Probing the dissociation pathway of a kinetically labile transthyretin mutant

Abstract

Aggregation of transthyretin (TTR) is associated with devastating amyloid diseases. Amyloidosis begins with dissociation of the native homotetramer (a dimer of dimers) to form a monomeric intermediate that assembles into pathogenic aggregates. This process is accelerated in vitro at low pH, but the process by which TTR dissociates and reassembles at neutral pH remains poorly characterized due to the low population of intermediates. Here we use ¹⁹F-NMR and a highly sensitive trifluoromethyl probe to determine the relative populations of the species formed by dissociation of a destabilized variant, A25T. The A25T mutation perturbs both the strong dimer and weak dimer-dimer interfaces. A tetramer⇌dimer⇌monomer (TDM) equilibrium model is proposed to account for concentration- and temperature-dependent population changes. Thermodynamic and kinetic parameters and activation energetics for dissociation of the native A25T tetramer, as well as a destabilized alternative tetramer (T*) with a mispacked F87 side chain, were extracted by ¹⁹F-NMR van’t Hoff, lineshape analysis, saturation transfer and transition state theory. The ¹⁹F and methyl chemical shifts of probes close to the strong dimer interface are degenerate in the dimer and T* species, implicating interfacial perturbation as a common structural feature of these destabilized species. All-atom molecular dynamics simulations further suggest more frequent F87 ring flipping on the nanosecond timescale in the A25T dimer than in the native A25T tetramer. Our integrated approach offers quantitative insights into the energy landscape of the dissociation pathway of TTR at neutral pH.

Graphical Abstract:

Introduction

Transthyretin is a homotetramer, organized as a dimer of dimers, that functions as a transporter of thyroid hormone and the holoretinol-binding protein in blood plasma and cerebrospinal fluid (CSF). Transthyretin amyloidosis is caused by deposition of pathogenic TTR aggregates in tissues and organs including the heart and central nervous system.¹ The TTR aggregation pathway begins with dissociation of the native tetramer to form an aggregation-prone monomeric intermediate,² a process that is accelerated at mildly acidic pH.³ However, wild-type (WT) TTR, as well as most pathogenic TTR variants, remains predominately tetrameric at physiological concentrations in the blood^{4, 5} with a protomer concentration ranging from 7 to 21 μM.⁶ However, in the CSF, the TTR concentration range is lower, 0.16–1.6 μM,⁶ favoring dissociation of variants with destabilized tetramers such as A81T, which exhibits a monomer population of 17% at pH 7.0 and 1.25 μM concentration.⁵

Several mechanisms for TTR tetramer dissociation and reassembly have been proposed, involving dimeric and/or trimeric intermediates,^7–9 and dissociation pathways via an intermediate dimerized across the strong dimer interface have been simulated by molecular dynamics (MD).^{10, 11} Previous experimental studies of dissociation pathways have been carried out at lowered pH or in the presence of detergent or denaturant to destabilize TTR,^{12, 13} or by using crosslinking to shift equilibria,⁸ or by relying on surface-induced gas phase dissociation to drive tetramer dissociation.¹⁴ Direct experimental observation of weakly populated dissociative intermediates in solutions that are free of detergent or denaturant has not been possible for WT TTR. Elucidation of the mechanism of TTR dissociation and the partitioning between reassembly and aggregation at physiological concentrations and pH is of central importance for understanding potential pathways that result in TTR amyloidosis and disease.

The tetramer formed by the A25T variant is kinetically unstable and is one of the most rapidly dissociating TTR variants reported to date.¹⁵ Compared to the slow dissociation of WT TTR, on the timescale of days, the dissociation half-time of A25T is on the timescale of minutes.¹⁵ The A25T mutation is located at the weak dimer-dimer interface (Figure 1A), where mutations are known to destabilize the native tetramer.¹⁶ A25T aggregates aggressively in the CSF and is associated with TTR amyloidosis in the central nervous system.^{15, 17} In the present work, we used ¹⁹F-NMR to probe the dissociation equilibria of the A25T-C10S-S85C mutant, labeled with 3-bromo-1,1,1-trifluoroacetone (BTFA; the labeled construct is denoted as A25T^F), over a range of temperatures at neutral pH. The simplest model that accounts for the ¹⁹F-NMR data is a tetramer⇌dimer⇌monomer (TDM) equilibrium. Global van’t Hoff analysis offers key insights into the thermodynamics of the dissociation pathway of A25T at neutral pH. The analysis revealed the presence of a minor, destabilized form of the tetramer, previously denoted as T*,¹⁸ that is more prone to dissociation than the native tetramer. Combining ¹⁹F-NMR magnetization transfer, lineshape analysis and transition state theory, we determined rate constants and activation energy barriers of the two-step TDM dissociation pathway. Population analyses using ¹⁹F- and ¹³C-HMQC NMR reveal that the chemical shifts of three probes for the T* and D states are degenerate, indicating similar probe environments in these states. All-atom MD simulations further suggest that both the strong and weak dimer interfaces are perturbed in the A25T dimer. Our work provides new quantitative data on the thermodynamics and kinetics of the TTR dissociation pathway at neutral pH.

Figure 1.

Comparison of the X-ray structures of WT (gray, PDB 5CN3¹⁹) and A25T (red, PDB 8T5X, this work). A. Cartoon representation of the strong dimer present in the asymmetric unit of the A25T structure. Labels show motifs involved in the strong dimer interface (formed by the F and H strands in one protomer and the F’ and H’ strands in the neighboring protomer) and the location of the AB loop that forms a part of the weak dimer-dimer interface. Both interfaces are indicated by dashed lines. The site of attachment of the ¹⁹F probe at residue 85 is indicated by a red sphere. B. Expanded view from the box in panel A showing residue 25 in the AB loop and the critical D18-Y78 hydrogen bond (orange dashed line) that stabilizes interactions between the AB loop and the nearby EF helix. The short D strand and the EF helix are also labeled.

Results

X-ray crystal structure of A25T

The A25T mutant was crystallized in space group P2₁2₁2 and diffraction data were collected to a resolution of 1.63 Å. The strong dimer formed by two neighboring protomers was observed in the asymmetric unit (Table S1). Overall, the A25T structure is very similar to the WT TTR X-ray structure (Cα RMSD = 0.3 Å to PDB 5CN3,¹⁹ Figure 1A). The AB loop (residues 18 to 27, including the mutation site) and the nearby EF helix (residues 75 to 82) have a similar conformation in the A25T and the WT structures (Figure 1). There is no change in the crucial hydrogen bond between the Y78 hydroxyl and D18 carboxylic oxygen (Figure 1B) that stabilizes the TTR protomer: the hydrogen bond distance of 2.6Å is the same in both structures. As shown by over 200 TTR X-ray structures²⁰ and a previously published A25T X-ray structure (PDB 3TFB,¹⁷ Figure S1), X-ray structures of TTR seldom if ever reveal pronounced conformational perturbations caused by individual mutations.

Solution ¹⁵N-NMR spectroscopy of A25T

Solution NMR using ¹⁵N and ¹³C-labeled protein can give a better idea of the structural and dynamic changes that occur as a consequence of mutations. Several amide and Cα cross peaks are broadened or missing from 2D and 3D triple resonance spectra of A25T. Overlaid ¹H, ¹⁵N-TROSY spectra of A25T and WT TTR are shown in Figure 2A. Residues in the AB loop, the short D strand (residues 54 to 56) and the EF helix, as well as neighboring residues, are broadened or shifted substantially in the A25T spectrum relative to that of WT TTR, and V93 and V94 in the strong dimer interface are also broadened (Figure 2A). A number of resonances were completely missing from 3D triple resonance spectra (trHNCA and trHN(CO)CA) of A25T. Residues lacking discernible amide or Cα connectivities are mapped onto the A25T structure in Figure 2B. These residues are mostly located in the AB loop and the DAGH β-sheet (Figure 2B). The broadening also propagates to parts of the EF loop (residues 83 to 89) and the F strand (residues 91 to 97), which forms part of the strong dimer interface (Figure 2C). Both the weak dimer-dimer and, to a lesser degree, the strong dimer interfaces are perturbed by the A25T mutation. For backbone amides that can be assigned, chemical shift perturbations (CSP) are observed for residues extending from the middle of the E strand into the EF helix (Figure 2D). Interestingly, Y78 has similar chemical shifts in A25T and WT TTR, consistent with the similarity of the Y78 (side chain)-D18 hydrogen bond distances and orientations in the X-ray structures (Figure 1B).

Figure 2.

Perturbation of the TTR NMR spectrum by the A25T mutation. A. Superposition of 800 MHz ¹H-¹⁵N-TROSY spectra of A25T (red) and WT TTR (gray) in NMR buffer (10 mM potassium phosphate, 100 mM KCl, pH 7.0) at 298 K. Amide cross peaks of representative residues in the AB loop, B strand, D strand, E strand, and EF helix that are broadened or shifted substantially in the A25T spectrum relative to WT are labeled in black and cyan, respectively. Broadened cross peaks of V93 and V94 (located in the strong dimer interface) are also labeled. B. Residues lacking visible amide NH or Cα peaks in 3D trHNCA and trHN(CO)CA spectra are mapped as red spheres on two views of the A25T protomer structure with β-strands labeled. C. Mapping of residues with broadened cross peaks onto the symmetry-imposed tetramer structure of A25T; the majority of these residues are located at the weak dimer-dimer interface (W, indicated by the dashed lines) and strong (S) dimer interface. D. Chemical shift perturbation CSP = √[(Δδ_H)² + (Δδ_N/5)²] of A25T versus WT. The dashed line denotes mean +1 standard deviation (s.d.). Secondary structures are indicated above the panel. Red boxes on the x-axis denote residues with broadened resonances and the blue star indicates the site of the A25T mutation.

Dissociation equilibria of A25T TTR monitored by ¹⁹F-NMR

We then asked whether the perturbed dimer interfaces promote dissociation of the native A25T tetramer. To probe dissociation of A25T at low μM concentration, we introduced a sensitive trifluoromethyl probe at position 85, in the construct A25T-C10S-S85C-BTFA (denoted as A25T^F). This ¹⁹F probe senses distinct oligomerization states of TTR.²¹ At 310 K, pH 7.0 and a concentration of 1.25 μM TTR (corresponding to physiological levels in the CSF), three peaks were observed in the ¹⁹F spectrum of A25T^F (Figure 3A). The intense peak at −83.92 ppm arises from the native tetramer (T). The shoulder at −84.01 ppm was initially assigned to T*, a mispacked tetramer state with reduced stability.¹⁸ Later analysis (see below) indicated that this shoulder also includes the signal from a dimeric A25T intermediate. The third peak at −84.20 ppm corresponds to the monomer (M). The same ¹⁹F chemical shift difference (0.28 ppm) was observed between the T and M peaks of the corresponding WT construct C10S-S85C-BTFA (TTR^F) at pH 4.4.²¹ The M peak is also observed at neutral pH and 1.25 μM concentration in ¹⁹F spectra of the BTFA-labeled K80E^F, K80D^F, and A81T^F mutants, which destabilize the EF helix, but not in TTR^F spectra recorded under the same conditions.⁵ Incubation of A25T^F with the drug tafamidis results in loss of both the T*/D shoulder and the monomer peak and an increase in the intensity of the T peak (Figure 3B), showing that binding of tafamidis in the central hydrophobic cavity shifts the dissociation equilibrium to stabilize the native, but not the mispacked T*, tetramer.²²

Figure 3.

¹⁹F-NMR spectra of A25T^F under physiologically relevant conditions. (A) ¹⁹F spectrum of 1.25 μM A25T^F at 310 K and pH 7.0 with T, T*/D and M peaks labeled. (B) ¹⁹F spectra of 30 μM A25T^F at 281 K and pH 7.0 without (blue) and with 50 μM tafamidis (red). The incubation time was 60 mins. (C) The logarithms of T and T*/D peak intensities in ¹⁹F-DOSY measurements of 230 μM A25T^F at 277 K are a linear function of 10 z-gradient strengths. Solid lines denote linear fits. (D) The ratio of the translational diffusion constants for the T*/D peak relative to the T peak at 277 and 298 K for 230 μM A25T^F. The results show that a species smaller than tetramer is present under the shoulder and that it is more highly populated at the lower temperature. The error is propagated from fitting uncertainty from 50 bootstrapped data sets. (E) Concentration dependence of the A25T^F spectrum at 277 K. 3-state Lorentzian fits are shown as solid lines and data as open circles. For clarity, only selected concentrations are shown, with full data set fits in Figure S2. (F) Fits of relative populations using the TDM model, where the ¹⁹F chemical shifts of T* and D are degenerate and both resonances are under the shoulder peak. The relative populations of the various species (circular data points) were determined by fitting the ¹⁹F spectrum at each concentration using Lorentzian functions. Fits to the TDM dissociation model are shown in solid lines. Alternative models with worse or unconstrained fitting statistics are shown in Figure S3. Inset: relative populations of T* and D calculated from the TDM model (dashed lines) showing that both contribute to the observed shoulder peak (red data points and solid line model fit).

To identify the TTR species that contribute to the shoulder peak, we performed ¹⁹F diffusion-ordered spectroscopy (DOSY) experiments for A25T^F at 277 K, where the intensity of the shoulder peak is enhanced compared to higher temperatures (Figure 3B). The translational diffusion constant of the shoulder peak relative to that of the T peak is 1.13 ± 0.03 (Figure 3C), which is reduced to 1.02 ± 0.03 at 298 K (Figure 3D). This indicates that a species that diffuses faster than tetramer is more highly populated under the shoulder peak at 277 than at 298 K. To further probe the species under the shoulder peak, we recorded a series of spectra at decreasing A25T^F concentration at 277 K and determined the relative populations of the T, T*/D, and M species using 3-state Lorentzian fits (Figures 3E and S2). At the lowest protein concentrations, the intensities of the monomer and shoulder peaks are greatly increased relative to the tetramer peak. Different models for the A25T dissociation pathway were tested by fitting the concentration dependence of the T, T*/D, and M populations (Figure S3). The simplest dissociation model that can fit the concentration dependence is a tetramer⇌dimer⇌monomer (TDM) model (Scheme 1) where both the mispacked tetramer T* and the dimer D contribute to the shoulder peak (Figure 3F and inset).

Scheme 1.

The tetramer⇌dimer⇌monomer (TDM) dissociation model for A25T at pH 7.0. Transition states are labeled as $I_1^{\ne }$ to $I_3^{\ne }$.

Alternative dissociation models that omit D or T* do not fit the high concentration data (Figure S3A–B). Models invoking a trimeric species to explain the shoulder peak fail to fit the low concentration data (Figure S3C–E and G). A sequential dissociation model in which both T and T* first dissociate to form a trimer (R) that dissociates further to D and subsequently to M (TRDM model, Figure S3H), fits the concentration-dependent data nearly as well as the simpler TDM model (Figure S3F). However, the TRDM model is unable to constrain any of its four K_d parameters (Figure S3H). Fitting the concentration dependent ¹⁹F spectra of A25T^F at 277K to the TDM model yields constrained K_d1, K_d2 and K_d3 values for the three dissociation processes (T ⇌ 2D, D ⇌ 2M and T* ⇌ 2D) in Scheme 1 (Table S2). It is of note that T* and D are 6.6-fold and 8-fold more dissociation-prone than the native T species, respectively. Based on this model, the dimer is primarily responsible for the enhanced height of the shoulder peak at low concentrations (Figure 3F inset). The similar ¹⁹F chemical shifts of T* and D suggest similar environments for the CF₃ probe, which is located at S85C in the EF loop near the strong dimer interface (Figure 1A). This further suggests that the dimeric intermediate D probably corresponds to the strong dimer, with protomers packed through the F-F’ and H-H’ interface (Figure 1A).⁷ If dimerization occurred through the weak dimer-dimer interface, the environment around the S85C-BTFA probe would be more similar to M than to T*. This is clearly inconsistent with the distinct ¹⁹F chemical shifts observed for M and D in the A25T^F spectra. A comparison of TTR^F and A25T^F by SDS PAGE analysis also shows that a band corresponding to the strong dimer, which is resistant to SDS unless the sample is boiled, is maintained in A25T^F (Figure S4).

Thermodynamics of the A25T TDM dissociation pathway

We recorded ¹⁹F NMR spectra of A25T^F at 6, 30 and 230 μM concentration, at 8 temperatures ranging from 277 to 298 K, and quantified the relative populations of the T, T*/D, and M peaks using 3-state Lorentzian fits (Figures 4A–D and S5). Differences in ¹⁹F T₁ values, which are around 0.3 s for all species over this temperature range (Table S3), are too small to affect the population estimates given the 1-s recycle delay used. A global fit of the temperature titration data to the van’t Hoff equation gives values for K_d1, K_d2 and K_d3 in the TDM model in Scheme 1 that are statistically similar to those obtained by fitting of the concentration-dependent spectra at 277 K alone (first two rows in Table S2). Additional concentration-dependent data were acquired at 298 K and all five data sets were fitted globally (Figure 4E–F). The fitted thermodynamic parameters (one pair of ΔH and ΔS for each of the three equilibria associated with K_d1, K_d2 and K_d3) are listed in Table 1 and the corresponding K_d values at 277 and 298 K are in Table S2.

Figure 4.

Global van’t Hoff analysis of temperature and concentration dependent data for A25T^F. A. ¹⁹F spectra of 6 μM A25T^F at 277 and 298 K, aligned on the T chemical shift at 277 K. Data are shown in circles and 3-state Lorentzian fits are in solid lines. For clarity, data from only two temperatures are shown. Fitting results are shown in Table 1 and the full data set fits in Figure S5. B–D. Relative population of T, T*/D, and M as a function of temperature at A25T^F concentrations of 6 μM (B), 30 μM (C), and 230 μM (D). E–F. Changes in relative population of T, T*/D, and M with changes in concentration at 277 (E) and 298 K (F). These data sets were used in the van’t Hoff analysis of the TDM dissociation model for A25T^F. Data points are shown as circles and the TDM model fits as solid lines.

Table 1.

Fitted ΔH and ΔS parameters by van’t Hoff analysis for the TDM dissociation model of A25T^F

Equilibriuma	T ⇌ 2D	T* ⇌ 2D	D ⇌ 2M
ΔH (kcal/mol)	ΔH1 = −42 ± 4	ΔH3 = −38 ± 5	ΔH2 = 4 ± 2
ΔS (cal/mol/K)	ΔS1 = −179 ± 15	ΔS3 = −160 ± 16	ΔS2 = −10 ± 8
ΔG (kcal/mol) at 277 K	ΔG1 = 8.0 ± 0.1	ΔG3 = 6.8 ± 0.1	ΔG2 = 6.9 ± 0.1
ΔG (kcal/mol) at 298 K	ΔG1 = 11.8 ± 0.3	ΔG3 = 10.2 ± 0.3	ΔG2 = 7.1 ± 0.2

^aΔH and ΔS for the forward reaction in each equilibrium (from left to right).

Knowledge of ΔH and ΔS for each dissociation process allows computation of ΔG as a function of temperature (Figure 5A). The ΔH and ΔS for the formation of T and T* from D are both positive (Table 1), indicating that tetramerization is enthalpically unfavorable and driven mostly by a favorable entropy change. This implies a key role for hydrophobic interactions in driving tetramerization, likely associated with the formation of the central aromatic ligand-binding channel that occurs in T or T* but not in the strong dimer D.^{5, 22} The large increase in entropy upon tetramer formation presumably reflects release of hydration water molecules from the weak dimer-dimer interface as it becomes buried in the tetramer (Figure 1A). Lowering the temperature from 298 to 277 K weakens K_d1 for dissociation of T from 2 nM to 0.5 μM (Table S2), providing a thermodynamic rationale for the observed cold denaturation of the native TTR tetramer.²³ Between 277 and 298 K, K_d3 for dissociation of T* to form dimer is consistently weaker than the corresponding dissociation constant for T (K_d1) as T* is 1.2–1.6 kcal/mol less stable than T over this temperature range (Figure 5A–B).

Figure 5.

Free energies and populations derived from the van’t Hoff analysis. A. Temperature-dependent changes in free energy for T and T* dissociation to D (blue and gold, respectively) and for dissociation of D to M (red). B. Dissociation constants derived from the equilibria in A. Errors are 1 s.d. from 100 bootstrapped datasets. C–F. 2D-contour plots showing the relative protomer population in the T (C), T* (D), D (E) and M (F) states of A25T^F at pH 7.0 as a function of temperature and concentration.

By contrast, K_d2 for dimer dissociation remains relatively constant at around 4–6 μM between 277 and 298 K (Table S2), a much weaker temperature dependence than for K_d1 or K_d3 (Figure 5B). This temperature insensitivity arises because the absolute value of ΔH₂ (for dissociation of dimer to form monomer) is an order of magnitude smaller than ΔH₁ or ΔH₃. The formation of the strong dimer (D) is mainly driven by enthalpy (ΔH₂ = −4 kcal/mol), consistent with the extensive hydrogen-bonding between the H and the H’ strands (Figure 1A). The 2-dimensional contour plots of the relative populations of the T, T*, D and M species as a function of temperature and concentration summarize their thermodynamic stabilities (Figure 5C–F). High concentrations favor both T and T* tetramers at the cost of M, and D behaves like a dissociative intermediate. Due to its lower free energy, T is more favored at higher temperature than T*, especially above 10 μM. Within the physiological concentration range (0.1–20 μM), the population of dimer is less than 5% at 298 K, explaining the similar diffusion constants of the T*/D shoulder and T peak (Figure 3D) and the lack of a peak corresponding to dimeric TTR in the size-exclusion chromatogram (Figure S6).

Temperature dependent population shifts

The methyl ¹H,¹³C-HMQC spectrum of A25T (Figure 6A) shows tetramer cross peaks for the δ1 methyls of the five Ile residues, which were assigned by mutagenesis (Figure S7A–D). The I26δ1, I68δ1 and I84δ1 cross peaks are more intense than those from I73δ1 and I107δ1. Two additional minor cross peaks are observed at similar positions to those of the I68δ1 and I84δ1 cross peaks in spectra of the monomeric mutant F87E-V122I (Figure S7E), and likely arise from the corresponding Ile residues in a small population of A25T monomer. The population of the two minor peaks increases at low temperature, where tetramer dissociation is enhanced. However, the population of M predicted from the ¹⁹F-NMR van’t Hoff analysis (Table 1) underestimates the relative intensities of the minor peaks in the ¹³C-HMQC spectrum (gray line in Figure 6B). Rather, both the intensity and volume of the minor peaks and their temperature dependence can be accounted for only if it is assumed that they arise from all three of the intermediates T*, D, and M (green line in Figure 6B). We infer that the chemical shifts of the I68 and I84 δ1 methyls of T* and D are degenerate, as also observed for the ¹⁹F chemical shift in spectra of A25T^F (Figure 3F).

Figure 6.

Characterization of A25T using ¹³C-HMQC spectroscopy. A. Ile δ1 methyl region in the ¹³C-HMQC spectrum of 216 μM A25T at 277 K and pH 7.0 with assignments indicated. The tetramer cross peaks were assigned by mutagenesis and the minor peaks were assigned to I68δ1 and I84δ1of the monomer by comparison with the spectrum of monomeric V122I-F87E (Figure S7). B. The M species alone (gray line) cannot account for the observed populations of the minor peaks of I84δ1 and I68δ1 (blue, red and gold markers, by intensity or volume as labeled). The population of M and the sum of the T*, D and M populations (green line) were determined from the thermodynamic parameters in Table 1. The population analysis suggests that the methyl chemical shifts of the minor peaks of I84δ1 and I68δ1 are degenerate for T*, D and M.

Kinetics and activation energetics from ¹⁹F-NMR lineshape analysis

Rate constants for the TDM model (Scheme 1) were extracted by lineshape analysis of the ¹⁹F spectra of A25T^F at 277 and 298 K and at multiple concentrations (Figure 4E–F). Lineshapes calculated by numerically solving the Bloch-McConnell equation for chemical exchange^{24, 25} based on the 4-state kinetics model in Scheme 1 were fitted to experimental lineshapes as described in the Supporting Information. The following considerations were made to reduce the number of floating parameters. Based on the K_d1, K_d2 and K_d3 constraints (Scheme 1), we only need to fit one of the two rate constants for each equilibrium step. Since ¹⁹F chemical shifts for T, T*/D, and M are independent of concentration (Figure S8), peak center positions were constrained to further reduce the number of fitting parameters. The rate constants of k₃ and k₄ were determined by ¹⁹F saturation transfer and constrained by K_d2 (Figure S9 and green entries in Table 2) so that the only floating kinetic parameters are k₁ and k₅. At 277 K, T, T*/D, and M have distinct ¹⁹F R₂ values (Table S4) and the linewidth of each peak was fitted independently. However, at 298 K, the T and T*/D peaks have comparable ¹⁹F R₂ and their linewidths were constrained to have the same value while the linewidth of M was fitted separately. The fitted spectra are shown in Figure 7. The fitted and measured kinetic parameters are summarized in Table 2, and the extracted linewidths are listed in Table 3. Note that the 2D⇀T rate constant (~3 × 10⁹ M⁻¹ s⁻¹) approaches the diffusion limit at 298 K (~7 × 10⁹ M⁻¹ s^-1, estimated by the Smoluchowski coagulation equation²⁶ at 298 K).

Table 2.

Kinetic parameters for the A25T^F TDM dissociation model determined by lineshape analysis or measured by ¹⁹F saturation transfer (also see labeled kinetic scheme in Figure S15).

Equilibrium	T ⇌ 2D	T* ⇌ 2D	D ⇌ 2M
Forward ratea constant at 277 K	k1 7.5 ± 0.9 s−1	k5 4.4 ± 1.1 s−1	k3 1.9 ± 0.4 s−1
Reverse ratea constant at 277 K	k2 7.7 ± 1.2 μM−1s−1	k6 0.53 ± 0.17 μM−1s−1	k4 0.26 ± 0.05 μM−1s−1
Forward ratea constant at 298 K	k1 13 ± 1 s−1	k5 50 ± 8 s−1	k3 15 ± 6 s−1
Reverse ratea constant at 298 K	k2 ~3 ×103 μM−1s−1	k6 ~7 ×102 μM−1s−1	k4 1.2 ± 0.2 μM−1s−1

^aRate constants constrained by measured values and K_d2 are in green. Only k₁ and k₅ are fitting parameters; k₂ and k₆ are constrained by K_d1 and K_d3 (Table S2).

Figure 7.

¹⁹F-NMR lineshape analysis for A25T. A–B. Concentration-dependent ¹⁹F spectra at 277 (A) and 298 K (B) are shown in blue circles and lineshape fits are in solid red lines. The intensity of each spectrum was normalized to ensure the same total peak area for all spectra. Concentrations are indicated at the top of each panel.

Table 3.

Fitted linewidths from Bloch-McConnell lineshape analysis for the TDM dissociation model of A25T^F.

19F Peaks	Temperature (K)	Linewidth (Hz)a
T	277	29 ± 1
T*/D	277	25 ± 1
M	277	38 ± 1
T, T*/D	298	19 ± 1
M	298	13 ± 1

^aLinewidths do not include any broadening caused by inter-species exchange.

With both rate constants determined for the T* ⇌ 2D equilibrium, the Bloch-McConnell lineshape analysis predicts the ratio of diffusion constants of the T*/D peak relative to the T peak as 1.11 ± 0.02 at 277 K (Figure S10A and see SI Methods for numerical details), in excellent agreement with the experimental value of 1.13 ± 0.03 (Figure 3D). By contrast, the population averaged (i.e., assuming no or very slow exchange between T* and D) diffusion constant ratio for the shoulder peak relative to the T peak is 1.07, which is significantly less than the measured value. This result shows that exchange between T* and D on the second time scale contributes to the measured diffusion constant over a diffusion delay of 0.1 s.^{27, 28} The simulated diffusion constant ratio at 298 K (1.02 ± 0.03, Figure S10B) also matches the experimental value of 1.02 ± 0.05 (Figure 3D); at this temperature, the dimer is less populated (<5%) and thus the diffusion constant ratio is much closer to 1 since both T and T* are tetrameric. In addition, the kinetic parameters (Table 2) and the linewidth parameters from the lineshape analysis (Table 3) recapitulate ¹⁹F spectra collected at 1 μM at both 277 and 298 K (Figure S11), which were not included in the lineshape analysis. Collectively, these comparisons validate the fitting of the ¹⁹F-NMR lineshapes.

The Bloch-McConnell lineshape analysis based on the TDM model yields a linewidth for the M peak that is 1.3-fold larger than that of T at 277 K (Table 3), despite M having a lower molecular weight. By contrast, the ¹⁹F R₂ relaxation rate of the M state of A25T^F at 277 K, measured with a CPMG pulsing frequency of 4000 s⁻¹, is comparable to that for an F87A^F monomer or the monomeric F87E^F (Table S4). Thus, the linewidth of the A25T^F M species at 277 K in Figure 7A is increased by an exchange process that is quenched by 180^o refocusing pulses applied at a high frequency. Exchange between the ground state and an excited state of A25T^F M with different ¹⁹F chemical shifts could lead to such broadening. This was confirmed by ¹⁹F-CPMG dispersion experiments at 277 K, which revealed exchange at a rate of ~780 s⁻¹ between the ground and excited states of M. The exchange contributions to relaxation are quenched at a pulsing frequency of 4000 s⁻¹ as the fitted ¹⁹F $R_2^0=34\pm 3{\text{s}}^{-1}$ of A25T^F M (Figure S12) is fully consistent with the measured ¹⁹F R₂ of the A25T^F monomer (32 ± 5 s⁻¹, Table S4). By comparison, the A25T^F T*/D and T peaks do not exhibit relaxation dispersion at 277 K (Figure S12) so that the Bloch-McConnell modeled linewidth ratio of the T*/D shoulder peak relative to that of the T peak (0.86 ± 0.04) is within error of the independently measured ¹⁹F R₂ ratio for A25T^F (0.81 ± 0.03, Table S4). The linewidth ratio and ¹⁹F R₂ ratio are both smaller than 1, indicating a substantial contribution from the smaller dimer that contributes to the intensity of the T*/D shoulder (Figure 3F). At 298 K, the measured ¹⁹F R₂ ratio of the T*/D shoulder and T peaks of A25T^F is 1.0 ± 0.1 (Table S4), consistent with a greatly reduced dimer population at 298 K (Figure 5E) and with the similar translational diffusion coefficients of the T* and T peaks at 298 K (Figure 3D).

The determination of rate constants in the TDM model at two temperatures, 277 and 298 K, allows us to construct the free energy landscape for all three dissociation processes (Figure 8). Note that the activation energy for all the unimolecular reactions with first order rate constants k₁, k₃, and k₅ are concentration independent. By contrast, the activation free energy for bi-molecular reactions with second order rate constants k₂, k₄ and k₆ are concentration-dependent. At low concentrations, the activation free energy for these reactions is increased, reflecting reduced oligomerization rates and a more stable monomer relative to other species. Activation enthalpies and entropies determined using transition state theory and the rate constants at two temperatures are listed in Table S5. Of note, the formation of T and T* from D has the most positive activation enthalpy as these two reactions are associated with largest increase in rate constants from 277 to 298 K.

Figure 8.

Activation energy diagram for the TDM dissociation pathway of A25T determined from ¹⁹F-NMR lineshape analysis. A–D. Energy diagrams for A25T at 232 μM at 277 (A) and 298 K (C) and 2 μM at 277 (B) and 298 K (D). The free energy of the dimer, the common intermediate, is set to 0 kcal/mol as reference. For clarity, the dissociation pathway from T* via $I_3^{\ne }$ to D is shown in red.

Molecular dynamics simulations of the A25T dimer

The degeneracy of the ¹⁹F S85C-BTFA and I84δ1 methyl chemical shifts of D and T* indicate similar environments of the EF loop region in these two species. The F87 side chain is mispacked in the T* species.¹⁸ To obtain insights into the structure and dynamics of the A25T dimer we turned to molecular dynamics (MD) simulations. Two types of all-atom simulations were run: isothermal-isobaric ensemble (npt) and canonical ensemble (nvt). Multiple independent simulations show that flipping of the F87 ring about the χ2 dihedral angle occurs on the nanosecond time scale in the A25T dimer (Figures 9A and S13), but is much less frequent in the A25T tetramer (Figure 9B). Corresponding probability contour plots for the dimer and tetramer are shown in Figure S14A–B.

Figure 9.

All-atom isothermal-isobaric ensemble (npt) molecular dynamics simulations of the A25T dimer. A–B. Trajectories showing changes in F87 χ2 dihedral angle in A25T dimer (A) and tetramer (B). The F87 χ2 angles in the A25T X-ray structure (black) and a low pH structure of WT TTR (green, PDB 3D7P³⁰) are plotted for reference.

Discussion

The rate limiting step in TTR aggregation is dissociation of the tetramer to form an aggregation-prone monomeric intermediate.^{3, 4, 31, 32} Dissociation of WT TTR at neutral pH is on the order of days and previous in vitro studies of the TTR aggregation pathway were therefore performed at acidic pH to accelerate dissociation of the tetramer and promote local unfolding and aggregation of the resulting monomer.^{2, 3, 21, 31, 32} Although tetramer dissociation at neutral pH has been observed for WT TTR at sub-μM concentration³¹ and at μM concentration for the highly destabilized variants L55P,^{33, 34} V14D-V16E,³⁵ S112I^{36, 37} and D18G,³⁸ the thermodynamic equilibrium constants of individual steps, the molecular nature of dissociative intermediates, and the kinetics of the dissociation pathway are unknown. Understanding the kinetics and mechanism of TTR tetramer dissociation and reassembly at neutral pH is critical for an understanding of transthyretin amyloidosis under physiological conditions.

In this work, we used solution NMR based assays to quantitatively map the kinetics and energetics of the dissociation and reassembly pathway of the highly destabilized A25T variant using a tetramer⇌dimer⇌monomer (TDM) model (Scheme 1). The full kinetic scheme and the rate constants at 277 and 298 K determined by ¹⁹F lineshape analysis are summarized in Figure S15. The A25T mutation is located in the AB loop, which forms the weak dimer-dimer interface (Figure 1A–B). The enhanced propensity of A25T to dissociate, which occurs on a time scale of minutes compared to days for the WT TTR tetramer, is related to perturbations propagated from the site of mutation to both the weak and strong dimer interfaces, where amide resonances of affected residues are either broadened or shifted. In particular, the amide cross peaks of L17, V20, L110 and V121, which are located in the subunit interfaces and play a key role in mediating the dissociation of tetramer to form dimers in MD simulations,¹⁰ are broadened in the TROSY spectrum of A25T (Figure 2D), suggesting enhanced conformational dynamics (Figure 2A–C).

By coupling a BTFA probe at residue S85C, we are able to resolve peaks corresponding to T, T*/D, and M in ¹⁹F-NMR spectra of the A25T variant under physiological conditions (Figure 3A). The key to our analysis is a global van’t Hoff fit of multiple concentration and temperature titration data sets (Figure 4B–F) and lineshape analysis of the ¹⁹F spectra of A25T^F over a range of concentrations (Figure 7A–B). The initial intermediate formed on the dissociation pathway of A25T^F is a dimer, inclusion of which best fits the concentration series (Figure 3F and Table S2) and quantitatively explains the ¹⁹F-DOSY results (Figure 3D) as it diffuses faster than tetramer and exchanges with the T* state on a time scale of seconds (Table 2). The degenerate ¹⁹F chemical shifts of the T* and D peaks, which contribute to the shoulder on the T peak, suggest that the dissociative intermediate is the strong dimer. Tetramer dissociation via the strong dimer intermediate has been previously inferred and studied using Cys-crosslinked dimer constructs,^{7, 39} by surface induced dissociation in gas phase by native mass spectrometry,¹⁴ and by MD-based free energy calculations.^{10, 11} Our observation of dimer-mediated dissociation of A25T^F under equilibrium conditions at neutral pH is fully consistent with these previously reported results. Numerical simulations have suggested that the dimer-mediated formation of tetramer (dimer of dimers) can reduce the amounts of protein trapped as assembly intermediates and is employed by all homo-tetrameric proteins surveyed.⁴⁰ In this regard, the homotetrameric TTR is no exception.

The concentration dependence of the A25T ¹⁹F spectra can also be fit by a sequential tetramer-trimer-dimer-monomer dissociation (TRDM) model (Figure S3H). Although the TRDM model fits the data nearly as nearly well as the simpler TDM model, it does not yield constrained parameters for any of the four K_d values. Even when the K_d values for the common D ⇌ 2M and T ⇌ T* equilibria are fixed to the corresponding values from the TDM model, the remaining two K_d values for dissociation of T via trimer R to D (K_d5 and K_d6) remain unconstrained (Figure S16A–D). However, the product of K_d5 and K_d6 is constrained, which indicates that the calculated K_d1 for T ⇌ 2D is constrained and is within error the same as K_d1 directly fitted from the TDM model (Figure S16E–F). This analysis shows that invoking a trimeric intermediate in the tetramer dissociation pathway is not justified by the concentration series data. With that said, a very small population of trimer (~1%, Figure S16B) cannot be completely excluded. The favoring of a dimeric intermediate, formed by disruption of the weak dimer-dimer interface, over a trimeric intermediate in the dissociation pathway is consistent with their interface surface areas: the weak dimer-dimer interface has the smallest area compared to the strong dimer or the trimer/monomer interfaces,⁴¹ so that energetically the strong dimer is the most likely major dissociative intermediate, which is consistent with how most homotetrameric proteins are assembled.⁴⁰ We note that a monomer-dimer-trimer-tetramer (MDRT) reassembly model has been previously invoked to account for refolding kinetics data at high monomer concentration.⁸ At low monomer concentration, the trimeric intermediate is unstable and dissociates to reform unstable dimeric intermediate and monomer.⁸ Under equilibrium conditions, as in the present studies of A25T, the concentration of monomer is low and formation of a trimeric reassembly intermediate would thus be disfavored. Indeed, for WT TTR in the physiological concentration range (0.1–20 μM) at neutral pH, the estimated monomer concentration is in the nM range, based on the ~10²⁴ M³ tetramerization K_d⁴² and reassembly via a dimeric intermediate would be favored. This is consistent with a previous report that showed reassembly of the WT and V30M tetramers via an MDT model at low protein concentrations.⁴³

The equilibrium population of the dimer is less than 5% at 298 K when the A25T concentration is above 0.1 μM (Figure 5E). This small population highlights the need to combine a highly sensitive ¹⁹F trifluoromethyl chemical probe and NMR cryoprobes to study the dimeric species experimentally. The similar environments experienced by ¹⁹F and I84δ1 probes in D and T* offer structural insights into the otherwise elusive dimer species. Based on MD simulations, the F87 aromatic ring flips more frequently on the nanosecond time scale in the A25T dimer than in the tetramer (Figure 9A–B). These ring flipping events are likely coupled with altered EF loop conformations, resulting in a slightly upfield shifted ¹⁹F chemical shift of the shoulder peak from the main T peak (Figure 3A–C). This structural perturbation may be related to a cavity near the F87-A120 region¹⁸ that could contribute to protein breathing motions⁴⁴ and facilitate associated aromatic ring flipping events.⁴⁵ The mispacked T* state is observed under physiological conditions (Figure 3A) and all the species can be converted to T with tafamidis within 1 hour (Figure 3B), consistent with interconversion rate constants on the second time scale (Table 2). Interestingly, the energy difference between the T and T* states of A25T^F is 1.6 kcal/mol at 298 K (Figure 8C–D), which is identical to that of the parental TTR^F at the same temperature (1.6 kcal/mol or a T* population of ~7%, as previously described¹⁸). Thus, the A25T mutation does not change the energetic difference between the correctly packed (T) and mispacked (T*) forms of the TTR tetramer.

The quantitative insights from our analysis allow us to place our results in the context of existing TTR literature. First, the ΔH and ΔS of the forward T ⇌ 4M equilibrium of A25T are more negative than reported values for two EF helix mutants, K80E and K80D⁵ (Table S6). However, ΔG at 298 K is similar for all three destabilized mutants, ~7–8 kcal/mol smaller than for WT TTR (Table S6). Also, the ΔG at 278 K is 21.9 ± 2.4 kcal/mol for A25T, within error compared to that of 23.6 ± 0.6 kcal/mol previously reported.¹⁷ Second, the rate constants for the D ⇌ 2M transition have been reported for WT TTR refolded upon dilution from denatured monomers as 9–590 s⁻¹ (k₃) and 0.3–20 μM⁻¹s⁻¹ for (k₄) at 298 K⁸. The corresponding rate constants for A25T^F, k₃ = 15 s⁻¹ and k₄ = 1.2 μM⁻¹s⁻¹ at 298 K, fall within these ranges. Third, the tetramer dissociation rate constant, measured by urea-induced unfolding, for T* is 0.26 h⁻¹ and for T is 0.066 h⁻¹ in TTR^F at 298 K¹⁸. For A25T^F at 298 K, the rates of dimer formation by dissociation of T* and T (k₅ and k₁) are 50 and 13 s⁻¹, respectively, much faster than the corresponding dissociation rate constants for T and T* in TTR^F and consistent with previous reports that A25T is a highly kinetically destabilized variant.^{4, 15} Interestingly, the k₅/k₁ ratio for the T* and T states of A25T^F is 3.8 ± 0.7, closely mirroring the 3.9-fold increase in dissociation rate constant of T* relative to T in TTR^F.¹⁸ This similarity suggests that the A25T mutation likely affects the dissociation processes of T and T* similarly.

In summary, we demonstrate an integrated method based on ¹⁹F-NMR that allows simultaneous determination of all thermodynamic and kinetic parameters for dissociation of the A25T variant of TTR at neutral pH. Our approach provides free energies as well as activation energy barriers for all equilibrium steps, including ones that involve weakly populated species. We anticipate that our method will be broadly applicable to characterize dissociation equilibria of oligomeric proteins under a wide range of conditions.