Pathological α-synuclein triggers synaptic NMDA receptor dysfunction through altered trafficking


 Background

α-Synuclein misfolding and aggregation contribute to synaptic dysfunction in synucleinopathies, including Parkinson’s disease. However, the mechanism underlying the effect of α-synuclein on synaptic components remains unclear. Since the N-methyl-D-aspartic acid receptor (NMDAR) plays a key role in glutamate synapse pathophysiology, we here investigated its surface dynamics and functional distribution in neurons exposed to various pathological α-synuclein forms.
Methods

A combination of single-molecule tracking, immunochemistry, immunoblot and calcium imaging approaches were used to assess the changes in NMDAR membrane dynamics and functions. The NMDAR alterations were evaluated in rat cultured hippocampal networks, in which α-synuclein mutants were overexpressed or exposed to α-synuclein proteins (monomeric/PFF α-synuclein). The surface dynamics of NMDAR subtype was artificially tuned in order to test its instrumental role.
Results

We observed that mutant α-synuclein (A53T-α-synuclein) restricted NMDAR surface trafficking and impaired synaptic function. In contrast, wild-type α-synuclein did not affect synaptic NMDAR. Further, we found that chronic exposure to α-synuclein preformed fibrils induced molecular dysfunctions that mainly targeted the GluN2B-NMDAR subtype. The deficits of synaptic NMDAR have also been found in A53T transgenic mice α-synuclein. Upon fine-tuning of the surface dynamics of GluN2B-NMDAR, pathological α-synuclein gradually lost its synaptic toxicity.
Conclusions

Our findings indicate that pathological α-synuclein alters GluN2B-NMDAR synaptic dynamics and organization, which leads to glutamate synapse dysfunction.

Moreover, α-Syn is found in the postsynaptic compartment; however, its function and involvement in synaptopathies remain unclear.

Animals
Heterozygote A53T-α-syn transgenic mice (B6; C3-Tg (Prnp-SNCA*A53T) 83Vle/J) were originally obtained in breeding pairs from the Jackson Laboratory (004479). Wild-type mice were C57BL/6J purchased from Liaoning Changsheng biotechnology.co., Ltd (China). All mice in the colony were kept under speci c pathogen-free (SPF) conditions in a 14 h light /10 h dark cycle, and had free access to food and water. At 10 months, mice were scari ed for protein detections.
Neuronal electroporation was performed following the commercial instructions, using the P3 Primary Cell 4D-NucleofectorTM X Kit (LONZA, #V4XP-3024) and 0.75 µg of plasmid DNA (0.25 µg of Homer1C-Dsred and 0.5 µg of WT-/A53T-α-syn) to prepare the electro-buffer. The neurons (5 × 10 5 ) were resuspended in electro-buffer and transferred to a Single Nucleocuvette™. Using the Amaxa Nucleofector (Lonza), the cells were electroporated with program DC-104. Afterward, the cells were plated on the coverslips and cultured at the 37 °C in 5% CO 2 for 14 days of overexpression before use.

α-Synuclein treatments
The overexpression of α-syn (WT/A53T) was processed by either calcium phosphate transfections or the electroporations. Cells at DIV 9-11 were transfected with 0.5 µg WT-/A53T-α-syn plasmids via the calcium phosphate method for four days of expression. Or dissociated neurons were electroporated with 0.5 µg of WT-/A53T-α-syn plasmids and expressed for 14 days after seeding. To calculate the relative overexpression, we utilized antibody against α-syn to label the α-syn expression in transfected neurons, and the non-transfected neurons in the same coverslip were as basal condition. Then we quanti ed uorescence intensity of the α-syn positive signal in every whole neuron, including both cell bodies and dendrites. To perform quanti cations, 10 cells were chosen from each condition.

Single quantum dots tracking and analysis
Hippocampal neurons at DIV 13-14 were rstly incubated 10 min with polyclonal antibodies against green uorescence protein (GFP, Invitrogen, Thermo Fisher 1:100 000) to detect overexpressed GluN2Α/2B-sep-NMDAR, followed by 10 min of incubation with QD 655 Goat F(ab')2 anti-rabbit polyclonal antibodies (Invitrogen, Thermo Fisher Scienti c Inc.1:50 000). All the cells were incubated in the neurobasal medium supplemented with 1% bovine serum albumin (BSA) at 37 °C. Coverslips containing neurons were then mounted with Tyrode solution (30 mM D-glucose, 120 mM NaCl, 5 mM KCl, 2 mM MgCl 2 , 2 mM CaCl 2 , and 25 mM HEPES, pH 7.3-7.4) as described previously [38] and placed on a heated chamber for observations. Using a mercury lamp and appropriate excitation/emission lters, QD signals were detected with an electron multiplying charged-coupled device (EMCCD) camera (Evolve, Photometrics). Images were obtained every 50 ms (20 Hz) with up to 500 consecutive frames and processed using the Metamorph software (Universal Imaging Corporation, PA, USA). In each recording session, 3-4 structured dendritic areas were obtained from every coverslip obtained, and the procedure lasted for up to 20-25 min. α-Syn proteins (monomer or PFF) were incubated for 20 min before QD labeling and acquisitions. The calculations of single QD tracking followed the formulas and the rules described previously [38,39]. Brie y, the instantaneous diffusion coe cient 'D' was calculated for each trajectory, from linear ts of the rst four points of the mean-square-displacement versus time function using MSD(t) = < r2 > (t) = 4Dt. Using a Vogel algorithm, the two-dimensional trajectories were constructed by correlation analysis between consecutive images. This technique offers high resolution and is capable of measuring the dynamic distribution of GluN2B/2Α-NMDAR at synaptic and extra-synaptic sites. A synaptic area was de ned as the area with the transfected synaptic marker Homer1C-Dsred. The diffusion coe cient was calculated from GluN2-QD trajectories were present on both the inside and outside of the synaptic area.
To detect the endogenous and overexpressed proteins levels, neurons at DIV14 were xed with 4% paraformaldehyde (Sigma-Aldrich, #P6148) in phosphate-buffered saline (PBS) containing 4% sucrose (Sigma-Aldrich, #0389) for 15 min at room temperature (RT), followed by the 5 min of permeabilization with 0.25% Triton X-100 in PBS. Non-speci c signals were blocked by incubation with 10% BSA (SIGMA) in PBS for 1 h, after which the cell were incubated with primary antibodies for 2-3 h at RT, followed by the secondary antibodies for 1 h at RT. To label the contents of surface glutamatergic receptors (GluN2Α/2Bsep-and GluA1-sep-subunit), dissociated hippocampal neurons were live immunostained for 20 min, using anti-GFP antibody at 37 °C. After the xation, the cells were incubated with secondary antibodies for 1 h at RT. All the antibodies were prepared in 1% BSA in PBS. All samples were washed at least three times with PBS between the incubations, nally mounted in the Mowiol mounting medium.

Fluorescence microscopic imaging
All the neuronal images were acquired using a Photometrics Quantem 512 camera (EMCCD) and MetaMorph imaging software (Molecular Devices) on an inverted confocal spinning-disk microscope (Leica DMI6000B, Leica), using a Leica HCX PL APO CS 60×/1.4 or 63×/1.4 oil objective. In each individual experiment, images were captured in the same settings (the same laser intensity and exposure time) across all the samples. Images were exported in the grey scale from individual channels and pseudocolor overlays were prepared using ImageJ software. To perform quanti cations, 9-11 cells were chosen from each condition from every independent experiment were chosen. For each neuron, two or three dendritic regions were selected for analysis.

Calcium imaging
The experiments were performed similarly as formerly described [41]. Brie y, treated hippocampal neurons were transfected with GCaMP6 at DIV 9-10 for four days and transferred to Tyrode solution at least 3 h before acquisition at DIV 13-14. Before imaging, the cells were incubated in Mg 2+ -free Tyrode solution supplemented with 5 µM Nifedipine (Tocris) and 5 µM Bicuculline (Tocris) for 15 min. Time-lapse images were obtained every 50 ms for up to 3000 frames. Thereafter, 50 µM D-AP5 was applied directly to the chamber to abolish the calcium transient events. A movie containing another 3000 time-lapse frames was acquired as the baseline for each neuron.
Synaptosomes components were puri ed from homogenized hippocampus of 10-month-old WT-and A53T-α-syn transgenic mice, following the methods as previously described [39]. Hippocampal tissues (around 20 mg) were thawed in 300 µL fresh TPS buffer (0.32M sucrose 4 mM HEPES buffer, pH7.4) supplemented with a protease inhibitor cocktail with 15 strokes in glass-Te on homogenizer (500 µL), on the ice. After centrifugation at 1000 g for 8 min at 4℃, the pellet was discarded and the supernatant was centrifuged again at 13,500 g for 15 min at 4 °C. The resulting pellet was washed by TPS buffer once and resuspended in RIPA lysis buffer supplemented with protease inhibitor cocktail and PMSF for 15 min lysis. Finally, samples were collected and kept at − 80℃ for storage.

Ethics Statement
Animal experimentation: All experiments were carried out in accordance with the guidelines and regulations of University of Bordeaux and Northeastern University. The animal procedures were approved by the ethical committee of the University of Bordeaux (Laurent Groc experimentation authorization number 3306009) and Northeastern University, China.

Data and statistical analysis
All data and statistical analysis was performed by GraphPad Prism 8.00 (GraphPad Software, Inc., La Jolla, CA). Most non-normal data, such as single-particle tracking and immunocytochemistry of the data were assessed by Mann-Whitney test to compare differences between two groups and by Kruskal-wallis test followed by Dunn's Multiple Comparison test between three groups. Normal data such as expression levels using unpaired student's t test. Signi cance was assessed at p < 0.05 using two-tails tests unless otherwise speci ed in the gure legends. Symbols used are: *p < 0.05; **p < 0.001; ***p < 0.0001 throughout the manuscript.

Results
Wild-type-α-syn overexpression preserves NMDAR surface tra cking and distribution Wild-type-α-syn (WT-α-syn) is the physiological α-syn form with concentration-and genetic expressiondependent toxicity [42,43]. Given the decreased hippocampal GluN2A/GluN2B subunit ratio in animals with PD [24], we rst investigated whether WT-α-syn overexpression could alter GluN2A-and GluN2B-NMDAR surface tra cking in hippocampal neurons in active networks. Between 9 and 11 days in vitro, hippocampal neurons were co-transfected with WT-α-syn, Homer1C-Dsred, and/ or GluN2Α/2B-sep (sep, superecliptic pHluorin) (Fig. 1). WT-α-syn overexpression led to three-fold α-syn enrichment ( Fig. 1a and b, details see Methods). Using antibodies against tagged GluN2 subunits, we performed single-nanoparticle (quantum dot [QD]) tracking on transfected neurons (Fig. 1c). WT-α-syn overexpression did not signi cantly affect the membrane tra cking of synaptic GluN2A-NMDARs; speci cally, there was a slight change in the diffusion coe cients and no effect on the explored area ( Fig. 1d and f; Supplemental  Fig. 1a ). Moreover, GluN2B-NMDARs maintained similar synaptic surface-tra cking patterns in the presence of WT-α-syn ( Fig. 1e and g); however, there was a mild increase in the diffusion of extra-synaptic receptors (Supplemental Fig. 1b). This indicated that WT-α-syn overexpression had no, or very mild, effects on GluN2A-and GluN2B-NMDAR surface tra cking. To determine whether WT-α-syn overexpression altered the synaptic composition of these NMDAR subtypes, we performed live immunostaining of either GluN2Α-or GluN2B-NMDARs in hippocampal neurons with identi ed glutamatergic synapses based on the presence of Homer1C-Dsred clusters ( Fig. 1h and i). Similar to the minimal effects on NMDAR subtype membrane dynamics, there were no changes in the synaptic GluN2A-  Fig. 2a, b, d and e). To further con rm this observation, we measured synaptic NMDAR-mediated transmission through spontaneous calcium transients in cultured hippocampal neurons. Based on the neuronally expressed calcium indicator GCaMP6, we imaged calcium events in the presence of an L-type voltage-dependent calcium channel blocker (see Methods). The calcium events were fully stopped by NMDAR-competitive antagonist D-AP5 (50 µM), which indicated that they mostly originated from the NMDAR. In neurons that overexpressed WT-α-syn, there was an unaffected frequency of NMDARmediated calcium transients and it was slightly increased compared with the spine populations (Supplemental Fig. 3). Taken together, these ndings indicate that a 3-fold increase in WT-α-syn only minimally, if at all, alters NMDAR membrane tra cking and synaptic content.
Mutant A53T-α-syn alters NMDAR surface tra cking and content A53T-α-syn is among the earliest reported mutations in synucleinopathies and is known to favor rapid pathological aggregations [44]. Moreover, it has been shown to cause severe movement disorders in A53T transgenic mice [45,46] and to be involved in early-onset familial PD [44]. Furthermore, it has been recently shown to cause learning, memory, and synaptic plasticity de cits [47,48]; speci cally, only A53T-mutant, but not WT-or A30P-mutant, α-syn can induce postsynaptic dysfunction in transgenic mice [47]. Consequently, we assessed whether A53T-α-syn could alter NMDAR membrane tra cking and synaptic content. First, we ensured that neuronal expression of WT-and A53T-α-syn were within the same range ( Fig. 2a and b, Supplemental Fig. 4). Subsequently, using single-nanoparticle tracking, we assessed whether A53T-α-syn could alter NMDAR surface tra cking. Compared to WT-α-syn, A53T-α-syn signi cantly decreased GluN2Α-NMDAR surface diffusion ( Fig. 2d and f right). To further examine GluN2Α-NMDAR diffusion patterns, we plotted the mean square displacement (MSD) versus time lag. Compared to WT-α-syn, A53T-α-syn shifted the MSD curve of GluN2Α-NMDAR, which indicates a more con ned diffusion pattern and smaller explored area (Fig. 2f left). Moreover, A53T-α-syn signi cantly reduced GluN2B-NMDAR surface dynamics; however, this was to a smaller extent ( Fig. 2e and g). The similar surface tra cking changes in GluN2A-and GluN2B-NMDARs were mainly observed in the synaptic compartment; contrastingly, extra-synaptic receptors had minimal changes (Supplemental Fig. 1c and d). Taken together, these results show that A53T-α-syn downregulates synaptic GluN2Α-and GluN2B-NMDAR surface tra cking, which could involve previously de ned aberrant interactions with synaptic partners [20,40].
These altered receptor synaptic dynamics suggest that their retention is possibly also altered. Consequently, we evaluated the synaptic content of GluN2A-and GluN2B-NMDARs using live immunocytochemistry on hippocampal cells. A53T-α-syn-expressing neurons had signi cantly reduced GluN2A-NMDAR synaptic content; contrastingly, the total membrane amount remained stable ( Fig. 2h and  j, Supplemental Fig. 2a and c). Consistently, A53T-α-syn overexpression signi cantly decreased the GluN2B-NMDAR synaptic, but not surface, content ( Fig. 2i and k, Supplemental Fig. 2d and f). To con rm that A53T-α-syn impaired NMDAR synaptic transmission, we recorded NMDAR-mediated spontaneous calcium transients in WT-and A53T-α-syn-expressing neuron. Consistent with the imaging data, A53T-αsyn-expressing neurons had a signi cantly decreased frequency of NMDAR-mediated spontaneous calcium events (Fig. 3b-d). Compared to WT-α-syn neurons, A53T-α-syn-expressing neurons had a leftshifted cumulative distribution of calcium transient frequency (Fig. 3e). To test whether the A53T-α-syn effect on synaptic NMDARs is speci c for this receptor type, we assessed the GluA1-AMPAR synaptic content under similar conditions. Unlike in NMDARs, A53T-α-syn did not alter the surface or synaptic content of GluA1-AMPAR (Fig. 4). Taken together, our ndings indicate that A53T-α-syn expression downregulates GluN2Α-and GluN2B-NMDAR surface tra cking, as well as their synaptic retention and content, while leaving the AMPAR component intact. α-Syn preformed brils (PFFs) speci cally impair GluN2B-NMDAR membrane dynamics In sporadic PD, extracellular α-syn, including its brillary form, is considered the "toxic seed" and is crucially involved in PD pathogenesis and propagation [1,49,50]. Through prion-like propagation [51,52], toxic proteins are thought to act pre-and post-synaptically in neurotransmission and plasticity impairment. Toxic extracellular α-syn impairs glutamatergic synapse LTP through NMDAR alterations [10-12, 27, 34, 36]. Since mutated α-syn impairs synaptic NMDARs (see above), acute neuron exposure to α-syn PFFs (a toxic form of extracellular α-syn [49]) could cause similar NMDAR dysfunction. To assess this question, we examined NMDAR surface tra cking after acute neuron exposure to two different α-syn forms, i.e., monomers or PFFs (concentrations see Methods). Both forms did not alter GluN2A-NMDAR surface diffusion (Fig. 5A upper, B). Notably, α-syn PFFs signi cantly reduced GluN2B-NMDAR surface diffusion in the synaptic and extrasynaptic compartments (Fig. 5a lower and c). Moreover, α-Syn PFFs affected the MSD curve of GluN2B-NMDAR (Fig. 5c left), which indicated that PFFs con ned these receptors more. GluN2A-and GluN2B-NMDAR surface and synaptic content remained stable (Fig. 5d-g,  Supplemental Fig. 5a-c). However, α-syn PFFs signi cantly increased the global surface content of GluN2B-NMDARs (Supplemental Fig. 5d-f). Taken together, these ndings suggest that acute α-syn PFF administration downregulates GluN2B-NMDAR, but not GluN2A-NMDAR, surface tra cking. This implies a difference in the effect of α-syn PFFs on NMDAR sensitivity or signaling, which is consistent with previous reports [10,11,35]. To further study the effects of α-syn PFFs on NMDAR synaptic transmission, we evaluated the NMDAR-mediated spontaneous calcium transients in neurons exposed to acute α-syn PFF or monomer treatment. There was no signi cant difference in the frequency of calcium transient events and distributions between PFFs and monomers, which indicated that both did not affect NMDARmediated transmission (Fig. 5h-j). Overall, acute α-syn PFF administration could have acute and selective effects on GluN2B-NMDAR surface tra cking and expression.

Chronic neuron exposure to α-syn PFF decreases NMDAR content and function
Since the effect of α-syn aggregation on synaptic activity and plasticity occurs in a time-and dosedependent manner [49,50], the non-effect of acute α-syn PFF exposure on synaptic NMDAR could have resulted from the short exposure time. To check this, we extended the α-syn PFF incubation time to 48 h and 96 h and used calcium imaging and immunochemistry to examine the synaptic NMDARs. After incubation for 48 h, there was a tendency of decreased frequency of NMDAR-mediated calcium transients, which was indicated by the signi cantly left-shifted cumulative distributions between monomer and PFF (Fig. 6b). Similarly, this was observed after incubation for 96 h; moreover, PFF signi cantly decreased the NMDAR spontaneous calcium transients ( Fig. 6c and d). These results indicate that chronic (96 h) exposure to α-syn PFFs reduces NMDAR synaptic content and transmission. These ndings were further con rmed via immunochemistry, which indicated decreased GluN2A-and GluN2B-NMDAR synaptic content in neurons exposed to α-syn PFFs compared to those exposed to α-syn monomers (Fig. 6e, f right, g, and h right). Notably, there was no effect on GluN2A-and GluN2B-NMDAR membrane contents, which further indicates speci c synaptic alterations (Fig. 6f left, h left, Supplemental  Fig. 6). Taken together, our ndings indicate that α-syn PFF downregulates NMDARs synaptic content, which is consistent with the deleterious effects of A53T-α-syn.
Hippocampal GluN2A-and GluN2B-NMDAR de cits in A53T-α-syn transgenic mice To determine whether pathological α-syn induced synaptic NMDAR alterations in an animal model of synucleinopathies, we used Western blotting to quantify hippocampal NMDAR contents in A53T-α-syn transgenic mice (10-month-old). We observed a signi cant decrease in GluN2A-and GluN2B-NMDAR contents in A53T-α-syn transgenic mice; however, there was no effect on obligatory GluN1 subunits.
Modulation of GluN2B-NMDAR surface dynamics rescues the α-syn-mediated effect Since we found that the pathological α-syn forms (A53T and PFF) alter NMDAR membrane tra cking and synaptic content, we speculated that these effects could be prevented by NMDAR tra cking, but not ionotropic activity, modulation. First, we con rmed that pathological α-syn expression led to abnormal αsyn accumulation and deposition on the cell body [53] (Fig. 8b and c) and glutamatergic synapse loss [54][55][56] (Fig. 8d and e). Since pathological α-syn reduces NMDAR surface tra cking, especially in the GluN2B-NMDAR subtype (Fig. 5), we assessed whether arti cial increasing GluN2B-NMDAR surface dynamics could rescue the synaptic loss mediated by pathological α-syn. To achieve this, we increased GluN2B-NMDAR membrane dynamics using a competing peptide (TAT-2B) that prevented the interaction between the GluN2B-NMDAR C-tail and PDZ scaffold proteins, as previously described [38][39][40]57]. Although the control non-sense peptide (TAT-NS) did not alter synaptic maker density in A53T-α-synexpressing neurons, TAT-2B signi cantly increased the synaptic density (both pre-and postsynaptic markers were consistently increased (Fig. 8f). Therefore, increasing GluN2B-NMDAR surface dynamics over several days could prevent A53T-α-syn-mediated synaptic loss, which suggests that NMDAR disorganization by pathological α-syn mediated cellular toxicity and pathology.

Discussion
Synaptic dysfunctions could re ect early events in synucleinopathies before neuronal death [10,47,48,58]. Therefore, there is a need to understand the molecular mechanism(s) underlying this early effect on synapses. Using a combination of immunocytochemistry, calcium imaging, and single-molecule tracking techniques, we found that pathological α-syn, i.e., A53T-α-syn, impaired NMDAR surface tra cking, and synaptic content. Moreover, a chronic neuron exposure to α-syn PFFs downregulated NMDAR synaptic tra cking and content; further, there was an early speci c effect on GluN2B-NMDAR. Arti cially increasing GluN2B-NMDAR surface tra cking to counteract the effect of pathological α-syn prevented the deleterious effect of α-syn on synaptic loss. Taken together, our ndings support the model where αsyn mutations or multimeric conformation alters NMDAR synaptic organization and function. Consequently, this could affect various NMDAR-dependent processes, including synaptic viability. Therefore, our ndings could lead to the development of novel therapeutic strategies for circumventing NMDAR-dependent synaptic de cits associated with synucleinopathies.
There is increasing evidence that A53T-α-syn, but not WT-α-syn or other mutations, induces postsynaptic impairment [47,48], α-syn inclusion, [45,61] and neurotoxicity [45,46]. Furthermore, individuals with A53T mutation developed a severe PD that is often associated with dementia [62]. This is further indicated by our observation that A53T-α-syn signi cantly decreased NMDAR. Similarly, chronic exposure to α-syn PFFs downregulated NMDAR synaptic retention and spontaneous activity. This indicates that A53T and chronic exposure to PFFs might modify NMDARs via similar mechanisms. For example, A53T prefers aggregation [44,45,61] and four-day A53T expression allows aggregate formation similar to those induced by PFF. Moreover, oligomeric α-syn has been reported to interact with the membrane protein PrP C and form complexes with GluN2B-NMDAR, which leads to impaired synaptic function [10]. Decreased NMDAR membrane dynamics and increased GluN2B-NMDAR surface retention after acute treatment with PFFs supports the hypothesis that PFF induces extracellular signaling via GluN2B-NMDAR to impair synaptic function. We reported that PFF α-syn cause NMDAR dysfunction within 4 days, in line with the previously-described seeding effect of PFF α-syn 4 days after administration [49,63]. Altogether, these evidences suggest that α-syn, aggregated or not, could impair NMDAR synaptic function.
Additionally, GluN2-PSD95 dissociation e ciently promotes neuronal viability in neuropathology [57,69,70]. Taken together, these previous ndings suggest that PSD95 could be a candidate for NMDAR modulation in pathological α-syn conditions. There is a need for studies on the effect of pathological αsyn on scaffold proteins (e.g. PSD95) and potential α-syn partners.