Toward a molecular mechanism for the interaction of ATP with alpha-synuclein

The ability of Adenosine Triphosphate (ATP) to modulate protein solubility establishes a critical link between ATP homeostasis and proteinopathies, such as Parkinson's (PD). The most significant risk factor for PD is aging, and ATP levels decline dramatically with age. However, the mechanism by which ATP interacts with alpha-synuclein (αS), whose aggregation is characteristic of PD, is currently not fully understood, as is ATP's effect on αS aggregation. Here, we use nuclear magnetic resonance spectroscopy as well as fluorescence, dynamic light scattering and microscopy to show that ATP affects multiple species in the αS self-association cascade. The triphosphate moiety of ATP disrupts long-range electrostatic intramolecular contacts in αS monomers to enhance initial aggregation, while also inhibiting the formation of late-stage β-sheet fibrils by disrupting monomer–fibril interactions. These effects are modulated by magnesium ions and early onset PD-related αS mutations, suggesting that loss of the ATP hydrotropic function on αS fibrillization may play a role in PD etiology.


Introduction
4][5][6][7][8] One protein for which this modulatory effect of ATP is particularly relevantyet currently unclearis alpha-synuclein (aS), whose aggregation is closely linked to the pathology of Parkinson's Disease (PD). 9ge is the primary risk factor for idiopathic PD, with incidence of PD increasing from ∼1% for individuals over 60 to 5% for those over 85. 10 Interestingly, aging is also accompanied by a dramatic decline of ATP levels in multiple model organisms. 11Older Caenorhabditis elegans and Drosophila exhibit ATP levels approximately 80% and 50% lower than their younger counterparts, respectively. 11In agreement with these observations, ATP levels decrease with age in the mouse brain, as well as in the cardiac muscles of both mice and humans. 11Given the age-related parallels between increasing PD prevalence and decreasing ATP levels, it is possible that a modulating effect of ATP on aS aggregation could be a factor in PD onset. 12ndeed, in vivo studies show that adenine supplementation to increase ATP levels from 1 to 4 mM signicantly reduces the number of aS-GFP-positive foci/aggregates in yeast. 13Additionally, yeast strains with signicantly reduced ATP levels show increased sensitivity to aS-GFP, as measured by reduced growth, suggesting that higher ATP levels likely reduce aS cytotoxicity. 13herefore, the available data collectively suggest that ATP likely elicits a physiologically-and pathologically-relevant effect on aS aggregation. 14,15However, the mechanism by which ATP exerts this effect is currently unclear.
Alpha-synuclein is a 140-amino-acid protein divided into three regions with distinct net charges (Fig. 1a): a positivelycharged N-terminus (residues 1-60) rich in lysine residues, a predominantly hydrophobic, non-amyloid-b component or NAC region (residues 61-95) that drives aggregation and an acidic C-terminus (residues 96-140) that binds a variety of metal ions. 16,17Throughout the aS sequence the consensus motif "KTKEGV" is repeated nine times either completely or partially (Fig. 1a) and these pseudo-apolipoprotein-like repeats confer lipid-binding properties to aS. 17,18 However, under pathological conditions aS monomers form insoluble, b-sheet-rich amyloid brils as well as neurotoxic oligomeric intermediates. 18any factors inuence aS aggregation, including aS point mutations such as E46K and A53T, which are associated with familial PD and cause disease onset decades earlier than under idiopathic conditions. 19In addition, tetra-polyphosphates interact electrostatically with the "KTK" segments of the Nterminal "KTKEGV" repeats of aS monomers to enhance charge-driven aggregation by disrupting long-range electrostatic contacts between the N-and C-termini of aS monomers. 20hese contacts shield the aggregation-prone NAC region of aS monomers and thereby inhibit aggregation (Fig. 1a; closed monomer). 21][23][24] It is currently unclear to what extent the mechanisms of polyphosphate-aS interactions can be transferred to ATP-aS complexes.A model proposed to explain the hydrotrope-like effect of ATP on aggregation-prone proteins suggests that the aromatic purine ring of ATP clusters over protein hydrophobic patches while the triphosphate chain interacts with bulk water to prevent the formation of aggregates by repelling other such bound monomers (Fig. 1b). 25However, ATP can also elicit a pro-aggregation effect on positively-charged amyloidogenic proteins, e.g., tau and human muscle acylphosphatase, by electrostatically binding to lysine residues of multiple monomers to enhance nucleating dimer formation (Fig. 1c). 6,7Similar interactions have been reported for the complexes formed by ATP and proteins undergoing phase separation.Phase separation of Fused-In-Sarcoma (FUS) protein is modulated by the adenine ring of ATP forming p-p interactions with aromatic side chains, while the triphosphate interacts electrostatically with arginine and lysine residues. 8The ATP : protein ratio is also critical for determining the effect of ATP, as ratios less than 100 : 1 promote phase separation of TDP-43 and FUS proteins via bivalent binding, while ratios above 100 : 1 elicit the opposite effect. 5,8inally, Nishizawa et al. proposed that ATP interacts indirectly with aS via weak, non-specic interactions that are driven by Mg 2+ serving as a bridge between ATP and the aS Cterminus (Fig. 1d). 12ile these studies provide important clues on aS-ATP complexes, several questions remain open on how ATP interacts with aS and how these interactions are modulated by pathologically-relevant factors, such as PD-related aS mutations.The effect of ATP on aS monomer conformations and on pathologically-relevant aS aggregation is also still unexplored.Addressing these gaps is critical to understanding the physiological and pathological roles of ATP.
Here, we probe how ATP interacts with multiple aS species and how these interactions modulate aS intramolecular and monomer-bril contacts and aggregation.By complementing multiple Nuclear Magnetic Resonance (NMR) experiments with Thioavin T (ThT) uorescence, dynamic light scattering (DLS) and microscopy, we explore the effect of ATP on both early-and late-stage aS aggregation.We show that ATP causes an enhancement of early aS aggregation by disrupting longrange electrostatic contacts in aS monomers and thereby shortening the lag time for b-sheet bril formation.Strikingly, elevated ATP levels also signicantly inhibit late-stage aS aggregation as well as N-terminally-driven aS monomer-bril contacts that are critical for templated bril elongation and pathologically-relevant secondary nucleation. 9,26We also show that the triphosphate moiety of ATP drives its primarily electrostatic interaction with the lysine-and threonine-dense "KTKEGV" N-terminal repeats in aS monomers. 17,18These ATP-aS interactions are modulated by Mg 2+ , which sequesters ATP from aS as ATP-Mg and vice versa, as well as by aS mutations that perturb the lysine-and threoninedistributions (E46K and A53T).Our data reveal that the E46K and A53T aS mutations dramatically alter the effect of ATP on aS aggregation, causing a signicant increase in the population of soluble aggregates of intermediate-size.The N-terminus, NAC region and C-terminus are colored in blue, yellow and red, respectively. 162][23][24] (b) One hypothetical model of the ATP-aS interaction based upon the interaction of ATP with proteins in the crystalline lens, whereby the adenine group of ATP clusters over protein hydrophobic patches and the triphosphate electrostatically repels other bound monomers to inhibit aggregation. 25(c) Another viable model for ATP-aS interactions, suggesting that ATP could bridge aS monomers and enhance aggregation via the phosphate-mediated targeting of lysine residues (designated by "+" symbols) in the aS N-terminus, as is the case for tau protein. 6(d) Model for Mg 2+ -mediated ATP-aS interaction proposed by Nishizawa et al., whereby Mg 2+ ions interact with the aS C-terminus and "bridge" indirect, non-specific interactions between the protein and ATP. 12 Results and discussion ATP elicits a concentration-dependent effect on aS aggregation that is phosphate-dependent and is preserved in the presence of Mg 2+ 7][18] We found that ATP causes a concentration-dependent shortening of the lag time for aS aggregation (Fig. 2a).This effect is already evident at ATP concentrations as low as 0.25 mM, and is particularly signicant at 10 mM.Interestingly, the aggregation-accelerating effect of ATP on aS is similar yet distinct from the effect of polyphosphates, as ATP causes a signicant shortening of the aS lag time at much lower concentrations than do tri-polyphosphates. 20In addition, 10 mM ATP causes a statistically-signicant reduction of aS ThT uorescence at plateau that is different from the enhancements caused by 0.25 and 1 mM ATP (Fig. 2a inset).The decreased plateau ThT uorescence observed in the presence of 10 mM ATP reects a loss in the amount of late-stage aS aggregates/brils, as conrmed by Transmission Electron Microscopy (TEM) (ESI Fig. S1 †).Furthermore, comparative DLS analyses show that 10 mM ATP also reduces the formation of soluble, intermediate-size aS aggregates at plateau (ESI Fig. S2 †).The ability of 10 mM ATP to both enhance early aS aggregation by shortening the lag time and concurrently inhibit aS ThT uorescence at plateau reveals a unique bi-phasic effect of ATP on protein aggregation.Overall, our ATP-and time-dependent ThT data provide an initial explanation for how ATP inuences aS aggregation.
Since the formation of mature aS brils is a hallmark of PD pathology, we next sought to determine the driving force for the ATP-dependent modulation of aS ThT uorescence at plateau. 9 To this end, we compared the effect of 10 mM ATP on aS to that of ATP analogs which differ from ATP in the number of phosphate groups, i.e., Adenosine Monophosphate (AMP) and Adenosine Diphosphate (ADP).Interestingly, ADPbut not AMPis able to signicantly reduce aS ThT uorescence at plateau to a similar extent as ATP, suggesting that ATP's effect on aS at plateau requires two or more phosphate groups (Fig. 2b).Additionally, the triphosphate-mediated reduction of aS ThT uorescence at plateau is statistically different from that of ATP, suggesting that the observed effect of ATP on aS aggregation is not completely recapitulated by another small molecule with three negative phosphate groups (Fig. 2b).In addition, the effects of ammonium sulfate and guanidine hydrochloride, which were selected as they are at opposite ends of the Hofmeister series, are signicantly different than that of ATP (Fig. 2b).From this we conclude that the effect of ATP on pathologically-relevant late-stage aS bril formation is not likely a general salting effect.Given the importance of the triphosphate moiety, the effect of ATP is instead likely to specically require nucleotides and not inorganic salts (Fig. 2b). 9In this respect, the interaction between ATP and aS shares some common features with that between ATP and lysozyme, which was recently modelled by Ou et al. using MD simulations. 27Ou et al. found that although the phosphate groups of ATP tend to form salt bridges primarily with positively-charged lysozyme residues, the adenine and ribose moieties of bound ATP molecules also form hydrogen bonds with the protein. 27e next explored whether ATP is still able to shorten the lag time for aS aggregation and reduce its plateau ThT uorescence in the presence of equimolar Mg 2+ , since Mg 2+ complexation is required for ATP's biological function as an energy source. 28,29n addition, Mg 2+ levels decline with age and are oen greatly reduced in PD patients. 7,30We observed that ATP can still cause a concentration-dependent decrease in the aS lag time, even in the presence of high Mg 2+ (Fig. 2c).ATP concentrations above 5 mM also cause a signicant reduction in aS ThT uorescence at plateau (Fig. 2d).Thus, the bi-phasic effect of ATP on aS aggregation persists in the presence of high Mg 2+ concentrations, suggesting that Mg 2+ is unable to silence the effect of ATP on aS aggregation.
As aS cytotoxicity correlates inversely with aggregate size, it is likely that the effects of ATP on aS aggregation in both the presence and absence of Mg 2+ , as well as the altered levels of ATP and Mg 2+ with increasing age, play a pleiotropic role in PD pathology. 9,11,13,31To understand the mechanisms underlying the effect of ATP on aS aggregation, we turned to NMR to characterize the interactions between ATP and aS and to evaluate how these interactions are modulated by Mg 2+ .

The triphosphate group of ATP drives electrostatic interactions with N-terminal lysine and threonine residues in aS monomers
We next characterized the driving mechanism for the ATP-aS interaction by measuring the ATP-induced 1 H- 15 N HSQC-based chemical shis of WT aS monomers.Fig. 3a-c show that ATP induces concentration-dependent shis primarily in lysine and threonine residues within the N-terminal pseudoapolipoprotein-like repeats of aS, leaving the NAC and Cterminal imperfect "KTKEGV" repeats largely unaffected and thereby suggesting a level of specicity for the ATP-aS interaction. 16,17Such specicity is notable as the binding of ATP to aS is weak with K ds in the 4-10 mM range, as shown by the binding isotherms built using the most signicant, N-terminal aS ATPinduced chemical shis (Fig. 3e).Interestingly, these mM K d values are similar to the reported range of cellular ATP concentrations, indicating that a signicant proportion of cellular aS is likely bound to ATP. 4,5 Another notable feature of the ATP-induced chemical shis in Fig. 3a and b is their parallel pattern which is quantitatively conrmed by the cos q ij matrix in Fig. 3d and suggests that ATP approaches different aS residues in a consistent orientation. 32owever, based on the chemical shi maps of Fig. 3a-c, it is clear that the high local concentrations of polar and positivelycharged residues in the aS N-terminus is critical for the ATP-aS interaction and predominates over p-p stacking interactions since aromatic aS residues in the C-terminus and hydrophobic NAC region show less signicant, concentration-dependent ATP-induced chemical shis (Fig. 3b and c). 17These data rule out the hypothesis that the ATP-aS interactions are driven by the adenine ring of ATP clustering over protein hydrophobic patches (Fig. 1b), while ruling in other models in which ATP-aS binding is mediated primarily by the triphosphate (Fig. 1c). 6,7,25o conrm the hypothesis that the triphosphate moiety of ATP drives its electrostatic interaction with aS, we compared the aS chemical shis induced by ATP to those induced by AMP, ADP and triphosphate (Fig. 3f-h).These ATP analogs cause phosphate-dependent shis mainly in the same N-terminal threonine residues targeted by polyphosphates and result in higher average K d values for ADP and AMP vs. ATP and triphosphate (Fig. 3f-h). 20In addition, linear correlations between the aS chemical shis induced by ATP analogs versus ATP exhibit slopes comparable to the ratio of charges between the respective ATP analogs and ATP at physiological pH 7 (Fig. 3i), corroborating that charge is a primary driver of ATP binding to aS.Given this charge-dependence, we hypothesized that Mg 2+ complexation with the phosphates of ATP and the corresponding partial charge neutralization of the triphosphate moiety would inuence the ATP-aS interaction. 28gnesium modulates the interaction of ATP with aS monomers and vice versa Our data show that formation of the ATP-Mg complex attenuates but does not eliminate the signicant chemical shis induced by ATP or Mg 2+ at the N-and C-termini of aS, respectively (Fig. 4a-c).This observation is consistent with the notion that the charge neutralization of both ATP and Mg 2+ causes ATP-Mg to bind aS less strongly.As such, the increasing formation of ATP-Mg accounts for the concentration-dependent decreases in the C-terminal Mg-induced aS chemical shis as well as the absence of increasing N-terminal ATP-induced aS residue shis as more ATP is added to Mg-bound aS (Fig. 4d-f).The sequestration by ATP of Mg 2+ ions away from aS is further conrmed by CHESPA (ESI Fig. S3 †), a type of NMR chemical shi projection analysis summarized in detail by Narayanan et al. 33 Conversely, as increasing concentrations of Mg 2+ are added to ATP-bound aS, the N-terminal ATP-induced aS chemical shis exhibit concentration-dependent decreases that are not accompanied by increased C-terminal Mg-induced aS residue shis, indicating that the de-tuning effect of ATP on Mg 2+ binding to aS is reciprocal (Fig. 4g).The dynamic interplay between ATP, Mg 2+ , ATP-Mg and aS suggests that ATP could serve as a Mg 2+ "sink," buffering its effects on aS.
We next explored Mg 2+ sequestration by ATP's triphosphate moiety, and its consequences on the amount of free Mg 2+ available to bind aS, by comparing the effects of ADP versus ATP on Mg-bound aS (Fig. 4h). 28When considered relative to the Mginduced aS residue shis, the signicant C-terminal shis induced by ATP and ADP support our hypothesis that the nucleotides are able to sequester Mg 2+ away from aS in a phosphate-dependent manner (Fig. 4h).Fig. 4h also shows that the phosphate negative charges are the major driving force of the ATP-aS interaction, since despite the reduced de-tuning effect of ADP on the aS-Mg 2+ interaction, the reduced number of phosphate groups still hinders ADP's relative association with the aS N-terminus, as evidenced by the absence of increased Nterminal shis.Hence, our emerging model of an electrostatic, phosphate-driven and Mg 2+ -modulated effect of ATP on aS somewhat differs from previous hypotheses that Mg 2+ bridges ATP primarily with the aS C-terminus (Fig. 1d). 12However, it is possible that the Mg-sequestration and Mg-bridging models are not mutually exclusive and represent two viable mechanisms of ATP-aS interactions, one of which may prevail under diverse experimental conditions.
Considering the pathologically-relevant effect of free Mg 2+ on aS aggregation, we tested whether an equimolar ATP-Mg "buffer" leaves residual free Mg 2+ available to interact with aS.We measured the aS chemical shis in solutions of AMP-Mg, ADP-Mg and ATP-Mg, relative to those induced by AMP, ADP or ATP alone and saw a clear pattern of increasing, Mg-induced C-terminal aS chemical shis, suggesting that the decreasing numbers of phosphate groups in ATP, ADP and AMP, respectively, leave consecutively more Mg 2+ available to bind aS (Fig. 4i).Nevertheless, the pattern of C-terminal aS shis induced by ATP-Mg versus ATP indicate that ATP is unable to chelate 100% of equimolar Mg 2+ away from aS, suggesting that some free ligands in a biologically-relevant ATP-Mg "buffer" can bind aS (Fig. 4i). 28This is consistent with our CHESPA results, which show that ATP-Mg accentuates the effect of residual Na + cations on the aS C-terminus by shiing it farther from the aS alone state relative to the ATP-bound state (Fig. 4j). 33Meanwhile, Mg 2+ complexation with ATP shis the aS N-terminus back toward the unbound state, as evidenced by the predominantly negative cos q CHESPA values, which are consistent with our hypothesis of ATP-Mg sequestering ATP from aS (Fig. 4j). 33ven the lack of major ATP-or Mg-induced NAC-region aS chemical shis (Fig. 3c; 4c and d), we hypothesize that the negative cos q NAC-region CHESPA values report on a perturbation of long-range contacts between aS residues, which have been shown to shield the NAC region. 20,21,33We therefore tested this hypothesis using intramolecular Paramagnetic Relaxation Enhancement (PRE) NMR experiments, which report on longrange contacts between the N-and C-termini of aS monomers. 20

ATP disrupts long-range electrostatic contacts in aS monomers
To explore the effect of ATP on long-range N-to C-terminal contacts in aS monomers, we measured residue-specic PRE  c) are reproduced in (g) and (h), respectively, for ease of comparison.aS net regional charges are shown above panels (c) and (h), with dark grey boxes representing the "KTKEGV" aS repeats. 17 G 2 relaxation rates of spin-labelled S87C aS in the absence and presence of 10 mM ATP (Fig. 5). 20,41Our G 2 data reveal that ATP causes widespread decreases in residue-specic aS G 2 values that are particularly signicant for the rst ∼50 residues as well as residues 110-140, indicating that these monomer regions are comparatively farther from the S87C spin label in the presence of ATP than in its absence (Fig. 5).Based on these results, we hypothesize that the phosphate-driven targeting of ATP to the aS N-terminus disrupts long-range electrostatic contacts between the N-and C-termini of 'closed' aS monomers, causing opening (Fig. 1a). 20,21][36][37] To test this hypothesis, we measured residue-specic transverse 15 N aS R 2 amide relaxation rates in the presence of late-stage WT aS amyloid brils in both the absence and presence of ATP (Fig. 6a).We observed dramatic reductions in N-terminal aS R 2 rates upon addition of ATP to preformed brils (Fig. 6).In fact, the general effect of ATP in the presence of brils is a widespread decrease in R 2 rates that, while most signicant in the N-terminus, extends throughout almost the entire aS sequence (Fig. 6b).By contrast, 10 mM ATP does not induce dramatic changes in R 2 rates of WT aS monomers in the absence of brils (ESI Fig. S4 †).These experiments were modelled aer those of Kumari et al., who showed that the primary interaction site between aS monomers and brils is the monomeric N-terminus, which displays concentrationdependent increases in R 2 relaxation rates with increasing amounts of brils. 9A signicant enhancement of amide ))) 1/2 . 42(i) aS DCCS induced by 10 mM ATP-Mg vs. ATP, ADP-Mg vs. ADP and AMP-Mg vs. AMP.DCCS are plotted versus aS residue number and calculated for panels (c), (d), (g) and (i) relative to aS. (j) Cos q profile from CHESPA analysis of 10 mM ATP-Mg binding to aS relative to aS and aS bound to 10 mM ATP, plotted versus aS residue number and with a 0.001 ppm cut-off.q angle is between the perturbation vector from aS and 10 mM ATP to aS and 10 mM ATP-Mg and the reference vector of aS to aS and 10 mM ATP. aS spectra in the absence or presence of 10 mM ATP are shown for reference.aS regional charges are shown above plots (c), (d) and (g)-(j), with dark grey boxes representing the imperfect "KTKEGV" repeats. 17 15 N aS +/− 10 mM ATP. G 2 differences between the samples without and with ATP greater than region-specific averages plus one standard deviation are labelled.Blue-labelled residues exhibit significantly shifted ppm values by 10 mM ATP.The 10 residues on either side of 87 are not shown, and the aS regional charges are shown.Grey boxes represent "KTKEGV" aS repeats. 17elaxation rates is expected if an isotopically-labelled aS monomer residue interacts with unlabelled brils. 9Therefore, the fact that ATP abolishes the bril-dependent, N-terminal R 2 increases of aS monomers suggests that ATP inhibits interactions between aS monomers and brils, particularly at the monomer N-terminus.
Since the aS monomer-bril interactions are predominantly electrostatic, it is likely that the muted interactions in the presence of ATP are due to ATP's N-terminal binding and positive-charge neutralization of aS monomers. 9Overall, the inhibition of monomer-bril interactions and the likely consequent suppression of templated bril elongation and secondary nucleation provide a viable explanation for the ATPmediated suppression of aS cross b-sheet brils at plateau (Fig. 2b). 9,26As secondary nucleation is the dominant mechanism of both aS oligomer and bril formation under physiologically-relevant quiescent conditions, it is likely that ATP's inhibition of the aS monomer-bril contacts vital for secondary nucleation may play a role in PD-relevant aS aggregation in vivo. 35e PD-related mutations E46K and A53T alter the effect of ATP on aS Finally, we characterized how the effect of ATP on aS is inuenced by the presence of PD-related point mutations E46K and A53T, which modify residues targeted by ATP, as shown by chemical shi mapping of WT aS (Fig. 3a-c). 19Fig. 7a shows that the overall patterns of ATP-induced residue shis are similar for WT and A53T aS.By contrast, the residue-specic effect of ATP on E46K aS is extremely pronounced relative to WT, with signicant chemical shi changes occurring throughout the aS sequence (Fig. 7a).Since E46K aS monomers exhibit increased N-to C-terminal contacts relative to WT, we hypothesized that ATP binding could disrupt these electrostatic contacts and result in widespread changes in overall monomer conformation, leading to the signicant chemical shis shown in Fig. 7a. 24Indeed, intramolecular PRE experiments involving S87C E46K aS reveal larger differences in residue-specic G 2 values caused by ATP at both the N-and C-termini of E46K aS monomers relative to WT (Fig. 7b).These data suggest that ATP causes a more signicant disruption of long-range electrostatic contacts in aS monomers in the presence of the PD-related E46K mutation. 24We hypothesize that the presence of an additional positive charge in the E46K aS N-terminus leads to an enhanced electrostatic interaction with ATP, which then causes more signicant N-terminal charge neutralization and overall monomer conformational changes toward open states.
To assess whether the residue-specic effects of ATP on E46K and A53T aS monomers lead to altered, pathologically-relevant aggregation, we next measured how ATP inuences the ThT uorescence of E46K and A53T aS at plateau and analyzed the aggregated species by DLS, sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and TEM (Fig. 7c-j, ESI S5 †). 9 Contrary to its inhibitory effect on WT aS at plateau, ATP causes a signicant enhancement of E46K aS plateau ThT uorescence, suggesting that the E46K mutation eliminates the inhibitory effect of ATP on late-stage aS b-sheet brillization (Fig. 7c).By contrast, the effect of ATP on A53T aS plateau ThT uorescence is not signicant (Fig. 7d), suggesting nonetheless that the A53T mutation attenuates the inhibitory effect of ATP on aS brillization (Fig. 2b).Consistent with the ThT data, our TEM images (Fig. 7g-j) show that E46K and A53T aS brils form in both the absence and presence of ATP.
In addition, our DLS data (Fig. 7e) shows that ATP shis the population of soluble A53T aS away from small, low-molecularweight species towards intermediately-sized (∼100 nm) aggregates in a manner distinct from its effect on WT.This effect is moreover independently conrmed by SDS-PAGE (Fig. 7f), which shows that the A53T mutation increases the population of High-Molecular-Weight (HMW) assemblies relative to Low-Molecular-Weight (LMW) species.Interestingly, similar mutation-induced DLS and SDS-PAGE changes are observed for E46K aS, albeit less extreme (Fig. 7e and f).Nonetheless, these changes are likely pathologically-signicant, as Emin et al. recently showed that soluble, smallbut not monomeric -aS oligomers induce the greatest release of tumour necrosis factor a, a pro-inammatory cytokine linked to PD progression, upon addition to mouse microglia. 31As the size of aS aggregates correlates inversely with their cytotoxicity, the fact that ATP affects the size distribution of soluble aS aggregates is likely particularly pathologically-relevant. 31 Overall, our ThT, TEM, DLS and SDS-PAGE data consistently indicate that the A53T and E46K aS mutations markedly inhibit the ATP-induced suppression of intermediate-size and brillar assemblies observed for WT aS, consistent with their pronounced earlyonset PD phenotype. 19

Experimental
Detailed experimental procedures and ESI data are described in the ESI.†

Conclusions
Here we provide a residue-resolution picture of the aS-ATP interactions and show how ATP inuences the aggregation of  WT aS as well as its PD-related variants E46K and A53T. 19Our results reveal that ATP directly binds aS via its triphosphate moiety and elicits a direct and bimodal effect on WT aS aggregation by disrupting long-range electrostatic contacts in monomers to shorten the aggregation lag time (Fig. 8).ATP also inhibits late-stage cross b-sheet aS bril formation as well as the aS monomer-bril interactions needed for secondary nucleation (Fig. 8). 9We also show that Mg 2+ inhibits the ATP-aS interactions and vice versa, suggesting that ATP-Mg serves as a "sink" to buffer the amount of free ATP and Mg 2+ available to interact with aS (Fig. 8).4][45][46][47][48][49] As well, our results reveal that the loss of this novel function of ATP caused by known aS pathogenic mutations offers a new perspective on early-onset PD progression. 19

Fig. 1
Fig.1Hypothetical models of aS-ATP interactions.(a) Upper panel: WT aS amino acid sequence, containing nine pseudo-apolipoprotein-like repeats (underlined), acidic (bolded) and basic (italicized) residues.17The N-terminus, NAC region and C-terminus are colored in blue, yellow and red, respectively.16Lower panel: general aS aggregation mechanism involving disruption of the long-range N-to C-terminal monomer contacts, leading to opening and subsequent aggregation of aS into oligomers and fibrils.[21][22][23][24](b) One hypothetical model of the ATP-aS interaction based upon the interaction of ATP with proteins in the crystalline lens, whereby the adenine group of ATP clusters over protein hydrophobic patches and the triphosphate electrostatically repels other bound monomers to inhibit aggregation.25(c) Another viable model for ATP-aS interactions, suggesting that ATP could bridge aS monomers and enhance aggregation via the phosphate-mediated targeting of lysine residues (designated by "+" symbols) in the aS N-terminus, as is the case for tau protein.6(d) Model for Mg 2+ -mediated ATP-aS interaction proposed by Nishizawa et al., whereby Mg 2+ ions interact with the aS C-terminus and "bridge" indirect, non-specific interactions between the protein and ATP.12

Fig. 2
Fig. 2 ATP elicits a concentration-dependent effect on aS aggregation that is phosphate-driven and is preserved in the presence of magnesium.(a) ThT fluorescence of fresh 300 mM WT aS in ThT buffer (20 mM K 2 HPO 4 , 5 mM KH 2 PO 4 , 100 mM KCl, 200 mM EDTA, 0.05% NaN 3 ), incubated for 85 h in a 37 °C plate reader.Plotted are average ThT measurements across four wells per condition, with measurements taken every six minutes with 30 s orbital shaking prior to each read and normalized to the final measurement for the aS sample.Inset shows non-normalized data, with standard deviations for the aS and aS + 10 mM ATP conditions shown.Error bars for ATP < 10 mM are not shown to avoid overcrowding, but are comparable to 10 mM ATP.(b) ThT fluorescence of fresh 300 mM WT aS in ThT buffer, incubated for 72 h in a 37 °C shaker at 150 rpm then in a plate reader for 20 additional hours at plateau.Plotted are well-specific average ThT measurements from multiple independent experiments, taken at plateau every five min with 30 s orbital shaking prior to each read and normalized to the average measurement of all aS samples.Ligand concentrations are all 10 mM.Chemical structures of AMP, ADP, ATP and triphosphate at pH 7.4 are shown above the panel.(c) ThT fluorescence of fresh 440 mM WT aS in ThT buffer, incubated for 47 h in a 37 °C plate reader.Plotted are average ThT measurements across four wells per condition, with measurements taken every six minutes with 30 s orbital shaking prior to each read and normalized to the final measurement of the aS + 10 mM MgCl 2 sample.(d) Well-specific average ThT fluorescence measurements of panel (c) samples at plateau (40-47 h), normalized to the average measurement of all aS + 10 mM MgCl 2 samples.Straight lines between samples in panels (b) and (d) represent no significant difference.Sample comparisons in panels (b) and (d) represent significance levels: * = p < 0.05, *** = p < 0.001 and **** = p < 0.0001.

Fig. 3
Fig. 3 The triphosphate moiety of ATP drives electrostatic interactions between ATP and the N-terminal pseudo-apolipoprotein repeats of WT aS.(a and b) 1 H-15 N HSQC spectral regions of WT aS monomers with increasing ATP concentrations.(c) ATP-induced Compounded Chemical Shifts (DCCS) of WT aS monomers, with DCCS greater than the region-specific average shift (dashed line) plus one standard deviation (dotted line) labelled.(d) Cos q cross-peaks > 0.97 for aS WT ("apo") versus 10 mM ATP-bound aS WT ("holo").(e) 0-20 mM ATP-induced DCCS of significantly-shifted N-terminal aS residues fitted to a one-site specific binding model. 38-40Upper and lower-limit model-calculated K d values are shown.(f) Average significant aS DCCS induced by AMP, ADP, ATP or triphosphate, fitted to one-site specific binding models and with approximate fitted K d values shown.Error bars represent the standard deviation of well-resolved peaks at each concentration.(g) HSQC spectral regions used to calculate panel (h) DCCS.(h) aS DCCS induced by 10 mM AMP, ADP, ATP or triphosphate.The HSQC spectrum of aS in the presence of 10 mM ATP and the resulting ATP-induced aS DCCS profile shown in (a) and (c) are reproduced in (g) and (h), respectively, for ease of comparison.aS net regional charges are shown above panels (c) and (h), with dark grey boxes representing the "KTKEGV" aS repeats.17(i) Correlations between 10 mM ATP-induced aS DCCS and those induced by 10 mM AMP, ADP or triphosphate.Approximate slopes are shown with lines of best fit and errors.Structures of AMP, ADP, ATP and triphosphate at pH 7.4 are shown.DCCS were calculated as DCCS = (0.5*((dHAMP, ADP, ATP or triphosphate − dH aS ) 2 + (0.15*(dN AMP, ADP, ATP or triphosphate − dN aS ) 2 ))) 1/2 . 42 Fig. 3 The triphosphate moiety of ATP drives electrostatic interactions between ATP and the N-terminal pseudo-apolipoprotein repeats of WT aS.(a and b) 1 H-15 N HSQC spectral regions of WT aS monomers with increasing ATP concentrations.(c) ATP-induced Compounded Chemical Shifts (DCCS) of WT aS monomers, with DCCS greater than the region-specific average shift (dashed line) plus one standard deviation (dotted line) labelled.(d) Cos q cross-peaks > 0.97 for aS WT ("apo") versus 10 mM ATP-bound aS WT ("holo").(e) 0-20 mM ATP-induced DCCS of significantly-shifted N-terminal aS residues fitted to a one-site specific binding model. 38-40Upper and lower-limit model-calculated K d values are shown.(f) Average significant aS DCCS induced by AMP, ADP, ATP or triphosphate, fitted to one-site specific binding models and with approximate fitted K d values shown.Error bars represent the standard deviation of well-resolved peaks at each concentration.(g) HSQC spectral regions used to calculate panel (h) DCCS.(h) aS DCCS induced by 10 mM AMP, ADP, ATP or triphosphate.The HSQC spectrum of aS in the presence of 10 mM ATP and the resulting ATP-induced aS DCCS profile shown in (a) and (c) are reproduced in (g) and (h), respectively, for ease of comparison.aS net regional charges are shown above panels (c) and (h), with dark grey boxes representing the "KTKEGV" aS repeats.17(i) Correlations between 10 mM ATP-induced aS DCCS and those induced by 10 mM AMP, ADP or triphosphate.Approximate slopes are shown with lines of best fit and errors.Structures of AMP, ADP, ATP and triphosphate at pH 7.4 are shown.DCCS were calculated as DCCS = (0.5*((dHAMP, ADP, ATP or triphosphate − dH aS ) 2 + (0.15*(dN AMP, ADP, ATP or triphosphate − dN aS ) 2 ))) 1/2 . 42

Fig. 4
Fig. 4 Magnesium modulates the ATP-aS monomer interaction and vice versa.(a and b) 1 H-15 N HSQC spectral regions of aS in the absence or presence of 10 mM ATP, ATP-Mg or MgCl 2 , colored as per legend in (a).(c) DCCS from panels (a) and (b) spectra.(d) aS DCCS calculated from panels (e) and (f) spectra, showing aS pre-incubated overnight with 10 mM MgCl 2 or with the subsequent addition of 5 or 10 mM ATP. Panels (e) and (f) are colored as per legend in (e) and include the aS alone spectrum for reference.(g) 10 mM ATP-induced aS DCCS from pre-incubation overnight or with subsequent addition of 5 or 10 mM MgCl 2 .(h) 10 mM MgCl 2 -induced aS DCCS from preincubation overnight or with subsequent addition of 5 or 10 mM ADP or 10 mM ATP. DCCS for this panel calculated as: DCCS = (0.5*((dH (ADP or ATP)+10 mM MgCl 2 − dH aS+10 mM MgCl 2 ) 2 + (0.15*(dN (ADP or ATP)+10 mM MgCl 2 − dN aS+10 mM MgCl 2 ) 2))) 1/2 . 42(i) aS DCCS induced by 10 mM ATP-Mg vs. ATP, ADP-Mg vs. ADP and AMP-Mg vs. AMP.DCCS are plotted versus aS residue number and calculated for panels (c), (d), (g) and (i) relative to aS. (j) Cos q profile from CHESPA analysis of 10 mM ATP-Mg binding to aS relative to aS and aS bound to 10 mM ATP, plotted versus aS residue number and with a 0.001 ppm cut-off.q angle is between the perturbation vector from aS and 10 mM ATP to aS and 10 mM ATP-Mg and the reference vector of aS to aS and 10 mM ATP. aS spectra in the absence or presence of 10 mM ATP are shown for reference.aS regional charges are shown above plots (c), (d) and (g)-(j), with dark grey boxes representing the imperfect "KTKEGV" repeats.17

Fig. 5
Fig.5ATP disrupts long-range N-to C-terminal interactions in aS monomers.Residue-specific G 2 values for spin-labelled, fresh 120 mM S87C15 N aS +/− 10 mM ATP. G 2 differences between the samples without and with ATP greater than region-specific averages plus one standard deviation are labelled.Blue-labelled residues exhibit significantly shifted ppm values by 10 mM ATP.The 10 residues on either side of 87 are not shown, and the aS regional charges are shown.Grey boxes represent "KTKEGV" aS repeats.17

Fig. 6
Fig. 6 ATP inhibits N-terminally-driven aS monomer-fibril interactions.(a) 15 N-R 2 profiles of mM WT aS monomers in the presence of 1.3 mM WT 14 N fibrils +/− 10 mM ATP.(b) R 2 differences between the two profiles shown in panel a.

Fig. 7
Fig. 7 Effect of ATP on E46K and A53T aS.(a) 10 mM ATP-induced DCCS of WT, E46K and A53T aS monomers, calculated relative to the corresponding variant alone.(b) PRE-derived G 2 value differences induced by 10 mM ATP on spin-labelled, fresh 120 mM 15 N S87C or S87C E46K aS, calculated relative to the corresponding S87C variant alone.The ten residues on either side of 87 are not shown.(c and d) Well-specific plateau ThT fluorescence measurements of fresh 300 mM E46K (c) or A53T (d) aS in ThT buffer +/− 10 mM ATP, normalized to the average measurement of the corresponding variant protein alone and incubated as for Fig. 2b.(e) DLS measurements of soluble aS from panels (c) and (d) plateau samples, following centrifugation to pellet large aggregates.(f) SDS-PAGE of soluble aS from panels (c) and (d) samples following centrifugation that were subsequently either retained by (IMW) or passed through (LMW) a Pall Nanostep 100 kDa centrifugal filter.Resuspended centrifugation pellets constitute HMW samples.Bands are in reference to PageRuler Prestained Protein Ladder (26616, Thermo Scientific), shown in the left-most lane.Numbers indicate a quantification of the ∼15 kDa monomer band intensities by ImageJ, expressed as the ratio of the HMW/ LMW lane monomer bands for each sample.(g-j) Negative stain TEM images of pelleted large aggregates from panels (c) and (d) plateau samples: E46K (g) or A53T (i) aS alone or in the presence of 10 mM ATP (h and j).All scale bars represent lengths of 400 nm.

Fig. 8
Fig. 8 Proposed summary model of aS/ATP interactions and their effects.