α-Synuclein plasma membrane localization correlates with cellular phosphatidylinositol polyphosphate levels

The Parkinson’s disease protein α-synuclein (αSyn) promotes membrane fusion and fission by interacting with various negatively charged phospholipids. Despite postulated roles in endocytosis and exocytosis, plasma membrane (PM) interactions of αSyn are poorly understood. Here, we show that phosphatidylinositol 4,5-bisphosphate (PIP2) and phosphatidylinositol 3,4,5-trisphosphate (PIP3), two highly acidic components of inner PM leaflets, mediate PM localization of endogenous pools of αSyn in A2780, HeLa, SK-MEL-2, and differentiated and undifferentiated neuronal SH-SY5Y cells. We demonstrate that αSyn binds to reconstituted PIP2 membranes in a helical conformation in vitro and that PIP2 synthesizing kinases and hydrolyzing phosphatases reversibly redistribute αSyn in cells. We further delineate that αSyn-PM targeting follows phosphoinositide-3 kinase (PI3K)-dependent changes of cellular PIP2 and PIP3 levels, which collectively suggests that phosphatidylinositol polyphosphates contribute to αSyn’s function(s) at the plasma membrane.


Introduction
Aggregates of human a-synuclein (aSyn) constitute the main components of Lewy body inclusions in Parkinson's disease (PD) and other synucleinopathies 1 . aSyn is expressed throughout the brain and abundantly found in presynaptic terminals of dopaminergic neurons, where it is involved in synaptic vesicle clustering and trafficking 2 . Whereas isolated aSyn is disordered in solution, residues 1-100 adopt extended or kinked helical conformations upon binding to membranes containing negatively charged phospholipids 3 . Complementary electrostatic interactions between lysine residues within aSyn's N-terminal KTKEGV-repeats and acidic phospholipid headgroups align these a-helices on respective membrane surfaces 4 .
Membrane curvature 5 , lipid packing defects 6,7 and fatty acid compositions 8,9 act as additional determinants for membrane binding. aSyn remodels target membranes 10,11 , which likely relates to its biological function(s) in vesicle docking, fusion and fission 2 . Furthermore, aSyn multimerization and aggregation may initiate at membrane surfaces, which holds important ramifications for possible cellular scenarios in PD 9 . Early aSyn oligomers bind to and disrupt cellular and reconstituted membranes 12,13 , whereas mature aggregates are found closely associated with membranous cell structures and intact organelles in cellular models of Lewy body inclusions 14 and in post-mortem brain sections of PD patients 15 .
Phosphatidylinositol phosphates (PIPs) are integral components of cell membranes and a universal class of acidic phospholipids with key functions in biology 16 . Reversible phosphorylation of their inositol headgroups at positions 3, 4 and 5 generates seven types of PIPs, which act as selective binding sites for folded and disordered PIP-interaction domains 17 .
In eukaryotic cells, PIPs make up less than 2% of total phospholipids with phosphatidylinositol 4,5-bisphosphate, PI(4,5)P2, or PIP2 hereafter, as the most common species 18 . PIPs function as core determinants of organelle identity 19 . PIP2 is predominantly found at the inner leaflet of the plasma membrane (PM), where it acts as a signaling scaffold and protein-recruitment 4 platform 20 . Carrying a negative net charge of -4 at pH 7 renders it more acidic than other cellular phospholipids such as phosphatidylserine (net charge -1) or phosphatidic acid (net charge -1) 21 .
Disordered PIP2-binding domains contain stretches of polybasic residues that establish complementary electrostatic contacts with the negatively-charged phosphatidylinositolphosphate head-groups 18 reminiscent of how aSyn KTKEGV-lysines interact with acidic phospholipids 22 . Indeed, aSyn has been shown to bind to reconstituted PIP2 vesicles in vitro 23 .
Here, we set out to investigate whether native aSyn interacted with plasma membrane PIP2 and PIP3 in mammalian cells. Using confocal and total internal reflection fluorescence microscopy, we show that endogenous aSyn forms discrete foci at the PM of human A2780, HeLa, SH-SY5Y and SK-MEL-2 cells and that the abundance and localization of these foci correlate with pools of PM PIP2 and PIP3. We further delineate high-resolution insights into aSyn interactions with reconstituted PIP2 vesicles by nuclear magnetic resonance (NMR) spectroscopy and establish that aSyn binds PIP2 membranes in its characteristic helical conformation. 5

PM localization of endogenous aSyn
To determine the intracellular localization of aSyn, we selected a panel of human cell lines (A2780, HeLa, SH-SY5Y and SK-MEL-2) that expressed low but detectable amounts of the endogenous protein. Confocal immunofluorescence localization in A2780 cells with an antibody that specifically recognizes aSyn without cross-reacting with its b-and g-isoforms (Figure 1 -figure supplement 1A), revealed a speckled intracellular distribution with distinct aSyn foci at apical and basal PM regions ( Figure 1A). We verified overall antibody specificity by downregulating aSyn expression via siRNA-mediated gene silencing, which established that aSyn foci corresponded to endogenous protein pools ( Figure 1B and To investigate colocalization of aSyn with PM PIP2, we co-stained A2780 cells with antibodies against aSyn and PIP2 ( Figure 1C). In 10-20% of cases, we detected clear superpositions of aSyn and PIP2 signals, which we confirmed by measuring fluorescence intensity profiles over individual cell cross-sections ( Figure 1D). To test whether changes in cellular PIP2 levels affected aSyn abundance at the PM, we transiently over-expressed green fluorescent protein (GFP)-tagged phosphatidylinositol-4-phosphate 5-kinase PIPKIg 25 .
PIPKIg localizes to the PM via a unique di-lysine motif in its activation-loop 26 . Upon kinase expression, confirmed by GFP fluorescence, we detected increased amounts of aSyn at the PM of transfected cells ( Figure 1E). By contrast, expression of GFP alone did not alter aSyn levels.
We obtained similar results in HeLa and SH-SY5Y transfected cells ( Figure 1E and Figure 1 -figure supplement 1C). These findings suggested that PM localization of endogenous aSyn correlated with cellular PIP2 levels. To better resolve the presence of aSyn at the PM, we resorted to total internal reflection fluorescence (TIRF) microscopy. Employing a narrow evanescent field depth of ~50 nm, we detected endogenous aSyn at PM foci in A2780, HeLa, 6 SH-SY5Y and SK-MEL-2 cells, which correlated with the abundance of total aSyn determined by semi-quantitative Western blotting (Figure 1 -figure supplement 1D).  To test whether aSyn directly bound PIP2 membranes under physiological salt and pH conditions (150 mM, pH 7.0), we added N-terminally acetylated, 15 N isotope-labeled aSyn to reconstituted PIP2 vesicles. Circular dichroism (CD) spectroscopy revealed characteristic helical signatures 27,28 (Figure 2A), whereas NMR experiments confirmed site-selective linebroadening of N-terminal residues 1-100, confirming membrane binding 29,30 (Figure 2B and (I/I0) of unbound and PC-PIP2-bound aSyn, we found that residues 1-10 constituted the primary interaction sites, whereas residues 10-100 displayed progressively weaker membrane contacts.
In agreement with our experiments on PIP2-only vesicles, we detected no contributions by Cterminal aSyn residues. These findings confirmed the tri-segmental nature of aSyn-PIP2 9 interactions and the importance of anchoring contacts by N-terminal aSyn residues, similar to other membrane systems 29, 31, 32 .   To further validate our conclusions, we performed NMR experiments with mutant forms of aSyn in which we deleted residues 1-4 (DN) 33 , substituted Phe4 and Tyr39 with alanine (F4A-Y39A) 34 , or oxidized aSyn Met1, Met5, Met116 and Met123 to methionine-sulfoxides (MetOx) 35  To investigate the reversibility of aSyn-PIP2 interactions, we prepared PC-PIP2 vesicles bound to 15 N isotope-labeled aSyn to which we added catalytic amounts of unlabeled PLC. We reasoned that PLC will progressively hydrolyze PIP2 binding sites and, concomitantly, release aSyn. In turn, we expected to observe an increase of aSyn NMR signals corresponding to the fraction of accumulating, unbound protein molecules. Indeed, we detected the recovery of aSyn NMR signals upon PLC addition ( Figure 2E and  Next, we asked whether aSyn binding to PC-PIP2 vesicles was sensitive to calcium, a competitive inhibitor of many protein-PIP2 interactions 38 . Whereas overall binding was greatly reduced, we found that the first ten residues of aSyn displayed residual anchoring contacts with PC-PIP2 vesicles even at high (2.5 mM) calcium concentrations ( Figure 2E and

aSyn-PM localization correlates with changes in PIP2-PIP3 levels
Following these results, we asked whether reversible aSyn-PIP2 interactions were present in cells. To answer this question, we transiently overexpressed different PM-targeted PIP phosphatases in A2780 cells and quantified PM localization of endogenous aSyn by confocal immunofluorescence microscopy ( Figure 3A). Specifically, we expressed MTM1-mCherry-CAAX, which targets PI(3)P to yield phosphatidylinositol (PI), INPP5E-mCherry-CAAX to produce PI(4)P from PIP2, and PTEN-mCherry-CAAX to create PIP2 from PI(3,4,5)P3, as described 39 . In agreement with our hypothesis, only the conversion of PIP2 to PI(4)P by INPP5E led to a marked reduction of endogenous aSyn at the PM ( Figure 3A).
Together with earlier kinase results, these findings corroborated that PM localization of cellular aSyn was modulated by PIP2-specific enzymes. Next, we asked whether signaling-dependent with the accumulation of bundled Actin fibers and in line with expected PI3K-signalingdependent rearrangements of the cytoskeleton 24 . Strikingly, aSyn colocalization with these peripheral PIP2-PIP3 speckles was significantly higher than at earlier time-points ( Figure 3B and Figure 3 -figure supplement 1A). We independently confirmed these results with single time-point measurements by TIRF microscopy (Figure 3 -figure supplement 1B). To investigate whether other PI3K pathways caused similar effects, we stimulated SK-MEL-2 with insulin, which triggers PI3K activation via receptor tyrosine kinase (RTK) signaling 42 . We verified that SK-MEL-2 cells endogenously expressed the insulin-like growth factor receptor 1b (IGFR-1b) by Western blotting (Figure 3 -figure supplement 1C). In support of our hypothesis, we measured increased aSyn-PM localization by TIRF microscopy upon insulin stimulation for 10 min (Figure 3 -figure supplement 1D). Given the short exposure times to histamine and insulin in these experiments, we reasoned that observed PM accumulations likely reflected enhanced recruitment of existing aSyn pools rather than de novo protein synthesis and PM targeting, thus providing further evidence that aSyn abundance at the PM correlated with signaling-dependent changes of PIP2 and PIP3 levels.

Discussion
Our results establish that clusters of endogenous aSyn are found at the plasma membrane of human A2780, HeLa, SH-SY5Y and SK-MEL-2 cells, where their native abundance correlates with PIP2 levels (Figure 1). Specifically, we show that targeted overexpression of the PIP2-generating kinase PIPKIγ increases endogenous aSyn at the PM ( Figure 1C), whereas the PIP2-specific phosphatase INPP5E reduces the amount of PM aSyn ( Figure 3A). We further demonstrate that PIP3-dependent histamine and insulin signaling redistributes aSyn to the PM (Figure 3B and Figure 3 -figure supplement 1), which collectively suggests that changes in PM PIP2 and PIP3 levels affect intracellular aSyn localization in a dynamic and reversible manner. Aiming for a stringent analysis, we investigated PM interactions at native aSyn expression levels and in a strictly unaltered sequence context, i.e., without modifying the protein with fluorescent dyes or fusion moieties.
These requirements precluded live-cell imaging experiments to determine PM-localization kinetics, although histamine and insulin experiments suggest that endogenous aSyn pools redistribute readily. Although we cannot rule out that additional secondary protein-protein interactions contribute to PM targeting, we demonstrate that aSyn directly interacts with reconstituted PIP2 vesicles in vitro (Figure 2A-D). Importantly, the biophysical characteristics of these interactions are indistinguishable from other previously identified, negatively charged membrane systems 29,31,43 . Based on the known membrane-binding preferences of aSyn, PIP2 and PIP3 lipids constitute intuitive ligands. Not only because of their highly acidic nature 21 , but also because of the compositions of their acyl chains, containing saturated stearic-(18:0) and polyunsaturated arachidonic-acids (20:4), the latter conferring 'shallow' membrane defects 44 ideally suited to accommodate aSyn's helical conformations 7,45 . Thus, from a biophysical point of view, phosphatidylinositol polyphosphates satisfy many of the known requirements for efficient membrane binding. From a biological point of view, PIPs are ubiquitously expressed 21 and stringently required for exocytosis and endocytosis, especially in neurons, where highly abundant PIP2 and PIP3 clusters (up to ~6 mol%) mark synaptic vesicle (SV) uptake and release sites 46 . Multiple PIP-binding proteins mediate key steps in SV transmission and recycling 47,48 and although aSyn has been implicated in synaptic exocytosis and endocytosis, its role(s) in these processes is ill defined 49 . Nonetheless, we believe that key conclusions of our study are generally valid. The affinity of aSyn to PIP2-vesicles has been reported to be in the low µM range 51 , similar to most other reconstituted membrane systems containing negatively-charged phospholipids 5,23,29,30,31 . In comparison, average dissociation constants for canonical PIP-binding scaffolds such as PH, C2, FYVE and ENTH domains vary between µM and mM 17,18 . By contrast, disordered polybasic PIP-binding motifs target negatively-charged membranes with much weaker affinities and in a non-discriminatory fashion based on complementary electrostatic interactions 20 . aSyn-PIP binding may define a third class of interactions that are comparable in strength to folded protein domains, but driven, to large parts, by electrostatic contacts similar to those of poly-basic motifs 9 . Based on these affinity considerations, we speculate that aSyn may successfully compete for cellular PIP2-PIP3 binding sites with other proteins, especially when their abundance is in a comparable range. For binding scenarios at presynaptic terminals, this is likely the case. 22 Our findings are additionally supported by recent data showing that intracellular aSyn concentrations directly influenced cellular PIP2 levels and that protein reduction diminished PIP2 abundance, whereas aSyn overexpression increased PIP2 synthesis and produced significantly elongated axons in primary cortical neurons 52 . Conspicuously, these effects depended on aSyn's ability to interact with membranes and were absent in a membrane-binding deficient mutant 52 Table 1. 1 µg of plasmids was used in all cases. Following transfection, cells were grown for 24 h before analysis.      Table 2.

Image Analysis and Quantification
Image analysis and quantification were performed in Fiji 66 . For confocal image quantification, focal planes of apical and basal PMs were selected manually. Images were segmented based on GFP signals by automatic thresholding according to Huang et al 67 .
Threshold regions were marked as regions-of-interest (ROIs), copied to the far-red channel (aSyn IF) and fluorescence intensities were determined. In the box plots of Figures 1E and 3A, B, each ROI corresponds to a single cell and is represented as a data point. For TIRF data in

Statistical Analysis
For box plots, data points considered 'outliers' were determined based on criteria defined in the Grubbs outlier test 69 and omitted. ANOVA tests with Bonferroni's post-tests 70,71 were used to determine the statistical significance of experiments with more than two samples, whereas Student's t tests were performed to assess statistical differences between samples 72 . Significance is given as * P < 0.05; * * P < 0.01; * * * P < 0.001.

Cell Lysate Preparation
Lysates  Table 2. Membranes were developed using the SuperSignal West Pico Plus reagent (Thermo) and luminescence signals were detected on a BioRad Molecular Imager.

Western Blot Quantification
Intensities of aSyn and b-Actin bands were quantified using the ImageLab software (BioRad). aSyn reference input was used to generate a standard curve. For cell lysate samples, aSyn intensity was normalized according to the b-Actin signal and cell lysate were calculated with respect to the aSyn standard curve. Error bars denote background signal (noise).

Recombinant Protein Expression and Purification
15 N isotope-labeled, N-terminally acetylated, human wild-type a-, b-and g-Syn were produced by co-expressing PT7-7 plasmids with yeast N-acetyltransferase complex B (NatB) 74 in Escherichia coli BL21 Star (DE3) cells using M9 minimal media supplemented with 0.5 g/L of 15 NH4Cl (Sigma). Protein purification under non-denaturing conditions was performed as described previously 75 . aSyn mutants DN and F4A-Y39A were generated by site-directed mutagenesis (QuikChange, Agilent) and confirmed by DNA sequencing.

Phospholipase C Reaction
Phospholipase C (PLC) was purchased from Sigma and the lyophilized powder was dissolved in NMR buffer at 1000 units (U)/mL. aSyn PC-PIP2 samples at 680-fold molar excess of PC-PIP2 lipids (60 µM aSyn, 40 mM PC-PIP2) were incubated while agitating at 37 °C for 45 min with 10 U of PLC and 1 mM PMSF in a total volume of 120 µL yielding a PLC activity of ~80 mM per min.

Nuclear Magnetic Resonance (NMR) Spectroscopy
For best comparison of protein reference and aSyn-lipid NMR data, final concentrations of 15 N isotope-labeled, N-terminals acetylated aSyn samples were adjusted to 60 µM, supplemented with 5% D2O and measured in 3 mm (diameter) Shigemi tubes in all cases. NMR experiments were acquired on a Bruker 600 MHz Avance spectrometer equipped with a cryogenically cooled proton-optimized 1 H{ 13 C/ 15 N} TCI probe. Reference and aSyn-lipid NMR spectra were acquired with identical spectrometer settings and general acquisition parameters. Specifically, we employed 2D 1 H-15 N SOFAST HMQC NMR pulse-sequences 77 with a data size of 128 x 512 complex points for a sweep width (SW) of 28.0 ppm ( 15 N) and 16.7 ppm ( 1 H), 128 scans, 60 ms recycling delay, recorded at 283 K. Inspection of the highly pH-sensitive His50 1 H-15 N chemical shift indicated that the sample pH changed from 7 to 6.5 during the PLC reaction ( Figure 2E). To accurately delineate I/I0 values, we recorded reference 33 NMR spectra at pH 6.5. All NMR spectra were processed with PROSA, zero-filled to four times the number of real points and processed without window function. Visualization and data analysis were carried out in CARA. NMR signal intensity ratios (I/I0) of isolated aSyn (I0) and in the presence lipids (I) were determined for each residue by extracting maximal signal peak heights in the respective 2D 1 H-15 N NMR spectra.

Circular Dichroism (CD) Spectroscopy
NMR samples of isolated aSyn and aSyn in presence of lipid vesicles were diluted with NMR buffer to a final protein concentration of 10 µM for CD measurements. CD spectra (200-250 nm) were collected on a Jasco J-720 CD spectropolarimeter in a 1 mm quartz cell at 25 °C. One replicate per sample was recorded. Six scans were averaged and blank samples (without aSyn) were subtracted from the protein spectra to calculate the mean residue weight ellipticity (qMRW).

Dynamic Light Scattering (DLS)
DLS measurements were acquired on a Zetasizer Nano ZS (Malvern Instruments) operating at a laser wavelength of 633 nm equipped with a Peltier temperature controller set to 25 °C. Data were collected on all NMR samples containing aSyn, isolated PC-PIP2 vesicles, and PC-PIP2 vesicles in presence of Ca 2+ and PLC, respectively. Using the Malvern DTS software, mean hydrodynamic diameters were calculated from three replicates of the same sample in the intensity-weighted mode.

Negative-Stain Electron Microscopy (EM)
NMR samples of aSyn at 30-and 680-fold molar excess of PIP2 and PC-PIP2 lipids were diluted to a protein concentration of ~10 µM in NMR buffer. 5 µL aliquots were added to 34 glow-discharged carbon-coated copper grids for 1 min. Excess liquid was removed with filter paper and grids were washed twice with H2O before staining with 2% (w/v) uranyl acetate for 15 s. Negative stain, transmission EM images were acquired on a Technai G2 TEM.