Membrane Binding of Parkinson's Protein α‐Synuclein: Effect of Phosphorylation at Positions 87 and 129 by the S to D Mutation Approach

Abstract Human α‐synuclein, a protein relevant in the brain with so‐far unknown function, plays an important role in Parkinson's disease. The phosphorylation state of αS was related to the disease, prompting interest in this process. The presumed physiological function and the disease action of αS involves membrane interaction. Here, we study the effect of phosphorylation at positions 87 and 129, mimicked by the mutations S87A, S129A (nonphosphorylated) and S87D, S129D (phosphorylated) on membrane binding. Local binding is detected by spin‐label continuous‐wave electron paramagnetic resonance. For S87A/D, six positions (27, 56, 63, 69, 76, and 90) are probed; and for S129A/D, three (27, 56, and 69). Binding to large unilamellar vesicles of 100 nm diameter of 1‐palmitoyl‐2‐oleoyl‐sn‐glycero‐3‐phospho‐(1′‐rac‐glycerol) and 1‐palmitoyl‐2‐oleoyl‐sn‐glycero‐3‐phosphocholine in a 1 : 1 composition is not affected by the phosphorylation state of S129. For phosphorylation at S87, local unbinding of αS from the membrane is observed. We speculate that modulating the local membrane affinity by phosphorylation could tune the way αS interacts with different membranes; for example, tuning its membrane fusion activity.

ning residues 1-100. [20][21][22] TheN -terminalh alf (residues1 -50) of the amphipathic helix is termed helix 1, and the other half (residues 51-100), helix 2. Thea ffinity of aSt o membranes depends on the negative charge density (1)of the membrane,w here 1 represents the molar fraction of anionic lipids present in the membrane. [23] Differentb inding properties were foundf or helix 1a nd helix 2. [24] There are three ways to generate protein constructs to study the effect of phosphorylation:1 )t op hosphorylate the respective residues enzymatically,w hich requires dedicatede nzymes/overexpression systems [25,26] and is reversible;2)byasemisynthetic approach, in which a(phosphorylated) peptide is linked to the correspondingo verexpressed protein; [27] and 3) by generatingm utants whose side chains mimic the chemical properties of the phosphorylated state (negative charge) and size,sometimes referred to as pseudophosphorylation. [28] Typically,Si sr eplaced by Do rE [13,17,29,30] to mimic phosphorylation, and alanine is used as the reference for the nonphosphorylated state,e specially for in vivo studies.
All three approaches have been used to study aSp hosphorylation in vivo and in vitro,s howing that in some cases,e nzymatically phosphorylated aS( P-aS) and pseudophosphorylated aSc an behave differently. [29,31] Fore xample,e nzymatic phosphorylation of aSa tS 129 has been shown to have an inhibitory effect on aSa ggregation, while pseudophosphorylation does not show such an effect. [29] Apparently,t he different behaviord epends strongly on the properties probed and the environment aSi se xposed to.I nt he present study, we focus on the phosphomimic approach with the S!Ds ubstitution to mimic phosphorylation, and investigate the constructs S87A or S129A (nonphosporylated);a nd S87D or S129D (phosphorylated).
To investigate membrane binding, we used spin-label electron paramagnetic resonance (EPR) spectroscopy. Fors pin labelling,t he amino-acid residue at the sequence position of interesti sr eplaced by ac ysteine,w hich is reacted with as uitablef unctionalg roupo ft he nitroxide spin label (seeF igure1b), an approach introducedb yt he Hubbell group. [35a] In this way,anitroxide, which contains an unpairede lectron and is therefore EPR active, is covalently attached to the protein. Then the properties of the protein can bep robed at the modified position by EPR. In the present study,w em ake use of the ability of EPR to detect the mobility of the spin label by room-temperature continuous-wave (cw) EPR. Characteristicl ine shapeso ft he spectra revealt he mobility of the spin label, with narrow lines corresponding to fast motion (i.e., rotational correlation times (t r )o fs everal hundreds of ps) and broad lines to slow motion, in the ns-regime.I no ur particular case,s low motion of the spin label showst hat the section of the protein to which the spin label is attached is bound to the membrane,w hereas fast motion shows detachment of the protein from the membrane. Them ethodology described was introduced before and has provenv aluable for determining the local binding of aStom embranes. [24,[32][33][34] Thes pin-labelled constructs are referred to as SLposi-tionaS/S87A(D) or SLpositionaS/S129A(D), such that, for example,S L27aS/S87D is the constructw ith the spin label at position 27 and is the phosphorylated varianta t position 87. We investigateds everal spin-label positions for each phosphorylation site,r esulting in at otal of nine constructs,ass ummarized in Table 1.
In this work, we show how phosphorylation affects the binding of aSt ot he membrane.I td ecreases the binding of aSt ot he membrane when phosphorylated at the S87 position, whereas no effect is seen whenp hosphorylated at the S129 position. We also show that phosphorylation at position 87 does not detach the protein completely from the membrane,b ut rather, causes local unbinding, which is particularly pronounced in the helix2region.

Results
We investigate the binding of phosphorylationv ariants of aSa tp ositions 87 and 129 to LUVs of 100 nm diameter. TheL UVs are composedo fa1:1m ixture of POPG and POPC,g eneratingamembrane of charge density 1 = 0.5. We first describe the results of phosphorylation at position 87, then at 129. Figure 2s howst he spectra of the spin-labelled constructs probing phosphorylation at position 87 in the presence of LUVs (fort he complete list of constructs,s ee Table 1). In this set, helix 1i sp robed in the middle,a t residue 27;h elix 2i sp robed at five probing positions, starting from position 56 and terminating at 90. Figure 2a shows the spectra of aSi nt he nonphosphorylated form and Figure 2b in the phosphorylatedf orm. Thes pectrai n Figure 2a differ from those in Figure 2b;m ost notably, eachs pectrumi nF igure 2b has narrower linest han its counterpart in Figure2a. As described in the introduction,n arrow linesderive from spin labelst hat are rotating fast. As discussed in mored etail below,f ast rotation shows that the section of the protein to which the spin label is attachedi sn ot bound to the membrane.M ore detailed information was obtainedb ys pectrals imulation of the experimentals pectra,w hich yields the parameters of mobility of the spinl abel, the rotationalc orrelation time(t r ), and in the case of multicomponent spectra, the amount by which each fraction contributes.T hesep arameters are given in Table 2. In Figure 2c, an example of as imulation is shown. Three fractions are visible:t he fast, the slow,a nd the immobile components,w hich have increasingly largel inewidths. Thei ndividual components add up to give the experimental spectrum. Table 2r eveals that all but two spectra consist of as uperposition of two components,t he fast and slow components,e xceptf or the SL56aS/S87A variant, which, in addition, has at hird, the immobile component, and the SL90aS/S87A and SL90aS/ S87D variants,w hich have only one component, the fast component. Each component reflects ap art of the protein population:t he fast fraction is due to protein in which the region around the site that is spinl abelled is not attached to the membrane,w hereas the slow and immobilized fractionsa re due to the sections bound to the membrane.T he amount by whiche ach component contributest ot he spectra ( Table 2, columns four and six) reflects the fraction of protein contributingt oe ach component. Thec orrelation times can be determined to several tens of ps in the case of the fast fraction, and several hundred ps for the slow fraction (see Table 2). Thec ontribution of the fast component of aSi nt he nonphosphorylated form is smaller than in the phosphorylated form for each probing position. Theo pposite is the case for the contribution of the slow components.B oth these trends reveal that phosphorylation reduces membrane binding. To illustrate the effect of phosphorylation atp osition 87, Figure 3s hows ap lot of the amount of the fast fraction for phosphorylation at position8 7a safunction of the sequence number at which mobility is probed.
Fora ll monitoring positions,t he amounto ft he mobile fraction is larger in the phosphorylated variant. At monitoring positions 27 and 56, the amount of mobilefractions of nonphosphorylated aSi sb elow 10 %, which indicates strong binding, but at later positions( helix 2), the amounto ff ast fractions increases to 70 %, indicating the loosening of the helix 2o faSw hen it is nonphosphorylated, in agreement with previous findings for wt aS. [24] For the phosphorylated aS, thea mount of the mobile fraction is higher than in the nonphosphorylated form for all positions monitored, enhancing the tendency for local unbinding in helix 2, until, at position 90, the bound fraction is so low that it becomes undetectable.
To determine if the phosphorylation reduces the overall membrane affinity of aS, i.e., if aSd etaches completely from the membrane, resulting in aSp rotein that is free in solution( physical unbinding), we separated the physically Table 1. The aSc onstructs used to study phosphorylation at position S87 and S129;SLdenotes the position of the spin label.
Spin-label positions S129A (nonphosphorylated) S129D (phosphorylated) SL27 SL27aS/S129A SL27aS/S129D SL56 SL56aS/S129A SL56aS/S129D SL69 SL69aS/S129A SL69aS/S129D unbound fraction of aSf rom the membrane-bound fraction, by filtrating the samplet hrough af ilter that retains the vesiclesa nd aSb ound to them.T he amount of physically unboundp rotein in the filtrate is then determined by EPR, as described in Drescher et al. [24] (for details,s ee Section4 ). Thea mount of unbound aSi sg iven in Table 3, and is below 16 %f or all constructs.T hus,t he amounto fp hysically unbound aSi ss ignificantly lower than the amount of the fast fraction measured by EPR (see Table 2), showing that the local unbinding far outweighs any physical unbinding. Thep ercentages in Table 3f or spin-label positions2 7a nd 56 are slightly lower than for the other positions. Given that the differences are justo utside the error margins of the procedure, we cannot draw conclusions. Forp hosphorylation at position 129, Figure 4s howst he superpositiono ft he spectra of nonphosphorylated and phosphorylated variants for three spin-label positions (see Table 1). In contrast to phosphorylation at position 87, A and Dv ariants at position1 29 have similars pectra,o bviating the need for detailed spectral analysis.A pparently, phosphorylation has am uch smaller influence at position 129 than at position8 7.

Discussion
We have investigated how the membrane binding of aS depends on the phosphorylation state of positions8 7a nd 129. Membrane binding is detected locally,v ia the mobility of spin labels attachedt os pecific positions in the protein. An increased spin-labelm obility shows that the protein detaches from the membrane around the position probed.
Them embrane composition was chosen to be conducive to intermediate binding, with ac harge density of 1 = 0.5, to avoid dominant electrostatic effects,w hich are observed at higher charge densities,w here they cause strong, undifferentiated binding and are nonphysiological, or low charge densities, causing overallu nbinding, [23,24,[32][33][34] as described in the Introduction. Them embrane was offered in the form of LUVs of ad iameter of 100 nm. We mimic phosphorylation by the phosphorylation-mutation approach, replacing Sb yD ,a na pproach used before [13,17,29,30] (for details,s ee Introduction). Althoughs ome studies showed that biochemically phosphorylated aSc an have different properties than phosphorylation mimics, [29,31] the latterc onstructs provide ar obust system to study phosphorylation effects in vitro, explaining their popularity.
Under the conditionso fo ur study,p hosphorylation at position1 29 has no noticeable effect on membrane binding, whereas 87 has,s imilar to whatw as observed by other techniques in the past. [13] In the following, we will first discuss the influence of phosphorylation at position 87 on aSm embraneb inding, and then comparet he re- sults obtainedo nb oth phosphorylation sites to previous findings in the literature.
When position8 7i sp hosphorylated, membrane binding is reduced relative to the nonphosphorylated case.A n almost constantr eduction of the binding is observed at positions 27 and 56 in the helix 1r egion:s ee Figure 3. Similar to wild-type aS, [24] also in the S87A variants, helix 2h as al ower membrane affinity than helix 1. Phosphorylation enhances this trend,u pt ot he pointt hat at probingp osition 90, the bound fraction becomes so low that is undetectablew ithin experimentale rror. Complete physical detachment of the phosphorylated protein from the membrane does not play ar ole:a ss een in Table 3, the physicallyu nbound fraction is below 16 %f or all constructs.T oplace this into perspective,the amount of physically unbound aSi sm aximally one-third of the amount of fast fraction determined fromE PR, showing that the majority of the fraction, seen by EPR, derives from protein that is attachedt ot he membrane,p resumably at the residues precedingt he probed sequence position,e .g.,f or sample SL27/aS87P,r esidues2 7a nd below. Fluctuations in the amount of fast fraction (Table 2, SL 63, nonphosphorylated (SL63/S87A)h as al arger amount of fast fraction than SL 69), and al arger amount of physically unbound aSf or SL positions in helix 2 ( Table 3), could indicate an influence of the spin label on aSm embraneb inding. If such an effect is present, it never exceeds ac ontribution of 10 %, and therefore is not relevant for the conclusions drawn.
Overall, we find that phosphorylation at position 87 decreases the membrane affinity of aS, particularly for helix 2. This effect is fully consistent with the change in the charge caused by the conversion of S!Do rb yp hosphorylation:Anegative chargei nt he helix 2w ill weaken the electrostatici nteraction with the negativelyc harged    membrane surface,a si tc ounteracts the effect of several lysines (Lys;K )i nt he aSs equence from residues 1-100. Reducedm embrane binding of S87E and P-S87 has been reported before,e .g.,R efs. [13] and [35b]. Reducedm embrane bindinga ffects the entire protein, but is most pronounced in the helix 2r egion,a nd may selectively influence the behavior of helix 2. Some models propose that the physiological functiono faSi nvolves vesicle fusion events in which helix 1a nd helix 2i nteract with differenttypes of membranes. [36] We therefore specu-late that phosphorylation at position8 7c ould be used to tune how aSoperates in vesicle trafficking.
Fort he aS129 A/D variants,t he differencei nm obility of the spin label for phosphorylated and nonphosphorylated forms is minute,s howing that under the membrane conditions employed here,p hosphorylationa tt his site does not affect membrane binding. TheC -terminus of aS is already negatively charged and was not found to interact with the membrane in previouss tudies, [20,21,24,37,38] which is fully consistentw ith the lack of changes in membrane binding observed in the presents tudyu pon phosphorylation at position1 29.
Ther esults of the presents tudy suggest that phosphorylation at position 87 tunes thosef unctions of aSt hat involve membrane binding and vesicle interaction,w hereas phosphorylation at position 129 acts on other aspects of aSi nt he organism. Previously, [13] severalp ossibilitieso f how phosphorylation at 129 could affect aS in vivo behavior have been described and the study of Kosten et al. [39] shows that the phosphorylation at position 129 depends on the phosphorylation state of position 125, suggesting ac omplex interplay of posttranslational modifications in the C-terminus.
Most of the current research is focusedo np hosphorylation at position 129, and the phosphorylation degree at this position is relatedt od isease effects,a sr eviewed in Ref. [ 40].I na greement with our results, several studies show that aSp hosphorylation at 129 has no or little effect on membrane binding;s ee,f or example,R ef. [ 28]; however, severalstudies find an influenceofphosphorylation at 129 on the aggregation of aS [28,29,41] and on membrane bindingo faSa ggregates, [41] suggesting that in vivo effects are linked to aggregation-sensitive processes.

Conclusion
In conclusion,t he large spectrum of phosphorylation effects on aS in vivo and in vitro [13,14,16,19,28-31,35b,40-50] furnishes the need for isolating the different factors that can be modulated by aSp hosphorylation in vitro.T he present study gives one such example,w here we show that in vitro phosphorylation mimics at position8 7( S87D) reduce aSm embrane bindingi nal ocal, sequence-dependent manner, whereas the same modification at position 129 (S129D)h as no influence on membrane binding. We expect that this approachp rovides af oothold to interpreting the challenging in vivo physiological and pathological functions of aS.

Protein Expression and Labelling
All aSm utants were expressed in Escherichia coli strain BL21(DE3) using the pT7-7 expression plasmid and puri- fied in the presence of 1mMD TT,a sp reviously reported [51,52] Serine-87 is substituted either by Alanine (S87A, represents phosphorylation-inactive form) or by Aspartate (S87D, represents phosphomimic form). Forl abelling, ac ysteine mutation was introduced at the desired residues.
Spin labellingw as done following the standard protocol, described briefly.B efores tarting labelling, aSc ysteine mutants were reduced with as ix-fold molare xcess per cysteine with DTT( 1,4-dithio-D-threitol) for 30 min at room temperature.T or emove DTT,s amplesw ere passed through aP ierce Zeba 5mld esalting column.I mmediately,aten-fold molar excess of the MTSLs pin label ((1-oxyl-2,2,5,5-tetramethylpyrroline-3-methyl)-methanethiosulfonate) was added (from a2 5mMs tocki n DMSO) and incubated for 1h in the dark at room temperature.A fter this,t he free spin label was removedb y using two additional desalting steps.P rotein samples were applied onto MicroconY M-100 spin columns to remove any precipitated and/or oligomerised proteins and were diluted in buffer (10 mM Tr is-HCl, pH 7.4). Spin-label concentrations were 2.5 mM at protein concentrations of 250 mM. Owing to the high reactivity of the label and the fact that the cysteiner esiduesw ere freely accessible in the poorly folded structure,n ear quantitative labelling could be achieved under thesec onditions. [37] Samples were stored at À80 8C.

Preparation of Vesicles
All lipids were purchased from Avanti Polar Lipids,I nc. as chloroform solutions and were used without further purification.L UVs were prepared from 1:1m ixtures of 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1'-rac-glycerol) (POPG) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC). Lipids were mixed in the desired ratio and then chloroform was evaporatedb yd ry nitrogen gas.T he resulting lipid films were kept under vacuum overnight. Dried lipid films were hydrated with 10 mM Tr is-HCl, pH 7.4 for 1hour at 30 8C, and the resulting milky lipid suspensionsw ere extruded through 100 nm pore size polycarbonatedm embranes using the mini extruder (catalogue no.6 10000) from Avanti Polar Lipids. Thes ize of the vesicles was determined by dynamic light scattering (DLS). TheD LS-experiments were performed on aZ etasizer Nano-ZS( Malvern). We obtainedv esicles with ah omogeneous size distribution around d = 100 nm.

Sample Preparation
Spin-labelled aSm utants were added from stock solutions (concentration between 150 mMa nd 250 mM) to the LUVs to obtain al ipid to protein ratio (L :P)o f2 50 :1, and incubated for 30 min at room temperature before measuring. All samples were prepared and measured at least three times.A ll spin-labelled aSc onstructsu sed in this work are shown in the Table 1.

Filtration Experiments
To determine,i faSp hysically detaches from the membrane,w ep erformed filtration experiments similar to those described in Drescher et al. [24] An aSv esicle solution, prepared as for the EPR experiments described above (sample preparation), was passed through a1 00 kDa cutoff filter device (Amicon Ultra 100k) which retained the vesicles,a nd thereby, the membrane-bound aSf raction, but was permeable for unbound aS. Thec oncentration of aSi nt he filtrate was too low to measure directly;t herefore,t he filtrate was concentrated using a3kDa cutoff filter device (Amicon Ultra 3k) and measured by EPR to determine the amounto faSi nt he filtrate.T he error in the final value,o ft he order of 20 %, is largely due to the errors in determining the volumes before and aftert he concentrations tep,a nd the erroro f the double-integralp rocedure to determine the spin concentration by EPR.

Simulation of cw-EPR Spectra
Spectral simulations were performed usingM atlab (7.11.0.584, Natick, Massachusetts,U .S.A.) and the Easy-Spin package. [53] Fora ll simulations,t he followings pectral parameters were used: g = [2.00906,2 .00687, 2.00300] [54] and the hyperfine tensor parameters A XX = A YY = 13 MHz. Usually,asuperposition of more than one component was required to simulate the spectra. Thep arameters were manually changed to check in which range acceptable simulations of the experimental spectra were obtained to determine the error margins.T os imulate spectra of aS bound to membranes,t he t r of the fastest component was kept at the t r value of the spectra of the respective protein construct, in the absence of vesicles.
entitled AS ingle Molecule View on Protein Aggregation, part of the research program of the Foundation for Fundamental Research on Matter (FOM), which is parto f NWO. We thank MykhailoA zarkh( Leiden Institute of Physics) for help with analyzingt he data, WimJ esse (Leiden Institute of Chemistry) for introduction to the DLS apparatus,a nd Edgar Groenen( LeidenI nstitute of Physics) for constanti nterest and fruitfuldiscussions.