The number of α-synuclein proteins per vesicle gives insights into its physiological function

Although it is well established that the protein α-synuclein (αS) plays an important role in Parkinson’s disease, its physiological function remains largely unknown. It has been reported to bind membranes and to play a role in membrane remodeling processes. The mechanism by which αS remodels membranes is still debated; it may either affect its physical properties or act as a chaperone for other membrane associated proteins. To obtain insight into the role of αS in membrane remodeling we investigated the number of αS proteins associated with single small vesicles in a neuronal cell model. Using single-molecule microscopy and photo-bleaching approaches, we most frequently found 70 αS-GFPs per vesicle. Although this number is high enough to modulate physical membrane properties, it is also strikingly similar to the number of synaptobrevins, a putative interaction partner of αS, per vesicle. We therefore hypothesize a dual, synergistic role for αS in membrane remodeling.

α -Synuclein (α S) is an intrinsically disordered protein of 14.3 kD that is abundant in the brain where it represents a considerable part of the cytosolic protein content 1 . The protein α S has been suggested to play a key role in the development of Parkinson's disease (PD), a neurodegenerative disorder associated with loss of dopaminergic neurons in the substantia nigra 2 . Although the involvement of α S in PD pathology is well established, the exact physiological role of this protein remains an enigma.
α S is enriched in the presynaptic termini of dopaminergic neurons. At these termini the soluble protein is in equilibrium with vesicle bound α S 3 . Membrane bound α S has been suggested to interact with other synaptic vesicle proteins. It has been reported to chaperone the formation of the SNARE complexes [4][5][6] and thereby mediate membrane fusion processes. In vitro experiments hint that, at high protein to lipid ratios, α S can also directly change the physico-chemical membrane properties. From in vitro experiments, α S has been reported to increase lipid packing 7 and induce positive mean and negative Gaussian curvatures in phospholipid bilayers 8 . Additionally, α S binding has been observed to cause tubulation and fragmentation of phospholipid vesicles at high α S concentrations [9][10][11] , and reduce membrane tension and increase undulations in small unilamellar vesicles 12 . Although these changes in membrane properties agree well with a role for the protein in membrane remodeling processes, it is still unclear if the changes in membrane properties observed in vitro are relevant in vivo.
The two suggested functions of α S -being an interaction partner for other proteins or changing the physico-chemical properties of the membrane -require distinctly different numbers of α S per vesicle. Whereas for specific interactions with synaptic vesicle proteins only one or a few membrane bound α S molecules may be needed, larger surface concentrations are required if the physiological function of the α S involves changing physico-chemical properties of the membrane. The copy number of most synaptic vesicle proteins is low. Interestingly, α S is absent in reports on the copy number of major protein constituents of synaptic vesicles 13 . This absence might result from the equilibrium association of α S with membranes and the loss of bound protein during synaptic vesicle purification.
Here, we investigated the number of α S per vesicle in cells, thereby avoiding complications arising from vesicle purification. Using photo-bleaching experiments on α S-GFP expressing cells, we find a most frequent number of 70 α S-GFPs per vesicle. We discuss the observed surface concentration in relation to the possible physiological function of the protein.

Results
In images of primary neurons immunostained against α S, two distinct types of fluorescence signals are visible: a diffuse signal spread throughout the cell, and localized, small, high intensity α S puncta (Fig. 1A,B). In these neuronal cells, the α S puncta possibly represent synaptic or endocytic vesicles 3,14 . The radius of synaptic vesicles has been reported to range from 15 to 37 nm with an average of 21 nm 13 , whereas endocytic vesicles in the synapse of neuronal cells have a typical radius of 40 nm 15 . Clearly, these sizes are below the resolution limit of conventional optical microscopy, which is roughly 250 nm in radius. Hence, synaptic vesicles appear as diffraction limited sized features in the microscopy images.
Although immunostaining gives excellent results in visualizing and localizing α S in primary neurons, we not only aim to visualize α S, but also to determine the number of α S per vesicle. Unknown and varying labeling efficiencies plus possible steric hindrance due to the size of the antibodies preventing occupation of all possible binding sites, could easily lead to incomplete labeling of membrane bound α S which makes immunostaining unsuitable for quantitative studies.
Single molecule detection based techniques are by now well established in molecular life science [16][17][18][19][20] . To determine the number of molecules in complexes or molecular aggregates, single molecule photo-bleaching of fluorescent labels has been shown to be a viable technique [21][22][23][24] . In a cellular context, these photo bleaching based techniques require recombinant labeling of the protein of interest. These strategies, primarily based on GFP labeling of α S, have been proven successful before even in a mouse model of Parkinson's disease 3 . A control experiment verifies that the GFP tag does not change the interaction between α S and membranes. Circular dichroism experiments show that α S and α S-GFP bind phospholipid vesicles with comparable affinity (See Supplementary  Fig. 1), which is in agreement with literature 25 . Therefore, we used α S tagged C terminally with GFP (α S-GFP) and expressed it in a SH-SY5Y cell model. The SH-SY5Y cell model has been shown to be, after differentiation, a suitable cell model system for neurons in neurodegenerative diseases like PD 26,27 .
Indeed we find close resemblance between immunostained primary neurons ( Fig. 1A,B) and immunostained differentiated SH-SY5Y cells (Fig. 1C,D). In all cell types we find both a diffuse distribution of α S and a localization of α S in well-defined puncta. For the photo-bleaching experiments, we subsequently expressed α S-GFP in the SH-SY5Y cell model system (See Supplementary Fig. 2).
Since our study is based on the observation of bleaching steps from single GFP molecules, the use of ultrasensitive single-molecule fluorescence spectroscopy is required. In Fig. 1E,F we present typical images of our α S-GFP expressing cell line using ultrasensitive microscopy. We observed both a diffuse α S-GFP signal and a spotted pattern of α S puncta throughout the differentiated cells. The observation of both a diffuse signal and localized puncta agrees well with the observations from immunostained primary neurons and immunostained differentiated SH-SY5Y cells (Fig. 1A-D). We attribute the diffuse background signal to cytosolic α S-GFP or α S-GFP associated with vesicles that cannot be resolved individually. The small, high intensity puncta are within the size range expected for diffraction limited structures and might hence represent α S-GFP associated with synaptic and/or endocytic vesicles. Since the size of these puncta cannot be further resolved with conventional confocal fluorescence microscopy, we used STED super-resolution microcopy. From STED microscopy images (see Supplementary Fig. 3), we conclude that the structures giving rise to the puncta in our differentiated SH-SY5Y expressing α S-GFP model cell system are smaller than 80 nm, the resolution limit of the used STED setup, and are hence small enough to represent synaptic or endocytic vesicles.
To confirm that the puncta are indeed vesicles, we subsequently investigated the co-localization of the fluorescent α S-GFP puncta with membranes. To label membranes, we used the membrane marker wheat germ agglutinin, tagged with the fluorophore Alexa Fluor ® 647 (WGA-AF647). The GFP and AF647 fluorophores can be excited and detected independently and by overlaying the images, co-localization can be visualized. To show co-localization of α S-GFP puncta with the membrane, WGA was incubated with the cells for six hours. Independent excitation and detection of GFP (excitation 485 nm, detection 550/88 nm) and WGA (excitation 640 nm, detection >665 nm) resulted in co-localization of 40% of α S-GFP puncta with WGA-AF647 (see Methods).
In our co-localization experiments, upon excitation of GFP, we detected more intense emission above 665 nm than expected from direct excitation of AF647 by 485 nm laser light, and leakage of GFP emission into the detection window beyond 665 nm. Förster Resonance Energy Transfer (FRET) between the initially excited α S-GFP and WGA-AF647 could be the reason for this observation. When the two fluorescent markers, GFP and AF647, are in nanometer proximity, they form a FRET system in which the GFP (donor) fluorescence is efficiently quenched by the presence of AF647 (acceptor). The characteristic Förster distance for the GFP and AF647 pair is expected to be slightly smaller than the 5.6 nm reported for the spectroscopically similar AF488 and AF647 pair. This smaller distance results from the smaller quantum efficiency of GFP compared to AF488.
Moreover, in our experiments, a large fraction of puncta appeared only in the red channel (emission > 665 nm, emission of AF647) after GFP excitation at 485 nm, indicating very efficient FRET between α S-GFP and WGA-AF647 (Fig. 2, top panel). Furthermore, after photo-bleaching the FRET acceptor AF647 with high intensity laser light of 640 nm, the emission of the FRET donor, α S-GFP, appeared in the green channel (emission detection 550/88 nm) (Fig. 2, low panels). After this acceptor bleaching of AF647, and resulting dequenching of α S-GFP, co-localization between α S-GFP and WGA-AF647 increased from 40% to 60%. The formation of such an efficient FRET system implies nanometer proximity between α S-GFP and WGA-AF647. This nanometer distance could result either from a direct interaction between WGA and α S or GFP, or juxta positioning due to high local concentrations. FRET experiments on solutions of mixtures of WGA and α S or GFP excluded relevant direct interaction between WGA and α S or GFP (see Supplementary Fig. 4), suggesting that a high local density of α S-GFP near the membrane marker is most likely responsible for the observed FRET. Furthermore, nanometer proximity between α S-GFP and the membrane bound WGA-AF647 indicates that α S-GFP is in all likeliness membrane bound. This conclusion is further supported by the observation that in some larger vesicular Scientific RepoRts | 6:30658 | DOI: 10.1038/srep30658 structures, α S-GFP outlines the vesicle contour (see Supplementary Fig. 5). We hence conclude that the observed fluorescent puncta indeed represent vesicle bound α S-GFPs.
After establishing that the observed fluorescent puncta in the studied model cells originate from vesicle bound α S-GFP, the number of α S-GFPs per vesicle was determined. Recently, a number of methods have been developed that use single-molecule photo-bleaching to determine the number of fluorescently labeled subunits in a molecular complex or aggregate [21][22][23] . For larger complexes, Leake et al. pioneered an elegant method to determine the number of labeled subunits per complex that is suitable for in vivo quantification 24 . The number of fluorophores in a complex is determined by dividing the background subtracted initial fluorescence intensity by the average intensity per fluorophore, which is determined using a statistical analysis of each bleaching trace (see Methods). To confirm the applicability of this approach to our experimental system with small cellular vesicles, we tested the approach on unilamellar vesicles (UVs) with an average radius of 50 nm and a known average number of membrane embedded fluorophores. We then experimentally determined the number of fluorophores per UV using single-molecule photo-bleaching and the analysis method described above. We find that the determined and expected values of fluorophores per UV agree very well (see Supplementary Fig. 6). Hence, our control experiment verifies the suitability of the method to determine the number of fluorophores per vesicle.
To determine the number of α S-GFP per vesicle in our cell model, we recorded and analyzed bleaching traces obtained from more than 200 distinct puncta, a statistically relevant number. For our study we considered only α S-GFP puncta in the thin parts of the cell (Fig. 1F) with a radius smaller than 350 nm in fluorescence images (see Supplementary Fig. 7 for size distribution). The cutoff size reflects the punctum size expected from the convolution of the physical size of a vesicle with the microscopes point spread function. By using this cutoff criterion we exclude a bias in our data resulting from bleaching traces from large vesicular structures or the presence of multiple small vesicles in the sampled volume. After selecting an α S-GFP punctum that fitted these criteria, high intensity laser excitation (485 nm, 800 W/cm 2 ) was used to completely photo-bleach the α S-GFP emission, while recording the fluorescence intensity over time (see example in Supplementary Fig. 8A). The number of α S-GFPs per vesicle was determined from the bleaching traces of 243 distinct vesicles, the copy number was determined and assembled in the histogram presented in Fig. 3.
The distribution of the number of α S-GFPs on the studied vesicles is broad, ranging from around 20 α S-GFPs per vesicle up to more than 300 α S-GFPs, with a most frequent value of approximately 70 α S-GFPs per vesicle. Finding a distribution of the number of α S-GFP per vesicles rather than a defined number matches expectations, since it is well known that the size distribution of vesicles in cells is polydisperse 13,28,29 . With increasing vesicle surface area the number of α S-GFPs per vesicle is also expected to increase. Interestingly, the shape of the vesicle surface area distribution obtained from the size distribution of synaptic vesicles 13 and the distribution of the number of α S-GFPs per vesicle we determined, agrees very well. Copy numbers on the high end of the distribution might originate from more than one vesicle in the observation volume, in spite of our size selection criterion.
Using the reported sizes of synaptic vesicles and the number of α S-GFPs per vesicle obtained from the photo-bleaching experiments, it is possible to determine the average distance between α S-GFPs on a vesicle. For the most frequent value of 70 α S-GFPs on an average synaptic vesicle of 21 nm in radius, and neglecting multimerization or co-localization with other proteins, we find an average center-to-center distance of 10 nm between α S-GFPs on the vesicles.
This average distance agrees well with our observation of efficient FRET between GFP and AF647 for some of the vesicles (see Fig. 2). Considering the stochastic nature of both the binding of α S-GFP and the membrane marker WGA-AF647 to the vesicle, variations in energy transfer between α S-GFP and WGA-AF647 per vesicle are to be expected at this average α S density. For relatively few α S-GFPs per vesicle, the FRET partners are spaced further apart than the characteristic Förster distance. Hence for these low copy numbers co-localization of α S-GFP and WGA-AF647 is not accompanied by efficient energy transfer (i.e. 40% co-localization before photo-bleaching of the acceptor). For higher copy numbers, the distances between α S-GFP and WGA-AF647 are below the Förster distance, resulting in quenching of the FRET donor. The higher copy numbers therefore result in an underestimation of the co-localization of α S-GFP and WGA-AF647 (i.e. an increase from 40 to 60% co-localization after photo-bleaching of the acceptor).

Discussion
We observed diffraction limited fluorescent puncta in a neuronal cell model system of differentiated SH-SY5Y cells expressing α S-GFP. We used STED microscopy and co-localization of the α S-GFP puncta with the membrane marker WGA-AF647 to show that these puncta are vesicles. Efficient FRET between GFP and the fluorescent label AF647 attached to WGA strongly indicates that α S-GFP is membrane bound, which is further supported by our observation that α S-GFP outlines larger membrane structures. We hence conclude that the α S-GFP puncta we observe in our neuronal cell model, are small vesicles with membrane bound α S. We used single-molecule photo-bleaching to determine the number of α S-GFPs per vesicle in the cells. We find a distribution of copy numbers per vesicle with a most frequent number of 70 α S-GFP per vesicle.
We exclude the possibility that the puncta represent the well-defined oligomeric species and amyloid fibrils that have been discussed in the context of Parkinson's disease [30][31][32][33] . For example, the aggregation number of α S oligomers formed, in vitro, under different conditions was determined using single-molecule photo-bleaching experiments 22,23 . These in vitro formed oligomers typically consist of 30 ± 5 monomers 22,34 . Any fibrillar form of α S is expected to contain thousands of α S monomers. It can, however, not be excluded that vesicle bound α S is present as multimeric species.
The distribution in copy numbers reflects the distribution of vesicle sizes and the stochastic nature of binding of α S-GFP to vesicles. Considering that the SH-SY5Y cells were differentiated to neuron like cells, we assume that the studied vesicles are synaptic vesicles. The most frequent copy number of 70 α S-GFPs per synaptic vesicle, is rather high compared to the copy numbers found for most other synaptic vesicle proteins which are in the range of 1 to 10 per vesicle 13,35 . In this regard, since the number of α S per vesicle is high, the absence of α S in listings of the copy number of the major protein constituent of synaptic vesicles 13 is most likely caused by the equilibrium nature of the association of α S with membranes and detaching of α S from the vesicles during vesicle purification. Interestingly, 70 α S per vesicle is a high enough number to alter the physical membrane properties 8 .
In the literature, α S has been reported to change the physical properties of phospholipid bilayers [8][9][10] . Based on experimental data and simulations, α S has been suggested to be involved in membrane curvature induction and binding of α S has been shown to cause tubulation and vesiculation of lipid bilayers [8][9][10] . At high molar ratios of protein to lipid; 1:10 to 1:40, α S induces vesiculation and tubulation in POPG lipid bilayers 10 . At lower molar ratios of 1:200, α S induces positive mean curvature and local negative Gaussian curvature in flat lipid bilayers 8 . At this lower protein to lipid ratio, α S was reported to lower the membrane surface tension and increase membrane undulations in small unilamellar vesicles by decreasing the negative pressure in the headgroup region of the outer membrane leaflet and increasing the positive pressure throughout the hydrocarbon core 12 . If synaptic or endocytic vesicles in neurons are stressed by high curvature because of their size, the effect of α S binding on surface tension could stabilize the vesicles and prevent fusion.
Assuming that the observed puncta are synaptic vesicles with an average radius of 21 nm, the average copy number of 70 would result in a molar ratio of 1:250 between α S and lipids. Our data thus supports the idea that membrane binding of α S and the concomitant partial insertion of amphipathic α -helices can stabilize the membrane curvature and control fusion processes of vesicles with a radius smaller than 40 nm. Besides α -helix insertion, the large hydrodynamic radius of the charged unstructured C-terminal domain of the protein may contribute to the promotion of membrane curvature by generating steric pressure 36 . Structural studies of α S bound to sodium dodecyl sulfate micelles show that the protein forms two antiparallel α -helices comprising residues 3-37 and 45-92, while residues 93-140 remain unstructured and are exposed to the solution 37,38 . Assuming that the 48 residues at the C-terminus can adopt a free random walk with a persistence length l p corresponding to roughly 3 amino acids, they would form a globule with a radius of gyration of approximately = = . .
Using the known average size of synaptic vesicles, we determined the average distance between α S-GFPs on the vesicle surface to be 10 nm. We therefore conclude that α S binding to the already crowded surface of synaptic vesicles may contribute to steric pressure. However, since we used a photo-bleaching approach to determine the copy number of α S-GFP per vesicle, we do not sample endogenous α S. To estimate the ratio between α S and α S-GFP inside cells, we performed a Western blot. Analysis of Western blots of cell lysate indicated that the ratio between the endogenous α S and α S-GFP in the differentiated SH-SY5Y cells was approximately 1:2 (see Supplementary Fig. 2). Since α S and α S-GFP bind membranes with comparable affinity we conclude that the total number of α S proteins, including both α S and α S-GFP, associated with a single vesicle can be as much as 100. Insertion of the helical domain of α S and possibly, in conjunction with other membrane bound proteins, steric pressure exerted by the unstructured α S domain may thus considerably contribute to establishing and maintaining membrane curvature.
As mentioned earlier, α S has been reported to interact with other proteins. These reported interaction partners of α S include a variety of proteins, including, but not limited to the cytoskeletal protein tau 39 , the chaperone proteins HSP70 40 , and the SNARE protein synaptobrevin 6,41,42 . Most of the synaptic vesicle proteins are present in low copy numbers, the copy number of α S is high. The formation of complexes of α S and synaptic vesicle proteins in stoichiometric ratios of 1:10 or even lower seems unlikely. There is however an important exception to the low copy number found for synaptic vesicle proteins. Synaptobrevin, a v-SNARE protein involved in vesicle fusion 43 , was reported to have an average number of 70 on a single synaptic vesicle 13 . The agreement between the number of synaptobrevin and α S molecules on the vesicle surface determined by us, is striking and supports a possible role for α S in SNARE complex formation. It has been suggested that α S promotes SNARE-complex formation by a non-classical chaperone activity involving binding to synaptobrevin 6 . In this interaction the unstructured C-terminal domain of α S binds to the first 28 N-terminal residues of synaptobrevin, which are not involved in SNARE complex formation. This interaction is thought to catalyze the formation of SNARE complexes by presenting the vesicle bound synaptobrevin in the optimal conformation to interact with the plasma membrane bound SNAREs. We were not only able to confirm the interaction between α S and the soluble part of synaptobrevin, but using microscale thermophoresis we were able to determine the equilibrium dissociation constant K d of the α S -synaptobrevin complex (see Supplementary Fig. 9). The micromolar binding affinity and observed α S surface concentration are in agreement with a functional biomolecular interaction 41,42 .
Based on the total of approximately 100 α S proteins per vesicle, the observed direct interaction of α S with the soluble part of synaptobrevin, and the ability of α S to modulate the physical properties of membranes reported in the literature [8][9][10][11][12] we hypothesize a dual purpose of membrane bound α S. α S may both modulate the membrane's physical properties and interact with the SNARE protein synaptobrevin. Since interactions limit conformational freedom of the unstructured C-terminal domain of α S, the formation of a α S-synaptobrevin complex may not only result in a presentation of this v-SNARE to its t-SNARE partner but also reduce the steric pressure exerted by α S. We hypothesize that the interaction between α S and synaptobrevin is regulated by a third, yet unknown, factor. Switching on the interaction between α S and synaptobrevin with this factor, may not only result in the presentation of synaptobrevin, but also cause stabilization of the disordered part of α S resulting in a decrease of steric pressure and hence possible destabilization of the vesicle. The latter two events facilitate interaction between vesicle-SNAREs and target-SNAREs which finally results in the release of synaptic vesicle content.

Methods
Primary neuronal cells. The extraction and culturing of primary neurons was performed as described elsewhere 44 .
In short, cells were obtained from newborn (P1) Wistar rat pups. Both (cortical) cerebral hemispheres were isolated in a sterile environment, minced and trypsinized. The minced hemispheres were dissociated by trituration after which the cells were plated on polyethylenimine coated coverslips (Sigma-Aldrich, USA) to 60% density. After 2 hours, adhered cells were washed with DMEM (Invitrogen, USA) and cultured in 900 μ l serum and antibiotics-free R 12 medium at 37 °C with 5% CO2 45 . All research involving animals has been conducted according to Dutch law (as stated in "Wet op de dierproeven"), and approved by DEC, the Dutch Animal Use Committee.
Construct design for αS-GFP DNA plasmid. For construction of human α S-GFP plasmid, the complete human α S cDNA was obtained from pT7-7 human α S via a PCR reaction 46 . For this, a forward primer 5′ -GAAGATCTAAGATGGATGTATTCATGAAAGGAC-3′ and reverse primer 5′ -GAAGATCTCCGGCT TCAGGTTCGTAGTCTTG-3′ were used on a pT7-7 human-α S template. The reverse primer was designed in such a way that the natural human α S stop codon was removed. Following digestion with BglII, the human α S Scientific RepoRts | 6:30658 | DOI: 10.1038/srep30658 cDNA was isolated and ligated into the BglII and BamHI sites of pEGFP-N1 plasmid. The sequence of the construct was confirmed by DNA sequencing analysis.
Cell culture, transfection and selection of SH-SY5Y cells. The culturing, differentiation, and transfection of SH-SY5Y cells was performed as described elsewhere 47 . In short, SH-SY5Y cells were grown in proliferation medium GlutaMAX ™ supplemented with 10% heat inactivated FBS, 1% non-essential amino acids, 10 mM HEPES buffer, and 1% Penicillin/Streptomycin, all obtained from Gibco ® . For differentiation, we seeded SH-SY5Y cells to 60% confluency and induced differentiation by adding starvation medium containing 1% FBS and 10 μ M retinoic acid for 3 days. All chemicals were obtained from Invitrogen, USA if not indicated otherwise.
DNA (250 ng/cm2) was diluted in Opti-MEM in reduced Serum medium (GIBCO ® ) including Lipofectamine ® LTX Reagent with PLUS ™ Reagent. The mix was incubated for 5 minutes at room temperature before adding Lipofectamine LTX. For every 250 ng of DNA, 0.5 μ l of Lipofectamine LTX and 0.19 μ l of Lipofectamine PLUS reagent were mixed, according to manufacturer's protocol. After Lipofectamine LTX addition, a 30 minute incubation at room temperature was performed. The medium of the cells was changed and the DNA, Lipofectamine LTX, and Lipofectamine ® LTX Reagent with PLUS ™ Reagent mix were added. After one day, the medium was changed to proliferation medium. Two days after transfection, cells were trypsinized and re-seeded in conditioned medium. The next day, G418 (500 μ g/ml) was added and cells were grown in G418 supplemented conditioned medium until selection by fluorescence-activated cell sorting (FACS). The α S-GFP positive cells were expanded in culture dishes and stocks were stored in liquid nitrogen.
Immunocytochemistry and cell staining. Cell  To label membranes, we used the membrane marker wheat germ agglutinin (WGA) tagged with the fluorophore AlexaFluor ® 647 (AF647) (Invitrogen, USA). WGA is a lectin protein that binds to N-acetyl-D-glucosamine and sialic acid on cell membranes 48 . For WGA-AF647 labeling, living differentiated GFP-SH-SY5Y cells were incubated with 5.0 μ g/mL WGA-AF647 in medium for 10 minutes to 6 hours. Subsequently, cells were fixed and imaged as described above. Although previous reports 49,50 suggest that incubation with WGA-AF647 might induce apoptosis, we did not observe any detectable increased apoptosis associated with either the amount of WGA-AF647 or incubation time in our experiments.
Confocal fluorescence microscopy. Confocal laser scanning microscopy images were obtained using acommercial Nikon A1 confocal microscope with a 100 x oil immersion objective (NA = 1.4, Nikon, Japan) and laser sources of 405 nm, 488 nm, and 561 nm. Emission was detected using appropriate dichroic mirrors and filter sets.

Single-molecule imaging and photo-bleaching experiments.
The photo-bleaching experiments were performed using an ultra sensitive custom-built inverted confocal microscope, described in detail elsewhere 23 .
In short, as excitation source, we used a pulsed diode laser operating at 485 nm at a repetition rate of 20 MHz (LDH-D-C-485, Picoquant, Germany). An epi-illumination configuration was used, i.e., the illumination and emission are collected through the same microscope objective (UPLSAPO 60XW, 60X, 1.2NA, Olympus). The remaining excitation light in the detection path was suppressed with a long pass filter (RazorEdge ® , 488 nm, Semrock, USA) and a notch filter (StopLine ® , 488/14 nm, Semrock, USA). An additional band pass filter for the green channel (BrightLine ® , 550/88 nm, Semrock, USA) and a long pass filter for the red channel (RazorEdge ® , 664 nm, Semrock, USA) were used. The emission was spatially filtered using a 15 μ m pinhole and was subsequently focused onto a single photon avalanche diode (SPCM-APQR-16, PerkinElmer), connected to a photon counting module (PicoHarp300, Picoquant, Germany). The sample was first scanned at a high scanning speed, 1 ms per pixel, and low excitation power, ~80 W/cm 2 , to prevent dye bleaching. In the initial scan, we located individual α S-GFP puncta in PFA-fixed cells, localized them in the focus of the objective, and sequentially collected fluorescence intensity time traces from distinct α S-GFP puncta. To obtain the intensity time trace we used higher excitation powers, ~800 W/cm 2 , to make sure that all fluorophores of a punctum photo-bleached within 1 minute. After acquiring the photo-bleaching intensity time trace for the punctum (see Supplementary Fig. 8A), a previously reported data analysis has been used 24 . In short, a forward-backward non-linear filtering was used to de-noise the trace. Then, a power spectrum of the pairwise difference distribution function for each of the traces was obtained (see Supplementary Fig. 8B). The unitary step size -average effective emitted photon count from a single GFP until it photo-bleaches in each punctum -has been determined based on the power spectrum following the method by Leake and coworkers (Nature 2006 355-358), where the first significant peak against the background noise gives the step size. In the assignment of the step size, the presence of the second order peak was included in the selection criteria. Using the determined unitary step size, the total number of emitters per punctum was calculated.
Intensity profiles of the puncta were determined with Symphotime 64 (Picoquant, Germany) by applying a script that applies a Gaussian fit and calculates the Full Width Half Maximum (FWHM).
Calculating co-localization percentage between αS-GFP puncta and membrane marker. Two images were made sequentially using the 485 nm and 640 nm excitation of the single-molecule sensitive microscope. To quantify the co-localization between α S-GFP puncta and the membrane marker WGA, the ratio of puncta in the green channel (485 excitation) which co-localized with signal in the red channel (640 excitation), to all puncta in the green channel, was calculated. Then, each punctum that appeared in the red channel (485 excitation) image, but not in the green channel, was photo-bleached using the 640 nm laser light. After photo-bleaching, the number of puncta that additionally appeared in the green channel of 485 excitation were added to the numerator and denominator of the earlier ratio calculation, and the co-localization percentage was re-calculated.