations made by us and others (14, 49, 50); (ii) D37A6 which is held a rodent specific antibody targeting an upstream region of the C-terminus (surrounding E105), did indeed not recognize the human monomer. However, and most strikingly, it did show an affinity for hSyn engaged into amyloid fibrils.
To further explore these conformation specificities, we used increasing concentrations of urea to provoke the gradual disassembly of the 3 fibrils strains back into monomers and to measure the impact of disassembly on immunoreactivity (Fig. 3B). In agreement with the previous experiment (Fig. 3A), disassembly did not modify the signal of the conformation-independent antibodies MJFR1, EP1646Y and MJFR14, while it collapsed the signal of all the fibril-specific antibodies which we identified, i.e., LB509, Syn505, SynF1 and D37A6. As a mirror image, progressive fibril disassembly led to the progressive appearance of Syn1 immunoreactivity as more monomers were released from the amyloid assemblies, confirming the monomer-selectivity of Syn1.
Finally, we tested the above antibodies on PD and MSA brain homogenates comparing their sarkosyl insoluble and soluble fractions (Fig. 3C and Fig. S3). The results confirm that in a biochemical assay, LB509, SynF1 and Syn505 recognize pathological α-Syn and showed that Syn1 and D37A6 behave as amyloid state-dependent antibodies. Note that to our knowledge, changes in immunoreactivity towards Syn1 is the first amyloid conversion-dependent change of α-Syn taking place at the proper fibril core level and amenable to detection by an immunological method.
Amyloid conversion assays in neurons
Considering these results on synthetic and extracted fibrils, we explored the use of the antibodies which we retained conformation-dependent (see above) in primary cultures of mouse cortical neurons overexpressing hSyn.
We first performed double-immunofluorescence imaging using SynF1 or Syn505 in combination with EP1536Y which detects pS129 α-Syn. To discriminate aggregated α-Syn and its soluble forms, we performed the same staining on neurons permeabilized prior to fixation (Figs. 4A,B).
Both antibodies confirmed their preferential affinity for aggregated α-Syn, the quantification of the signal integral showing a significant increase for each of the antibodies in the PFF-treated neurons compared to the control conditions (Fig. 4A). SynF1 was also sensitive to overexpression, but the differential intensity between PFF-treated and untreated conditions was larger than with LB509 (Compare Fig. 2E and Fig. 4A,B). The signal shown by Syn505 in these experimental conditions was generally weak but provided a more clearcut readout of α-Syn aggregation in the PFF-treated neurons. For both antibodies, permeabilization of the neurons prior to fixation, which allowed the retention of aggregated α-Syn and the release of the non-aggregated forms, improved the difference between the PFF-treated and non-treated conditions, providing a “filtered/contrasted” image of the α-Syn pathology (Figs. 4A).
However, when we considered pS129 co-staining with EP1536Y, we observed that similarly to what we noted with LB509, the neurons positive to the conformational antibodies SynF1 or Syn505, and those positive to EP1536Y only partially overlapped, both in non-permeabilized and permeabilized conditions (Fig. 4B). This suggested that pS129 is not a comprehensive readout of aggregated α-Syn, since in some neurons or in some areas of the same neuron, aggregated α-Syn is not phosphorylated. This could be due to a difference in neuronal types, or, in the second case, it could be due to the “snapshot” of an evolving process inside a single neuron (progressive phosphorylation or dephosphorylation of the aggregates). On the other hand, we also observed neurons bearing pS129-positive aggregates that were insoluble, yet were negative to the conformational antibodies SynF1 or Syn505 (Fig. 4B).
Aiming at detecting/quantifying amyloid conversion in intact neurons irrespective of pS129 we decided to exploit the conformational properties of the Syn1 antibody. Again, the epitope targeted by this antibody (aa91-99) is part of the NAC domain of the protein which gets engaged and is instrumental in the protein-protein interactions established between stacked α-Syn monomers during assembly of the amyloid core. As a result, once the monomers are trapped inside the fibrillar structure, the epitope is no longer accessible to the antibody. Indeed, when we treated AAV-syn-infected neurons with PFFs we observed the appearance of inclusions with a collapse of Syn1 immunoreactivity in inclusions yet positive to MJFR1 (Fig. 4C). Thus, using a double staining with a conformation-independent antibody like MJFR1 (or EP1646Y, not shown) in combination with Syn1, we can easily differentiate soluble α-Syn from amyloid α-Syn assemblies and identify directly the neurons containing fibrils (Fig. 4C). Indeed, in neurons not seeded with PFFs, MJFR1 and Syn1 staining completely overlapped and both signals disappeared if the neurons were permeabilized prior to fixation. This indicates that normal neurons only contain soluble, non-amyloid α-Syn. When neurons were seeded with PFFs instead, a Syn1 negative MJFR1-positive population appeared which was unsensitive to permeabilization, corresponding to the neurons in which α-Syn underwent amyloid conversion (Fig. 4C). Note that amyloid conversion can similarly be tracked using EP1536Y to reveal pS129 α-Syn together with Syn1 (Fig. 4D). In this case, amyloid α-Syn is EP1536Y-positive and Syn1-negative.
In conclusion, pS129 is a surrogate marker of α-Syn aggregation that can generate both false positives and false negatives, especially in condition of α-Syn overexpression. The commercial conformation-dependent antibodies we tested and that do bind preferentially to aggregated α-Syn present only a relative specificity and can miss a significant fraction of the amyloid α-Syn fibrils present in neurons (false negatives). In other words, they are not able to properly discriminate aggregation in situ. Moreover, it should be highlighted that these conformation-dependent antibodies target the arrangements the N or the C terminals which derive only secondarily from the aggregation process and are not involved in the assembly of the amyloid core. In contrast, the loss of accessibility of the NAC epitope recognized by Syn1 is a direct indication of the enrollment of the protein in an amyloid structure and can thus be used in combination with conformation-independent antibodies to specifically highlight amyloid α-Syn assemblies in situ.
High Content Analysis amyloid α-Syn conversion in 96 well primary neuronal cultures: revisiting the impact of α-Syn mutations.
We thus decided to take advantage of this new method for investigating a series of α-Syn mutations – either disease-relevant or preventing S129 phosphorylation – to determine their propensities to act as a stand-alone trigger of α-Syn amyloid conversion inside neurons and/or to accelerate the process once seeded by PFFs (Fig. 5). We infected neurons with 7 distinct AAVs carrying the cDNA of hSyn bearing point mutations that have been associated with the emergence of autosomal-dominant inherited PD: A29E, A30P, E46K, G51D, H50Q, A53E, A53T (51). The mechanisms that link these mutations to PD are generally thought to depend on a facilitation of the amyloid assembly of α-Syn, but although appealing, this assumption is mostly based on historical protein-only observations made in vitro. We also included in our exploration the experimental S129A mutant, coding for a non-phosphorylatable form of the protein at S129 because here also, there is a debate on the functional impact of this phosphorylation on the process of spontaneous or of seeded α-Syn assembly (52–54).
Using MJFR1 we quantified the levels of neuronal overexpression achieved for variants and found that the different AAV infections yielded comparable expression levels, except for A30P which was slightly less expressed than its counterparts, and for E46K which was barely expressed and detectable (this observation was repeated with different AVV-E46K α-Syn production batches, not shown) (Fig. S5). In addition, expression of E46K appeared to be neurotoxic in our primary cultures of cortical neurons, confirming previous observations (55). We thus did not consider or discuss the possible impact of the E46K mutation on the amyloid conversion of α-Syn in situ.
Figure 5A shows that as it was the case for wt hSyn, all disease-associated mutants were significantly phosphorylated in basal conditions under the simple effect of overexpression (compare with S129A which cannot be phosphorylated on this residue). The phosphorylation level in the unseeded conditions was comparable for all disease-associated mutants.
Upon seeding with PFFs, a modest 2–3 fold increase of pS129A was observed for all variants, excepted for S129A which is non phosphorylatable (note that with this scale adapted for overexpression conditions, the phosphorylation of endogenous α-Syn upon seeding exists but is invisible), E46K which showed virtually no response because it was barely expressed, and A30P which exhibited the strongest basal phosphorylation in spite of its lower expression level, with little room left to detect a further impact of seeding on pS129 .
Figure 5A underlines that under overexpression conditions, little can be said on the impact of specific mutations on either basal or seeded α-Syn aggregation compared to wt α-Syn, apart perhaps the basal pS129 levels of A30P which could be interpreted as the indication of a more pronounced spontaneous aggregation of this variant. The next experiments indicate that this is not the case.
In order to track true α-Syn amyloid conversion in these conditions, we used the MJFR1-Syn1 staining combination and derived the % of neurons bearing amyloid inclusions (i.e., bearing MJFR1-positive Syn1-negative inclusions) (Figs. 5B,C). As for wt α-Syn, in unseeded conditions virtually all the neurons were perfectly double stained for all mutants (including S129A), indicative that none of the variants induced spontaneous aggregation with amyloid conversion. This indicates that the pS129 signal observed in unseeded conditions (Fig. 5A) is not due to spontaneous aggregation but simply to overexpression and phosphorylation of non-amyloid forms of the protein. These results reveal that in cortical neurons, the α-Syn mutations that cause autosomal dominant PD fail to trigger the spontaneous amyloid conversion of α-Syn into fibrils. The same holds true for the S129A variant showing that preventing phosphorylation at S129 is not sufficient not trigger spontaneous fibrillization.
We thus reasoned that in a neuronal context, these mutations might instead favor the seeded assembly of α-Syn into amyloid fibrils. Upon seeding with PFFs, we observed a massive burst of the population of neurons bearing amyloid inclusions (MJFR1-positive Syn1-negative) (Figs. 5B,C). Unexpectedly enough, the extent of amyloid conversion here also appeared comparable for all the variants: none of the mutations appeared to favor the seeded assembly process. At the opposite, the A53E variant even inhibited the amyloid conversion of α-Syn in neurons confirming several previous observations made in vitro regarding this mutant (56, 57). Collectively, these data suggest that the mechanisms by which familial α-Syn mutations might cause autosomal dominant PD are neither related to triggering nor to facilitation of α-Syn fibrillization in intact neurons. In addition, phosphorylation of S129 which can be misleading as a marker of fibrillization, does not seem to inhibit spontaneous or seeded α-Syn fibrillization either.
Probing amyloid conversion in histological brain sections: revisiting the status of α-Syn in pathological inclusions
These results prompted us to put amyloid α-Syn conversion under scrutiny in brain sections presenting clear signs of α-Syn inclusion pathology. We used in parallel sections from wt mice sacrificed 6 months after an intra-striatal injection of human PFFs and post-mortem sections from a patient with sporadic PD (Fig. 6). In mice (Fig. 6A-C), the inclusion pathology was revealed using the antibody pair pS129 and Syn1 like in the primary cultures of Fig. 4D. The figure shows a region of the right basolateral amygdala (BLA) filled with pS129-positive α-Syn neuronal inclusions of 3 types: LNs, neuronal perikaryal inclusions with a more or less compacted “Lewy Body-like” appearance, depending on the maturation level of the inclusion, and a few neuronal intranuclear inclusions (Fig. 6A, in green). Co-staining with Syn1 (Fig. 6B, in purple) revealed the physiological pool of non-amyloid α-Syn present in the synapses that were scattered all over the field of view. However, while as expected many pS129 inclusions appeared Syn1-negative (green inclusions in the overlay of Fig. 6C, see a few examples pointed by empty arrowheads), indicating inclusions exclusively made of amyloid α-Syn, a significant number of inclusions alternatively appeared partially or totally Syn1-positive (white inclusions in the overlay of Fig. 6C, see a few examples pointed by empty arrows), indicative of the presence of non-amyloid α-Syn inside the inclusions. It can be noted that the presence of non-amyloid α-Syn in experimental inclusions seeded in mice by PFFs does not seem to depend on the inclusion type since all of them can be concerned by the possible presence of non-amyloid α-Syn detectable by Syn1.
Panels of Fig. 6D-J and K-Q show the focused exploration of the amyloid status of α-Syn respectively in a LN and in a Lewy body both present in a post-mortem SN brain section of a sporadic PD patient. The section was first double-labeled and revealed using EP1536Y and Syn1 as before. The section was then hybridized with a third fluorophore-coupled MJFR1 antibody which detects hSyn irrespective of its conformation and phosphorylation (see Figs. 3, 5 and S5). The LN of Fig. 6 is pS129-positive on most of its length (green) with a superposable staining pattern shown by MJFR1 (red) (see in particular the pS129 and MJFR1 traces in the line scan of the overlay Fig. 6J). Instead, staining with Syn1 (purple) is variable along the length of the LN: some regions are Syn1-negative indicating a purely amyloid composition (empty arrowhead), while others are Syn1-positive showing the presence of non-amyloid α-Syn (empty arrows). The line scan of Fig. 6J makes it clear that the MJFR1 and the Syn1 signals were completely decoupled in the left end of the LN highlighting the exclusive presence of fibrils in this portion of the inclusion. At major variance, it appeared difficult to identify amyloid-only subregions in the LB. The 3 antibodies showed comparable staining patterns within the inclusion, with spatially correlated MJFR1 and Syn-1 signals (see the MJFR1 and Syn1 traces in the line scan of the overlay Fig. 6Q and compare with Fig. 6J). In contrast with the somatic inclusions seeded in mice (Fig. 6A-C), and with the LN shown before, Syn1 positivity was observed for all the LBs we investigated. This indicated that LBs in PD are pathological inclusions containing non-amyloid α-Syn. This is in line with the poor amyloid status of PD brain extracts compared to MSA ones [see companion paper Lafrerrière et al.] and confirms the pioneering observations of Shahmoradian and colleagues who reported that α-Syn fibrils were identifiable in many but not all α-Syn inclusions in PD (58).