Unveiling transient protein-protein interactions that modulate inhibition of alpha-synuclein aggregation by beta-synuclein, a pre-synaptic protein that co-localizes with alpha-synuclein

Pathology in Parkinson’s disease is linked to self-association of α-Synuclein (αS) into pathogenic oligomeric species and highly ordered amyloid fibrils. Developing effective therapeutic strategies against this debilitating disease is critical and βS, a pre-synaptic protein that co-localizes with αS, can act as an inhibitor of αS assembly. Despite the potential importance of βS as an inhibitor of αS, the nature, location and specificity of the molecular interactions between these two proteins is unknown. Here we use NMR paramagnetic relaxation enhancement experiments, to demonstrate that βS interacts directly with αS in a transient dimer complex with high specificity and weak affinity. Inhibition of αS by βS arises from transient αS/βS heterodimer species that exist primarily in head- to- tail configurations while αS aggregation arises from a more heterogeneous and weaker range of transient interactions that include both head-to-head and head-to-tail configurations. Our results highlight that intrinsically disordered proteins can interact directly with one another at low affinity and that the transient interactions that drive inhibition versus aggregation are distinct by virtue of their plasticity and specificity.

One approach to developing effective therapeutic strategies against this debilitating disease is to identify inhibitors of α S aggregation. Small molecule inhibitors of α S have been proposed [9][10][11][12][13] . In addition, proteins such as heat shock proteins (Hsp40, Hsp70, Hsp90, α B-crystallin) 14,15 , and the intrinsically disordered β -synuclein (β S), a homologue of α S with which it co-localizes have been shown to interfere with α S assembly. A number of studies have established a neuroprotective role for β S [16][17][18][19][20][21][22][23][24] . Masliah and co-workers have shown that β S is expressed at similar levels as to α S in the central nervous system. However the ratio of β S to α S at the mRNA level is significantly decreased in diseased brains, suggesting a regulatory role within the synuclein family 24 . In vivo it has been shown that over-expression of α S with β S in mouse models significantly decreases the number of plaques formed 25 and that intracerebral injection of the lenti-β S virus reduces the formation of α S inclusions in transgenic mice 26 . In vitro, it has been shown that the presence of β S with α S slows its aggregation [16][17][18][25][26][27] . Despite the fact that β S has a very similar sequence to α S, it does not form fibrils on its own 16,[28][29][30] , but may form aggregates whose toxicity is debated 31,32 . The in vivo data clearly suggest that β S plays an important regulatory role in inhibition of α S pathology but at this stage there is no molecular information about the nature, location and specificity of the protein-protein interactions that initiate the inhibition of α S by β S.
To understand the mechanism by which the intrinsically disordered protein β S interacts with α S, we use NMR to map the monomer-monomer interactions that lead to inhibition or promotion of aggregation. Despite the importance of these interactions, the molecular details are extremely difficult to obtain due to their transient nature and low population. Paramagnetic relaxation enhancement experiments (PRE) offer an excellent tool for characterization of weak and transient interactions because they are able to probe states that exist at low populations (even 0.5-5%) and exhibit short life times (250-500 μ s) 33,34 .
Here we use inter-chain NMR PRE experiments to identify and characterize weak transient complexes of α S and β S 35 . We show that α S homo-dimers sample a heterogeneous range of population distributions, including head-to-head and head-to-tail configurations, while α S/β S hetero-dimers exist primarily in head-to-tail configurations.
NMR inter-chain PRE titration experiments previously used on folded proteins 36 are applied here to intrinsically disordered proteins. These experiments allow us to obtain residue specific dissociation constants to inform us about the specificity and affinity of dimer interactions in the different regions of the transient disordered complexes. Our results show that the hetero-dimer transient head-to-tail interactions between α S and β S are approximately 5 times stronger than the interactions observed in the homo-dimer α S species suggesting that these α S/β S interactions create a kinetic trap which delays or inhibits the formation of α S fibrils. The novel insight presented in this paper not only defines residue specific contacts between two intrinsically disordered proteins but also links the homo and hetero-complexes with distinct pathways that lead to aggregation versus inhibition.

Results
βS inhibits αS fibril formation in a dose dependent manner. α S and β S are part of the synuclein family. These can be described as: the N-terminus that contains KTKXGV repeats and forms helices at membranes 37 , the non-amyloid-β S component (NAC) region, and the highly acidic and solubilizing C-terminus (Fig. 1). α S and β S have similar sequences, particularly at the N-terminus, but very different fibrillation and oligomerization properties (Fig. 1). The N-terminus for all synucleins is highly conserved, with only 6 substitutions between α S and β S sequences. In contrast, the C-terminus is the least conserved region with more prolines and more negatively charged residues with a net charge of -12 in α S and -15 in β S. β S has an 11 residue deletion in the NAC region that was thought to be important in preventing fibril formation but substitution of this region into β S does not recover full fibrillation potential of α S 28,29 .
Similarly to α S it has been established that the physiological form of β S and its pathological mutants in vivo are N-terminally acetylated 38 . All experiments in this study are performed on the acetylated forms of the protein that we will refer to as α S and β S. All previous characterization of β S was performed on non-acetylated protein 16,17,28,39 , therefore we use NMR and other biophysical approaches here to determine whether N-terminal acetylation affects the conformation or oligomerization state of the protein. NMR and other biophysical techniques including dynamic light scattering (DLS) and circular dichroism (CD) show that acetylated β S, similarly to acetylated α S, is primarily monomeric and unfolded (Supplementary Fig. 1 & 2). Secondary structure propensities indicate the formation of a transient N-terminal helix relative to the non-acetylated form of β S similarly to α S (Supplementary Fig. 1 & 2). In addition, the C-terminus of acetylated β S is more extended that the C-terminus of acetylated α S consistent with previous results on non-acetylated protein 40 . Electrospray ionization mass spectroscopy (ESI-MS) experiments (Supplementary Fig. 1) show that acetylated β S can populate an extended and a compact form with a higher population of extended conformation relative to acetylated α S.
Addition of acetylated β S to acetylated α S inhibits fibril formation in a dose dependent manner, consistent with previous findings for the non-acetylated forms of the proteins ( Fig. 2A) 16 . As the concentration of β S is increased we observe a significant change in the rate of the elongation phase and in the total ThT intensity as well as a small change in the lag phase. Previous work has shown that ThT can be used as a semiquantitative method to estimate relative amounts of fibril formed, although caution should be applied in interpreting ThT intensities 41,42 . In this case, the inhibition of fibril formation as shown by the ThT experiments is supported by Transmission Electron Microscopy (TEM) data which shows significant changes in fibril morphology of α S/β S relative to fibrils of α S alone ( Fig. 2B-D). β S does not form fibrils and the α S/β S mixture forms significantly fewer and shorter fibrils than α S alone as seen in TEM (Fig. 2C,D).

Mapping of residue specific transient interactions in αS/αS homo-and αS/βS hetero-dimer complexes using NMR inter-chain PRE experiments.
ThT fluorescence experiments and TEM data ( Fig. 2) have established that β S alters the aggregation kinetics of α S, however there has been no evidence to date of a direct interaction between these proteins. To determine the existence of, and to characterize transient inter-chain interactions between homo-dimers (α S/α S, β S/β S) and hetero-dimers (α S/β S, β S/α S), inter-chain NMR paramagnetic relaxation enhancement experiments were performed. In this experiment, NMR blind 14 N-MTSL labeled protein is mixed with NMR visible ( 15 N) unmodified protein and broadening of signal is limited to residues on the NMR visible chain that interact with the MTSL labels on the NMR blind chain (Supplementary Fig. 3). Signal broadening therefore reflects interactions between the NMR blind and NMR visible protein [33][34][35]43 .  Supplementary Fig. 4). There is evidence for interactions between the homo-dimer complexes of α S, and the hetero-dimer complexes of α S and β S; interactions between β S are essentially non-existent. We have introduced four spin labels along the sequence at positions 11, 44, 90 and 132 for α S and 11, 44, 80, 134 for β S to probe interactions in the N, NAC and C-terminal regions. Experimental results are presented as heat maps ( Fig. 3A-D), where each strip shows color-coded values of residue-specific inter-chain paramagnetic relaxation enhancement rates (PRE rate-Γ ) induced by the proximity of the MTSL label to another chain. Strong interactions (> 12 Hz) are defined relative to the lack of interactions otherwise observed in the β S homo-dimer contact maps. Under the conditions of the experiment, α S and β S do not form fibrils or oligomers; therefore dimer detection arises as a result of low populations of dimers existing in equilibrium with the monomer precursor (for α S lack of higher order species in the sample has been confirmed by ESI-IMS-MS experiments 44 and for hetero-species this has been confirmed with ESI-MS. αS populates a heterogeneous range of transient complexes while αS/βS hetero-complexes sample primarily head-to-tail non-propagating interactions. The heat map for the α S/α S homo-dimer shows strong interactions between the N-terminal α S-MTSL labeled positions A11 and T44 with the N-terminal region 36 to 44, and the C-terminal region 124 to 140 (Fig. 3A). Earlier studies by our group 35 on α S showed transient inter-chain interactions between the N-and C-termini (N-C), but by increasing the number and positions of the MTSL spin labels we now observe new inter-chain interactions between the N-termini showing that α S can populate multiple dimer configurations. In noticeable contrast, the α S/β S hetero-dimers show strong interactions between the N-terminal α S-MTSL labeled positions A11 and T44 and the C-terminus from residues 105 to 134 and extremely weak N-N terminal interactions between 37 and 41 (Fig. 3B). According to the heat map, the hetero-dimer interactions between the N-terminus of α S and the C-terminus of β S appear to be more extensive and stronger (residues 105-134 compared to residues 124-140 in α S/α S homo-dimers) ( Fig. 3A-C) than those in α S. β S shows extremely minimal interactions with itself supporting the view that β S does not form fibrils (Fig. 3D).
A schematic representation of the dominant homo and hetero-dimer interactions shows that both N-N and N-C configurations are sampled by the α S homo-dimers while only N-C terminal interactions are sampled in the α S/β S hetero-dimers (Fig. 3E,F). The favorable interactions of the N-terminal hydrophobic region of α S with itself (N-N), and with the C-terminus of α S and β S suggest that they act as an aggregation initiation region. The highly interactive N-terminal region encompassing residues 38-45 is referred to as 'hot spot' region (Suppl. 4A). In the α S homo-complexes, the interactions detected by Nand C-terminal probes show symmetry, implying that the interactions we observe are not experimental artifacts. The NMR PRE data show that α S homo-dimers can sample a heterogeneous range of populations, including head-to-head and head-to-tail configurations while α S/β S hetero-dimers sample only head-to-tail dimers. This suggests that the hetero-dimers have a more limited range of conformational preferences for the dimer species.
Aggregation versus inhibition of αS by βS is due to a balance between specificity and affinity of transient interactions. The inter-chain NMR PRE experiments described above are powerful as they provide us with direct evidence for the existence of transient interactions and allow us to pinpoint the specific residues involved in these encounter complexes, however the specificity and affinity of the interactions remains unknown. We extend an earlier approach 36,45 designed for folded proteins to transient encounter complexes of IDPs. We obtain residue-specific equilibrium dissociation constants (K D ) by performing titrations of 15 N labeled α S with MTSL-labeled 14 N-α S. K D s are obtained by fitting the intermolecular transverse 1 H relaxation rates to a titration curve (see methods) (Fig. 4).
NMR PRE titration experiments and data analysis is complex as K D s are anticipated to be weak due to the disordered nature of the monomers. The titration curves are grouped according to their profiles and three different types of patterns emerge (Fig. 4A-C). Representative examples of titration curves arising from interactions between α S-44-MTSL and residues 38-41 of the N-terminus (Fig. 4A), and residues 125-140 of the C-terminus (Fig. 4B), show that they are distinct from one another. The titration curves for the N-N interactions show a linear dependence between PRE values and concentration indicating non-specific interactions at this site, while interactions with the C-terminal 125-140 exhibit non-linear titration curves suggesting specific interactions. The range of K D values in this group is between K D ~ 500 μ M (range 90-1200 μ M) using the data analysis described in methods. While we observe both non-specific (N-N) (Fig. 4A) and specific interactions (N-C) (Fig. 4B) in homo-α S complexes, the hetero-complexes α S/β S (Fig. 4C) have only specific N-C interactions. The titration profiles of α S-44-MTSL with β S in the region 115-134 have a narrower range of K D values (K D ~ 100 μ M, range 40-350 μ M) than those associated with the K D values of α S-44-MTSL α S with its own C-terminus (K D ~ 500 μ M, range 90-1200 μ M) suggesting more uniform behavior across this region and higher specificity and affinity by approximately 5 fold (Supplementary Tables 1&2). The differences between the strengths of the interactions and the range of residues over which the interactions are occurring is seen clearly in the 3D plots where the interaction regions in α S are more rugged while the interactions between α S and β S are smoother, more uniform and extend over a wider range of residues (Fig. 4D,E).

Discussion
β S plays a role in the inhibition of α S aggregation but the mechanism by which this occurs and the stage in the aggregation pathway at which β S first interacts with α S has been unknown. We demonstrate that the monomer species of α S and β S interact directly with one another at specific sites suggesting that inhibition may begin at the very earliest stages of the fibril formation process. The molecular interactions and affinities obtained in the NMR PRE experiments described here support the view that early stages of aggregation versus inhibition may be due to a balance between conformational heterogeneity, specificity and affinity. Early stages of aggregation in α S may be promoted by sampling or searching conformational structures with weak transient affinities, including both non-specific head-to-head and weak specific head-to-tail dimers (Fig. 5A). In contrast early stages of inhibition may be favored by sampling only head-to-tail interactions with higher affinity and specificity within the α S/β S hetero-complex (Fig. 5B).
Head-to-tail interactions exist in both the homo-and hetero-dimer complexes, which strongly suggests that they may play a regulatory role on α S folding or misfolding. In light of the fact that the fibril state of α S assembles into in-register parallel cross-β -structure 46,47 , the NMR data suggests that head-to-tail α S/β S dimers would have to undergo conformational rearrangement to reach the final fibril form. This may thereby delay or inhibit the kinetics of fibril formation. We propose that the α S homo-complexes are more aggregation prone than α S/β S complexes for two reasons: first, α S can sample head-to-head interactions that are potentially aggregating promoting while α S/β S does not sample these; second, the non-propagating head-to-tail α S/α S complex has weaker, lower affinity interactions relative to the α S/β S complex suggesting that the conformational rearrangement required for fibril formation may be more facile for the α S/α S homo-complex.
Despite the fact that both the α S/α S homo-dimer and the α S/β S hetero-dimer sample head-to -tail interactions, there are notable differences in terms of the strength of the interactions and the range over which the interactions extend. For homodimers the interaction of the α S "hot spot" with the C-terminus of α S extends over ~15 residues, while the interaction of the α S "hot spot" with the C-terminus of β S extends over a broader range of residues from 105 to 134. In both cases the dimers are flexible and are able to probe a big surface area, possibly adopting multiple dimer conformations. In addition, the α S and β S C-termini are highly negatively charged and contain aromatic and hydrophobic residues suggesting that initial interactions are mediated by electrostatics, which are anchored and stabilized through the hydrophobic and aromatic interactions.
Using NMR PRE titration experiments we demonstrate that interactions between the C-terminus of β S and the N-terminus of α S are approximately 5 times stronger and more extensive than the interactions of Titration profiles for α S/β S show a higher degree of saturation consistent with more specificity and higher affinity than those for α S/ α S (D,E) 3D representation of titration curves in the C-terminus (D) 15 N-α S C-terminus titrated with T44C-MTSL labeled 14 N-α S (E) 15 N-β S C-terminus titrated with T44C-MTSL labeled 14 N-α S titration. The x-axis shows the C-terminal residues of both α S and β S after sequence alignment, the z-axis depicts the PRE values, the y-axis the ratio of T44C-MTSL labeled 14 N-α S to 15 N sample concentration. The surface is colored using a rainbow palette, where low H N Γ 2 values are red and the highest values are purple (according to legend below the plot).
the C-terminus of α S with its own N-terminus. This increased specificity and affinity may be attributed to two factors: a higher content of negative C-terminal charges thereby enhancing electrostatic interactions, and a higher proportion of proline residues thereby altering the conformational ensembles sampled by the C-terminus. NMR and ESI data indicate that the C-terminal of β S has a higher population of more extended species (Supplementary Fig. 1) and may therefore provide a more accessible surface area for interactions with the N-terminus "hot spot' of α S. These differences suggest that small changes in the binding surface, even in these highly dynamic IDP complexes, can lead to substantial changes in interactivity that can modulate the pathway of protein aggregation versus inhibition.
In conclusion, our studies highlight that transient and weak interactions are important for protein recognition pathways of IDPs that lead to diseases such as amyloidosis where the proteins self-associate and propagate to highly ordered fibrils. Work by Radford et al. 48 have shown that weak interactions are also important for folded proteins in directing aggregation versus inhibition. By performing NMR inter-chain PRE titration experiments we have identified and characterized the strength and affinity of transient interactions between α S and β S, both IDPs. As IDPs are highly flexible, binding affinities are likely to behave in a non-cooperative manner across the protein 49 . Our data support this view and highlight the variable nature of the binding affinities that result in aggregation promoting versus aggregation inhibiting configurations.
Knowledge about the distinct dissociation constants of different interaction regions provides a new framework for thinking about therapeutic intervention by providing direct information about which regions to target for small molecule intervention and about the conformational features that may be most effective at intervention. There have been some small molecule inhibitors that have targeted the N-terminus of α S but now based on our detailed molecular understanding of α S/β S interactions we can optimize the surface interactions to design novel inhibitors [10][11][12][50][51][52] . In addition, the powerful methods used here to identify the early stages of interaction of α S with β S can be extended to design inhibitors with biologics and can be applied even more widely to study other cross amyloid interactions, such as those between α S and amyloid-β -protein, or α S and tau, that have been shown to play a critical role in cross-seeding in neurodegenerative disease 48,[53][54][55][56] .  Supplementary Fig. 4. The NMR data show that α S homo-dimers can sample a heterogeneous range of poulations including N-to-N and N-to C-terminal configurations while α S/β S hetero-dimers sample only N-to C-terminal configurations with higher affinity and specificity. In light of the fact that the fibril form of α S forms β -strands that assemble into in register parallel β -sheets this would suggest that the N-C terminal dimers would have to undergo conformational rearrangement to reach the final fibril form, thereby delaying or inhibiting the kinetics of fibril formation.
Scientific RepoRts | 5:15164 | DOi: 10.1038/srep15164 Methods Mutagenesis, expression and purification. Cysteine mutants of α S (A11C, T44C) and β S (A11C, T44C, A80C, A134C), were prepared by site-directed mutagenesis using AccuPrime pfx from Invitrogen. To obtain N-terminal acetylated forms of α S and β S proteins, co-expression with the NatB plasmid N-Acetyltransferase B was performed, as described previously 44 . Protein purification was performed according to previous protocols 57 . Similarly, MTSL spin label conjugation to cysteine mutants was performed using previously established protocols 35 35 . Samples were prepared as follows: lyophilized samples of 14 N-MTSL-cysteine mutants or 15 N non-modified proteins (α S or β S) were separately dissolved. Samples were passed through a 100 kDa filter to remove higher order oligomers, and then concentrated using 3 kDa filters to be able to dilute the sample to a final concentration of 250 uM. Low sample concentration was chosen to minimize non-specific interactions. The total sample concentration was 500 uM, with 250 uM non-modified 15 N protein and 250 μ M 14 N-MTSL labeled protein. All combinations of proteins were mixed in order to see all possible interactions. Diamagnetic samples were prepared by reducing samples with 10× excess of Ascorbic Acid and 5× buffer exchange using 3kDa cutoff filters from Millipore Inc. All the controls have similar patterns and are in the range of experimental error. Additionally, the pattern for the reduced diamagnetic control is consistent with the pattern for the mixture of 14 N and 15 N non-modified samples. The 1 H-15 N HSQC of the cysteine mutants with the reduced spin label did not disrupt the HSQC pattern of α S and β S.
All 1 H-R 2 measurements of paramagnetic and diamagnetic (reduced) samples were acquired on a 600 MHz Varian at 15 °C using previously published pulse sequence and protocols 33,35 . The inter-chain paramagnetic relaxation enhancement rate (PRE rate-Γ 2 ) is the residue specific difference of the 1 H-R 2 values of the paramagnetic and diamagnetic samples. Increased 1 H-R 2 relaxation rates on the 15 N visible chain indicates that the NMR blind 14 N -MTSL labeled protein is in the proximity of specific residues in the NMR 15 N labeled visible chain. Para-and diamagnetic 1 H-R 2 were analyzed and processed using nmrpipe 58 and sparky 59 . For all experiments 10 relaxation delays were used: 12, 32, 104, 12, 124, 64, 48, 94, 64, 20 ms. Two data points (12 ms, 64 ms) were repeated in the experiment to obtain good statistics for the error analysis. Errors of Γ 2 were calculated using error propagation, and errors were below 2 Hz. All of the interactions that were considered significant were at least 2 times higher than the mean value, and all of them were higher than the 3rd quartile and at least 8Hz.

NMR PRE titration experiments.
For NMR PRE titration experiments we used protocols described before 36,45 . Spin label sample concentrations were reduced to low volume > 32 uL and were added to 350uL of 250uM 15 N sample of either α S or β S; the changes in the sample volume were less that 10% of the overall sample volume. For α S titrations we used the following ratio of 14 N-α S-44-MTSL labeled samples to 15 N NMR visible samples: 0, 0.25, 0.5, 0.75, 1, 1.5, and in the case of β S we went up to ratio of 2. α S in this ratio exhibited shifts in the HSQC spectra, thus we removed this point from the analysis. We ran 6 data points ranging from 12-125 ms with the first point repeated twice for statistics. The PRE profiles during the titration did not show significant contributions from the non-interactive regions. The relaxation rates for the non-interactive NAC region for the titration ratio 15  where x is the concentration of 14 N-α S (T44C-MTSL) in solution, Γ 2 free -represents paramagnetic relaxation enhancement for free protein, and Γ 2 bound represents the maximum observed saturation value. The fitting scheme was based on the papers by Bax 45 and Clore 36 . We used a nonlinear regression model with three fitting parameters Γ 2,free , Γ 2 bound and K d . The χ -squared statistic that measures the difference between the observed and predicted increase in 1 H N Γ 2 app was optimized using the R statistics package minpack.lm, via the Levenberg-Marquardt algorithm 60 . The resulting fit was robust to small changes in Γ 2 free , K D , and Γ 2 bound and provided us with residue specific K D values between 14 N-α S-44-MTSL and 15 N-α S or 15 N-β S. The K d calculations were performed for residues whose 5 th titration point had PRE values higher that 15 Hz (Supplementary table 1&2). Results are summarized in the supplementary tables 1&2.
Thioflavin T (ThT) aggregation assays. The experimental set up for measuring time dependent fibril formation with ThT has been previously described 57 . The conditions used for the ThT assay match Scientific RepoRts | 5:15164 | DOi: 10.1038/srep15164 the conditions of the NMR PRE experiments described earlier. 5-10 mg of lyophilized α S and β S was dissolved in 10 mM MES, pH 6, and centrifuged for 10 min. at 14000 rpm to remove big oligomers, and purified using size exclusion chromatography (Superdex 75 GL 10/300, from GE Healthcare Life Sciences). Protein was subsequently concentrated using 3kDa centrifugal units (Millipore Inc). The final concentration of α S and/or β S was 70 μ M. Addition of β S to α S was performed in multiples of 70 μ M. ThT fibrillation rates of the following samples were measured: (1) α S alone (2) β S alone; (3) ThT fluorescence of 1:1 the mixture α S + β S, and (4) ThT fluorescence of 1:5 the mixture of α S+ 5xβ S. Each condition was repeated 3 times in 37 °C, with linear shaking in the presence of the Teflon beads. For each sample 3 curves are plotted on the figure to show the variability of the aggregation profiles.