Characterizing Methyl-Bearing Side Chain Contacts and Dynamics Mediating Amyloid β Protofibril Interactions Using ¹³C_methyl-DEST and Lifetime Line Broadening

Abstract

Many details pertaining to the formation and interactions of protein aggregates associated with neurodegenerative diseases are invisible to conventional biophysical techniques. We recently introduced ¹⁵N dark-state exchange saturation transfer (DEST) and ¹⁵N lifetime line-broadening to study solution backbone dynamics and position-specific binding probabilities for amyloid β (Aβ) monomers in exchange with large (2–80 MDa) protofibrillar Aβ aggregates. Here we use ¹³C_methyl DEST and lifetime line-broadening to probe the interactions and dynamics of methyl-bearing side chains in the Aβ-protofibril-bound state. We show that all methyl groups of Aβ40 populate direct-contact bound states with a very fast effective transverse relaxation rate, indicative of side-chain-mediated direct binding to the protofibril surface. The data are consistent with position-specific enhancements of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ values in tethered states, providing further insights into the structural ensemble of the protofibril-bound state.

The formation of protein aggregates lies at the heart of numerous degenerative diseases, including Alzheimer’s disease in which the peptide amyloid β (Aβ) forms fibril plaques.^[1] Because smaller aggregates of Aβ, including oligomers and protofibrils, have been implicated as the primary toxic species, they are critical targets for structural characterization.^[2] Their large size, transient nature, and high level of structural disorder and heterogeneity, however, preclude characterization by standard structural techniques. Several new methods have been developed in an attempt to reveal structural details of these complexes.^[3] We recently reported a new solution NMR technique, dark-state exchange saturation transfer at backbone nitrogen positions (¹⁵N-DEST), to probe, with residue-level resolution, dynamic information (¹⁵N-R₂) and kinetic partitioning within the protofibril-bound state of Aβ.^{[4] 15}N-DEST and ¹⁵N lifetime line-broadening exploit the binding/unbinding of monomeric Aβ peptides to/from large protofibrillar aggregates and the vastly different transverse relaxation rates (¹⁵N-R₂) between monomers and protofibril-bound species to imprint atomic resolution information for the NMR-invisible (“dark”) protofibril-bound state onto the easily observed resonances of the free monomer. Unlike Aβ fibrils that are stabilized by β-sheets spanning ten residues in length in both the central and C-terminal hydrophobic regions,^[5] transient contacts with Aβ protofibrils are concentrated only in stretches of four to five hydrophobic residues. Since ¹⁵N-DEST reports only on the backbone, ¹³C-based experiments are required to probe side-chain contacts within the protofibril-bound states and to ascertain whether the transient interactions on the surface of the protofibrils are directly mediated by side chains or involve edge-on β-sheet contacts with effectively disordered side chains. Here we use ¹³C_methyl DEST and lifetime broadening to refine the atomic-level picture of the interactions between monomeric and protofibrillar Aβ40. We show that methyl groups in hydrophobic side chains directly mediate interactions on the protofibril surface and are sensitive reporters of dynamics in the bound-state ensemble.

U-[¹³C/¹⁵N]-labeled Aβ40 (rPeptide, Bogart, Georgia, USA) was prepared from NaOH treated stocks as described previously to remove preformed aggregates and seeds.^[4,6] NMR samples of 50 and 240 μM U-[¹³C/¹⁵N] Aβ40 were prepared in 50 mM HEPES buffer, pH 6.8 in 90% H₂O/10% D₂O from solutions stirred and filtered with Chelex 100 (Sigma) to remove trace metals. Samples were kept at 4 to 10°C and all experiments were conducted at 10°C. Assignment of the resolved methyl groups of monomeric Aβ40 was achieved using standard 3D heteronuclear correlation experiments (see the Supporting Information (SI), Figure S1) to correlate ¹H and ¹³C methyl resonances with the previously assigned backbone ¹H_N/¹⁵N and ¹H_α/¹³C_α resonances.^[4,7] Under these conditions, unstirred samples containing 240 μM Aβ40 spontaneously form large (2–80 MDa) polymorphic protofibrils, whereas dilute control (50 μM) samples remain >95% monomeric.^[4] After one week equilibration of the unstirred 240 μM Aβ40 sample, approximately 100 μM Aβ40 remains monomeric, whereas the balance of the peptide is sequestered in the protofibrillar aggregates visible by transmission electron and atomic force microscopy, and observed by analytical ultracentrifugation and dynamic light scattering.^{[ 4]} After the initial week-long period of protofibril assembly, the monomeric population remains in pseudo-equilibrium with the protofibril form for about one month, during which time the monomeric peptide reversibly binds to and unbinds from the protofibrils, resulting in near-zero net further assembly.^[4]

The protofibril-bound species exchanging with the remaining monomeric pool is invisible to standard solution NMR methods, because it is sparsely populated (≈5% of the remaining monomer) and characterized by very rapid transverse relaxation rates (¹⁵N-R₂ up to ca. 19000 s⁻¹) due to its very large size.^[4] Interactions, however, between monomeric and protofibrillar Aβ are evident by measuring backbone-lifetime-broadening effects (ΔR₂), independent of field and nucleus, from the difference in ¹⁵N-R₂ or ¹H_N-R₂ values in the presence (high concentration) and absence (low-concentration control) of protofibrils.^[6b] The apparent first-order association rate constant ( $k_{\text{on}}^{\text{app}}$ for monomer-binding to the protofibril, given by the maximum value of ¹⁵N-ΔR₂,^[6b] is 2.5±0.3 s⁻¹ for U-[¹³C/¹⁵N] Aβ40 at 240 μM (SI, Figure S2).

To compare the values measured at the backbone to those measured at the side-chain methyl positions, effective ¹³C_methyl-R₂ values were determined from ¹³C-R_1ρ (1666 Hz spin-lock field-centered at 20.7 ppm) and ¹³C-R₁ values using standard experiments and procedures (SI, Figures S3 and S4).^{[8, 9] 13}C_methyl lifetime line-broadening, (¹³C_methyl-ΔR₂), given by the difference in ¹³C_methyl-R₂ values between Aβ40 samples in the presence and absence of protofibril aggregates (Figure 1A), shows a position-specific pattern of interactions with the protofibrils. ¹³C_methyl-DEST measurements confirm the residue-specific nature of the interactions (Figure 1B). Briefly, the ¹³C_methyl-DEST pulse sequence (SI, Figure S5) resembles a 2D ¹H-¹³C HSQC-based ¹³C-R₁ experiment, in which ¹³C_z polarization is prepared; however, the variable delay used to measure R₁ is replaced by a fixed 480 ms ¹³C saturation pulse (radiofrequency field strengths of 763 and 1523 Hz) applied at a series of offsets from the ¹³C carrier frequency at 19 ppm (−23.7, −20, −16, −10 −8, −6, −4, −2, 0, 2, 4, 6, 8, 10, 16, 20, 23.7 26 kHz). Whereas the ¹³C_methyl-DEST pulse sequence is similar to the recently described ¹³C_methyl-CEST experiment,^[10] the DEST and CEST experiments are not the same and should not be confused with one another: CEST relies on differences in chemical shifts between slowly exchanging species with the same or approximately the same R₂ rates, whereas DEST is entirely reliant on the exchange between species with very large differences in R₂ rates and requires no differences in chemical shifts. Further, distinct chemical shifts for the minor species will not be observed in a CEST experiment when the R₂ values for the minor species are very large. Thus, the experimental considerations for and quantitative analysis of the DEST and CEST experiments are quite different.

Figure 1.

¹³C_methyl ΔR₂ and DEST data for Aβ40. A) Observed ¹³C_methyl-ΔR₂ (blue circles), the difference in effective R₂ of methyl group ¹³C spins in the presence (240 μM total Aβ) and absence (50 μM Aβ) of protofibrils, are well-matched by values calculated from best-fits to the equilibrium direct-contact/tethered state binding model (black diamonds; cf. Figure 2). Values for residues (Val and Leu) with two methyl groups are averaged. Error bars are 1 s.d. The amino acid sequence is displayed on top with hydrophobic residues colored in green. B) Examples of experimental (open circles) and best-fit (solid lines) ¹³C_methyl-DEST profiles at two radiofrequency field strengths, 763 (orange) and 1523 (blue) Hz, for methyl groups of alanine (A2 and A21), leucine (L17 and L34), and valine (V18 and V40). The best-fits obtained by nonlinear minimization and numerical integration of the McConnell equations for the direct-contact/tethered state model^[4] (cf. Figure 2) by optimization of a single global ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{contact}}$ value and residue-specific ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ values and partition coefficients (K₃) between direct-contact and tethered states, with the global pseudo-first order association rate constant $k_{\text{on}}^{\text{app}}$ set to ${}^{15}\mathrm{N}-\mathrm{\Delta }{R}_2^{max}$ (2.5 s⁻¹; SI Figure S2), and the global dissociation rate constant k_off (51 s⁻¹) taken from Ref. [4]. The experimental data were recorded at a spectrometer ¹H frequency of 600 MHz.

The data from ¹³C_methyl ΔR₂ and DEST experiments were fit simultaneously to the same kinetic model previously used for the corresponding ¹⁵N data^[4] to provide atomic-resolution details on the role of methyl-side-chain interactions in protofibril-bound states. Our previous ¹⁵N data, which probed the backbone of Aβ, showed that the ensemble of binding modes in the protofibril-bound state could not be described by a simple global two-state exchange model.^[4] The same is true for the current ¹³C_methyl data, for which the simple two-state exchange model predicts ¹³C_methyl ΔR₂ values and DEST profiles that are larger and narrower, respectively, than measured. However, a simple extension, incorporating slowly interconverting exchange between distinct bound states for each residue, described by a residue-specific partition coefficient K₃ for the population ratio of direct-contact to tethered states (Figure 2), is sufficient to reproduce the ¹⁵N-DEST and ¹⁵N-ΔR₂ data.^[4] This phenomenological extension of a two-state exchange model concisely captures the average behavior of the potentially very large ensemble of bound conformations using only two distinct residue-specific parameters, the partition coefficients K₃ and the $R_2^{\text{tethered}}$ values for the tethered states. Here, we show that without expanding the model, we capture the ¹³C_methyl experimental data by constraining the global values of $k_{\text{on}}^{\text{app}}$ to the value of ${}^{15}\mathrm{N}-\mathrm{\Delta }{R}_2^{max}$ measured on the 240 μM ¹³C/¹⁵N-labeled Aβ40 sample (SI, Figure S2), and k_off to that determined previously,^[4] and varying the residue-specific ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ relaxation rates and partition coefficients K₃, as well as a single global value of the transverse relaxation rate ( ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{contact}}$) when a methyl group is in direct contact with the protofibril surface. (Note that k_off is consistent with a range of values from about 20 to 200 s⁻¹ with statistically equivalent fits vis-à-vis residuals; hence we chose to fix k_off to the same value as determined previously,^[4] which should be independent of the Aβ40 concentration and facilitate the comparison between the current results and our previous work.)

Figure 2.

Interaction of monomeric Aβ on the surface of Aβ protofibrils. A) Schematic of overall exchange process. B) The direct-contact/tethered bound-state model, which incorporates tethered and direct-contact states for each residue when bound to the protofibril surface, superimposed on a global two-state exchange model.

A comparison of the experimental and computed ¹³C-methyl ΔR₂ (Figure 1A) and ¹³C_methyl-DEST profiles (Figure 1B) for the best-fit solution demonstrates that the model in Figure 2B describes the ensemble of interactions well. The ability to quantitatively match the observed ¹³C_methyl-DEST profiles and ¹³C_methyl-ΔR₂ values (Figure 1) with K₃ partition coefficients similar to those determined for the backbone (Figure 3A) demonstrates the ability of the model to capture the essential interactions present in transient monomer–protofibril binding. The high correlation between the population of direct contact states (i.e. K₃) obtained from the ¹³C_methyl and ¹⁵N data for all residues (Figure 4B), including the long, highly flexible methionine side chain, indicates that the entire side chain plays a role in mediating direct contacts with the protofibrils. The values of K₃ determined from the ¹³C_methyl data are systematically larger than those from the ¹⁵N data. Whether this truly reflects a different partitioning of bound and tethered states as viewed by the methyl groups, or whether these differences, which are less than a factor of 2 in an equilibrium constant, simply reflect sample variation, cannot be ascertained, because almost equally good fits to the data can be obtained using the K₃ values determined previously from the ¹⁵N data^[4] (SI, Figure S6), further emphasizing the role of the side chains in mediating direct contacts at the protofibril surface.

Figure 3.

Best-fit values of A) the partition coefficients K₃ (blue) and B) ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ (blue). Values for the corresponding K₃ values obtained previously from ¹⁵N data^[4] (black) and experimental values of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{monomer}}$ for the monomeric state (black) are also shown in panels (A) and (B), respectively. Values for residues (Val and Leu) with two methyl groups are averaged. For reference, ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{contact}}$, the best-fit value of the transverse relaxation rate, common to all methyl groups when in direct contact with the protofibril surface, is 1330±80 s⁻¹. Error bars are 1 s.d. The amino acid sequence is displayed on top with hydrophobic residues colored in green.

Figure 4.

Correlation between A) experimental side-chain ¹³C_methyl-ΔR₂ and backbone ¹⁵N-ΔR₂ values, B) best-fit direct contact/tethered partition coefficients K₃ obtained from ¹³C_methyl and backbone ¹⁵N data; and C) best-fit ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ and ${}^{15}\mathrm{N}-R_2^{\text{tethered}}$ values for Aβ40. Values for residues (Val and Leu) with two methyl groups are averaged. Color coding: black, Ala; red, Ile (C_γmethyl); green, Leu; blue, Met; and grey, Val. ¹⁵N values could not be determined for Ala2 owing to water exchange and are represented here by values for Glu3. Error bars are 1 s.d.

The global best-fit value of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{contact}}$ (at a ¹H frequency of 600 MHz) is 1330±80 s⁻¹. This value, which corresponds to a homogenous line width of 3 ppm, is consistent with direct methyl group interactions with the surface of large (2–80MDa) protofibril aggregates. Importantly, these direct interactions, though less frequent, are present even for residues such as Ala2 (Figure 1B), in the hydrophilic N-terminal region of Aβ.

The most important additional observations enabled by the ¹³C_methyl ΔR₂ and DEST experiments relate to the behavior of the side chains in the tethered states. Our previous ¹⁵N results suggested that the backbone R₂ in the tethered state varies as the average number of residues between the direct contact residue(s) and the residue of interest varies, with the highest ${}^{15}\mathrm{N}-R_2^{\text{tethered}}$ values (shortest tether) for residues 17–20. However, the role of side-chain atoms in tethered states could not be ascertained, and the relative contribution of a single side chain is obscured, because backbone motions are highly correlated along the peptide chain.^[4]

The good correlation between the experimental backbone ¹⁵N-ΔR₂ and side chain ¹³C_methyl-ΔR₂ values (correlation coefficient of 0.88) demonstrates the consistency between lifetime line-broadening at backbone and side-chain positions (Figure 4A). However, the ¹³C_methyl-ΔR₂ values are about half of those of the ¹⁵N-ΔR₂ values, as expected since ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{contact}}$ is only about 20-fold higher than k_off, such that ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{max}}$ does not reach the limiting value of $k_{\text{on}}^{\text{app}}$.

The validity of the model used to analyze the ¹³C-methyl data is further evidenced by the high degree of correlation between $R_2^{tethered}$ values for side-chain methyl groups and the ¹⁵N backbone of alanine residues (correlation coefficient = 0.99). The effective values of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ (Figure 3B, blue circles) for the three alanine residues, whose methyl groups at Cβ reflect the backbone motion, are much higher for Ala21 compared to Ala30 and Ala2 (Figure 3B), just as those at the corresponding ¹⁵N positions (Figure 4C, black circles, and solid black line).

The deviation of other methyl-bearing residues from the linear correlation between alanine ¹³C_methyl and ¹⁵N $R_2^{\text{tethered}}$ values (Figure 4C, solid line) shows the ability of the ¹³C-methyl experiments to provide additional information regarding the motion of particular side chains in the tethered state. Some residues, such as Leu34 and Val39, fall along the same correlation line as the alanine residues, indicating that their side chains behave much like the backbone. The remaining residues, however, fall below the alanine correlation with correspondingly smaller, but still well correlated, ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ values relative to the ${}^{15}\mathrm{N}-R_2^{\text{tethered}}$ values (Figure 4C, dashed line with correlation coefficient of 0.85 and a slope value half of that for the alanine correlation) suggesting that the average distance of these methyl groups to the nearest direct contact point is further away than for the backbone.

Within the central hydrophobic region, the values of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ for Leu17 and Ala21 are high, consistent with our previous observations for the corresponding backbone ${}^{15}\mathrm{N}-R_2^{\text{tethered}}$ values.^[4] A lower than expected ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ value, however, is observed for Val18 compared to the adjacent Leu17 despite one fewer dihedral angle degree of freedom separating the methyl groups from the backbone, suggesting exclusion of Val18 from transient-tethered-state interactions. Conversely, although the ${}^{15}\mathrm{N}-R_2^{\text{tethered}}$ values for Val18 are 50% higher than those for Val12, the ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ value for Val18 is about 30% lower, suggesting a greater degree of involvement of the Val12 side chain in transient intermolecular contacts in the tethered states than discernable from ¹⁵N-DEST alone. Finally, the value of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ for Met35 is significantly lower than that for neighboring residues, indicating that its long side chain retains considerable mobility when not directly in contact with the protofibril surface. (It should be noted that one would not expect to directly observe methyl cross-peaks for the tethered states, even in cases, in which the ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ values are low, as at the C-terminus, because the tethered state may be identical in chemical shift to the monomer state, and more importantly any potential cross-peak for a tethered state would be highly broadened in both ¹H and ¹³C dimensions by the off-rate of 51 s⁻¹ due to the highly skewed populations of free and bound Aβ40 exchanging on the surface of the protofibrils.)

The fits reported for the ¹³C_methyl DEST and ΔR₂ data in Figures 1, 3, and 4 were obtained with optimized values of K₃. However, as noted above, almost equally good fits can be obtained by constraining the values of K₃ to those obtained previously from analysis of the ¹⁵N data.^[4] The ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ values obtained in this manner are highly correlated to those obtained with optimized K₃ values (SI, Figure S6) and hence do not impact the conclusions drawn above. The absolute values of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{tethered}}$ as well as the value of ${{}^{13}\mathrm{C}}_{\text{methyl}}-R_2^{\text{contact}}$, are about 60–70% higher when K₃ is fixed. This is simply due to the fact that the values of K₃ and ¹³C_methyl-R₂ in the bound state are partially anticorrelated such that higher values of K₃ (when optimized) lead to lower values of ¹³C_methyl-R₂ in the bound state.

Our current ¹³C_methyl data support a model in which protofibril-bound structures are stabilized by a network of residues (based on the consistency of ¹⁵N and ¹³C_methyl DEST profiles), whereas methyl-bearing residues contribute to substructures in the tethered states. Further, the current work confirms the validity of the “model-free” representation of the very large ensemble of states that can be sampled on the surface of the protofibrils (Figure 2).

In summary, our analysis combining ¹³C_methyl DEST and ΔR₂ measurements brings a new, detailed view of the structural and dynamic properties of side-chain methyl groups in exchanging systems involving very high molecular weight (>1 MDa), transient complexes, and provides direct insight into interactions in the vicinity of a given methyl side chain that are inaccessible by other methods. Moreover, the rapid rotation of methyl groups resulting in a large reduction of $R_2^{\text{contact}}$ relative to the backbone, should permit ¹³C_methyl DEST and lifetime line-broadening to be applied to a much larger variety of systems involving both larger NMR-observed species (e.g. greater than 100 kDa) for which the observation of backbone amides would not be feasible, larger dark states (in excess of 80 MDa), and faster exchange kinetics down to the millisecond time scale. Further extension of DEST to all nonmethyl ¹³C atoms is also possible, but lower signal-to-noise and significant overlap precluded the complete characterization of other side-chain positions in Aβ40.