Mispacking of the F87 sidechain drives aggregation‐promoting conformational fluctuations in the subunit interfaces of the transthyretin tetramer

Abstract

Aberrant formation and deposition of human transthyretin (TTR) aggregates causes transthyretin amyloidosis. To initialize aggregation, transthyretin tetramers must first dissociate into monomers that partially unfold to promote entry into the aggregation pathway. The native TTR tetramer (T) is stabilized by docking of the F87 sidechain into an interfacial cavity enclosed by several hydrophobic residues including A120. We have previously shown that an alternative tetramer (T*) with mispacked F87 sidechains is more prone to dissociation and aggregation than the native T state. However, the molecular basis for the reduced stability in T* remains unclear. Here we report characterization of the A120L mutant, where steric hindrance is introduced into the F87 binding site. The x‐ray structure of A120L shows that the F87 sidechain is displaced from its docking site across the subunit interface. In A120S, a naturally occurring pathogenic mutant that is less aggregation‐prone than A120L, the F87 sidechain is correctly docked, as in the native TTR tetramer. Nevertheless, ¹⁹F‐NMR aggregation assays show an elevated population of a monomeric aggregation intermediate in A120S relative to a control containing the native A120, due to accelerated tetramer dissociation and slowed monomer tetramerization. The mispacking of the F87 sidechain is associated with enhanced exchange dynamics for interfacial residues. At 298 K, the T* populations of various naturally occurring mutants fall between 4% and 7% (ΔG ~ 1.5–1.9 kcal/mol), consistent with the free energy change expected for undocking and solvent exposure of one of the four F87 sidechains in the tetramer (ΔG ~ 1.6 kcal/mol). Our data provide a molecular‐level picture of the likely universal F87 sidechain mispacking in tetrameric TTR that promotes interfacial conformational dynamics and increases aggregation propensity.

1. INTRODUCTION

Transthyretin (TTR) is an abundant tetrameric protein that functions as a transporter for thyroxine in blood. Aggregation of TTR leads to TTR amyloid disease, including wild‐type (WT) TTR cardiac amyloidosis and familial cardiomyopathy and polyneuropathy associated with hereditary mutations (Johnson et al., 2012; Sekijima et al., ²⁰⁰⁵). The TTR aggregation pathway begins with dissociation of the TTR tetramer to form monomers (Lai et al., 1996) or unstable dimers that rapidly dissociate (Sun et al., 2024). WT TTR cardiac amyloidosis affects up to 25% of the population with age greater than 80 years (Westermark et al., 1990). Pathogenic mutations that weaken the TTR tetramer are associated with earlier ages of onset of TTR amyloidosis (Benson and Kincaid, 2007; Sekijima et al., ²⁰⁰⁵). Thus, it is important to understand how the TTR tetramer becomes destabilized to facilitate dissociation.

The TTR tetramer is assembled as a dimer of dimers. The subunits of the strong dimer are held together by hydrogen bonds between the F/F′ and H/H′ β‐strands and an extended β‐sheet, while a pair of dimers associates through interactions between the AB and GH loops that form the weak dimer interface (Figure 1) (Blake et al., 1978). We have previously shown using ¹⁹F‐NMR that an alternatively‐packed TTR tetramer (T*) is more prone to dissociation and aggregation than the native TTR tetramer (designated the T state) (Sun et al., ^2018b). In the native TTR tetramer, the F87 sidechain packs into a hydrophobic pocket in the neighboring subunit across the strong dimer interface (Figure 1a), contributing to the stability of the tetramer. The pocket is lined by several hydrophobic residues, including A120 (Figure 1b). By contrast, the F87 sidechain is mispacked in the T* state, which in native human TTR has a population of ~7%. The low population of the T* state makes structural characterization challenging. To circumvent this problem, we have shown that the T* population can be increased by mixing WT‐TTR with a designed A120L mutant; steric clash arising from the bulky Leu sidechain forces the F87 sidechain to adopt an alternative conformation. In this work, we use A120L as a model for the T* state and solve its x‐ray structure to highlight how the F87 sidechain is mispacked in the T* state. We also study a naturally occurring A120S mutant that is more stable than A120L but less stable than WT‐TTR. Using both ¹⁹F and amide NMR probes, we show that the conformational dynamics are enhanced in several interfacial residues in the T* state, which is likely linked with its higher aggregation propensity.

FIGURE 1.

Structure of the strong dimer and location of dimer interfaces. The F87 sidechain (magenta) packs into a hydrophobic pocket in the neighboring subunit across the strong dimer interface in a WT TTR structure (PDB: 5CN3) shown in a global view (a) and a closeup view (b). WT TTR, wild‐type transthyretin.

2. MATERIALS AND METHODS

2.1. Protein expression and purification

WT human TTR and the A120S and A120L TTR variants were expressed and purified as previously reported (Sun et al., 2017). Purification of Cys‐derived oligomers based on C10S‐S85C or C10S‐S100C and labeling for ¹⁹F‐NMR with 3‐bromo‐1,1,1‐trifluoroacetone (BTFA) was performed as previously described (Sun et al., ^2018a). Unless noted, all experiments were performed in 10 mM potassium phosphate, and 100 mM KCl at pH 7.0 (NMR buffer). Uniformly ¹⁵N labeled WT TTR was expressed in M9 medium containing 1 g/L (¹⁵NH₄)₂SO₄ and 3 g/L ¹²C‐glucose.

2.2. Size exclusion chromatography

A Superdex 75 gel filtration column was pre‐equilibrated in NMR buffer. A120L or WT‐TTR was then loaded at a flow rate of 0.7 mL/min.

2.3. ¹⁹F‐NMR spectroscopy

¹⁹F‐NMR spectra were recorded at 298 K using Bruker Avance 600 or Bruker Avance 700 MHz spectrometers as described previously (Sun et al., ^2018a). The fitting of K _d for a tetramer ⇌ monomer (T ⇌ 4M) equilibrium for A120S was performed as previously described (Ferguson et al., 2021).

2.4. ¹⁵N‐filtered diffusion‐ordered (DOSY) spectroscopy

¹⁵N‐filtered DOSY experiments for ¹⁵N‐labeled A120L (100 μM) were performed using a Bruker Avance 900 MHz spectrometer in a transverse optimized pulse field gradient experiment (Horst et al., 2011). A diffusion delay of 200 ms was used for a total of 11 gradient strengths from 6.8 to 23.8 G/cm (Sun et al., 2017).

2.5. Heterotetramer population modeling

To estimate populations of WT‐A120L heterotetramers at given WT and A120L total concentrations ($c_{\mathrm{WT}}$ and $c_{\mathrm{A}120\mathrm{L}}$, respectively), a minimal set of five K _d affinity constants for the equilibrium T ⇌ 4M are defined as follows:

$$K_{\mathrm{d}1}=\frac{{\left[M_{\mathrm{WT}}\right]}^4}{\left[T_{\mathrm{WT}}\right]},$$ (1)

$$K_{\mathrm{d}2}=\frac{{\left[M_{\mathrm{A}120\mathrm{L}}\right]}^4}{\left[T_{\mathrm{A}120\mathrm{L}}\right]},$$ (2)

$$K_{\mathrm{d}3}=\frac{{\left[M_{\mathrm{WT}}\right]}^3\left[M_{\mathrm{A}120\mathrm{L}}\right]}{\left[T_{{\mathrm{WT}}_3\mathrm{A}120{\mathrm{L}}_1}\right]},$$ (3)

$$K_{\mathrm{d}4}=\frac{{\left[M_{\mathrm{WT}}\right]}^2{\left[M_{\mathrm{A}120\mathrm{L}}\right]}^2}{\left[T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}\right]},$$ (4)

$$K_{\mathrm{d}5}=\frac{\left[M_{\mathrm{WT}}\right]{\left[M_{\mathrm{A}120\mathrm{L}}\right]}^3}{\left[T_{{\mathrm{WT}}_1\mathrm{A}120{\mathrm{L}}_3}\right]}.$$ (5)

Each K _d is associated with a free energy change $\mathrm{\Delta }G=-\mathit{RT}\ln \left(K_{\mathrm{d}}\right)$, where R is the Boltzmann constant, and T is temperature (298 K). $K_{\mathrm{d}1}=9\times {10}^{-25}{\mathrm{M}}^3$ (corresponding to ${\mathrm{\Delta }G}_1=32.8\text{kcal}/\mathrm{mol}$) for WT TTR at 298 K (Hurshman Babbes et al., 2008). To estimate $K_{\mathrm{d}2}$, we make use of the measured translational diffusion constant of A120L (6.3 × 10⁻⁷ cm²/s at 100 μM) which is weighted by the tetramer and monomer populations. Exchange between M and T is on the hour timescale at pH 7.0, much slower than the diffusion delay of 200 ms used in our ¹⁵N‐filtered diffusion‐ordered experiments (Sun et al., ^2018b) so that the translational diffusion constant of A120L is a simple population‐weighted average with no contribution from T/M exchange. Using translational diffusion constants of tetrameric WT TTR (5.1 × 10⁻⁷ cm²/s) and a monomeric F87E variant (7.6 × 10⁻⁷ cm²/s) (Sun et al., ^2018b), we determined the expected protomer concentrations of A120L monomer and tetramer and obtained $K_{\mathrm{d}2}=4.1\times {10}^{-13}{\mathrm{M}}^3$ using Equation (2). This corresponds to ${\mathrm{\Delta }G}_2=16.9\text{kcal}/\mathrm{mol}$ for A120L at 298 K. Using $K_{\mathrm{d}2}$ and population averaging, we estimated the translational diffusion coefficient for A120L at 800 μM to be (5.3 ± 0.2) × 10⁻⁷ cm²/s.

The following two constraints are based on mass conservation:

$$\begin{array}{l}4\left[T_{\mathrm{WT}}\right]+\left[M_{\mathrm{WT}}\right]+3\left[T_{{\mathrm{WT}}_3\mathrm{A}120{\mathrm{L}}_1}\right]+2\left[T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}\right]\\ +\left[T_{{\mathrm{WT}}_1\mathrm{A}120{\mathrm{L}}_3}\right]\\ =c_{\mathrm{WT}},\end{array}$$ (6)

$$\begin{array}{l}4\left[T_{\mathrm{A}120\mathrm{L}}\right]+\left[M_{\mathrm{A}120\mathrm{L}}\right]+\left[T_{{\mathrm{WT}}_3\mathrm{A}120{\mathrm{L}}_1}\right]+2\left[T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}\right]\\ +3\left[T_{{\mathrm{WT}}_1\mathrm{A}120{\mathrm{L}}_3}\right]\\ =c_{\mathrm{A}120\mathrm{L}}.\end{array}$$ (7)

We fit ¹⁹F‐NMR measured T* populations ($f_{{\mathrm{T}}^{*}}$) of 10 μM C10S‐S85C‐BTFA (TTR^F) as a function of [A120L] to extract $K_{\mathrm{d}3}$, $K_{\mathrm{d}4}$, and $K_{\mathrm{d}5}$ or ${\mathrm{\Delta }G}_3$, ${\mathrm{\Delta }G}_4$, and ${\mathrm{\Delta }G}_5$. $f_{{\mathrm{T}}^{*}}$ was experimentally measured as a normalized ratio, $f_{{\mathrm{T}}^{*}}=P_{{\mathrm{T}}^{*}}/\left(P_{{\mathrm{T}}^{*}}+P_{\mathrm{T}}\right)$, where P represents the ¹⁹F‐NMR peak areas of TTR^F, a model to report on T*% in WT TTR (Sun et al., ^2018b). In the case of mixed WT and A120L, $f_{{\mathrm{T}}^{*}}$ is weighted by populations of tetramers with at least one WT protomer:

$$f_{{\mathrm{T}}^{*}}=\frac{\left(4\alpha \left[T_{\mathrm{WT}}\right]+3\beta \left[T_{{\mathrm{WT}}_3\mathrm{A}120{\mathrm{L}}_1}\right]+2\gamma \left[T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}\right]+\epsilon \left[T_{{\mathrm{WT}}_1\mathrm{A}120{\mathrm{L}}_3}\right]\right)}{c_{\mathrm{WT}}},$$ (8)

where coefficients α, β, γ, and ε are the relative T* populations for respective heterotetrameric species with α = 0.06 (basal T* population of TTR^F used in these measurements) (Sun et al., ^2018b), $\beta =\frac{1}{3}\times 1.06$, and $\epsilon =1$. There are two types of 2:2 heterotetramer $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$: (1) a cis‐heterotetramer with the same protomers in either of the strong dimers (WT:WT || A120L:A120L, where “:” and “II” denote the strong and weak dimer interfaces, T* population = 0.06) and (2) a trans‐heterotetramer with the heterodimer WT:A120L in both strong dimers (i.e., WT:A120L || WT:A120L, T* population = 1). Treating the two types of heterotetramer with two floating K _d parameters led to the two individual K _d values for cis‐ and trans‐heterotetramers within fitting uncertainties. Therefore, we used one K _d4 to characterize both types of 2:2 heterotetramer, present in equal populations, so that the averaged T* population coefficient $\gamma$ is $\left(0.06+1\right)/2$. A previously reported MATLAB routine to fit ${\mathrm{\Delta }G}_3$, ${\mathrm{\Delta }G}_4$, and ${\mathrm{\Delta }G}_5$ constrained by ${\mathrm{\Delta }G}_1$ and ${\mathrm{\Delta }G}_2$ (Ferguson et al., 2021) was modified to use the MATLAB function fsolve to numerically solve Equations (6) and (7) which depend on $K_{\mathrm{d}3}$, $K_{\mathrm{d}4}$, and $K_{\mathrm{d}5}$. The differences in $f_{{\mathrm{T}}^{*}}$ between experimental data and modeled values calculated using Equation (8) for a given set of ${\mathrm{\Delta }G}_3$, ${\mathrm{\Delta }G}_4$, and ${\mathrm{\Delta }G}_5$ were minimized using fminsearch.

2.6. Real‐time NMR mixing experiments

¹⁵N labeled WT TTR and the unlabeled A120L variant were mixed at a 3:1 molar ratio (WT:A120L = 600 μM:200 μM) or a 1:1 molar ratio (WT:A120L = 200 μM:200 μM) in NMR buffer with 10% D₂O added for the lock signal. At these concentrations, A120L is a mixture of tetramer and monomer (Sun et al., ^2018b). Standard [¹H,¹⁵N]‐heteronuclear single quantum coherence spectra (HSQC) were acquired using Bruker Avance 600 or 700 MHz spectrometers as previously described (Sun et al., 2022). In the 3:1 mixing experiment, for each delay point in the indirect ¹⁵N dimension, 12 scans were collected, giving a total acquisition time of 60 min per 2D spectrum. A total of 72 2D spectra were recorded over 3 days. In the 1:1 mixing experiment, 32 scans were collected for each ¹⁵N delay and a total of 45 spectra were recorded for 5 days. The backbone amide assignments were transferred from those of WT human TTR (BioMagResBank code 27514) (Leach et al., 2018). The peak intensities of all non‐overlapped peaks were extracted by using CcpNmr (Vranken et al., 2005) and normalized to that of the C‐terminus E127. For peaks with pronounced intensity decays, a global fit to a single‐exponential function was performed using the MATLAB function nlinfit to extract plateau intensities. The error in the fitted rate constant was estimated as one standard deviation with 68% confidence level using nlpredci. All NMR data processing and analysis was carried out on the NMRBox platform (Maciejewski et al., 2017).

2.7. A120L and A120S x‐ray crystallography

Crystallization screens were set up using the sitting drop vapor diffusion method with the A120L and A120S mutants at a concentration of 9–10 mg/mL in NMR buffer. The crystallization condition for the A120L trial was 0.1 M sodium cacodylate (pH 6.5), 1 M sodium chloride, 10% (v/v) glycerol, and 30% (v/v) PEG 600 at 293 K. The cryoprotectant was 30% (v/v) PEG. The crystallization condition for the A120S trial was 0.1 M MES at pH 6.0, 10% (v/v) glycerol, 5% (w/v) PEG 1000, and 30% (v/v) PEG 600 at 293 K. The data were processed using HKL‐2000 (Otwinowski and Minor, 1997). Molecular replacement was performed using Phaser (McCoy et al., 2013) with the TTR structure (PDB ID: 2ROX) as a search model. Models were refined using phenix.refine (Liebschner et al., 2019), refmac (Murshudov et al., 2011), and Coot (Emsley et al., 2010). The coordinates have been deposited in the Protein Data Bank with accession codes 8W2W (A120L) and 8W1N (A120S).

2.8. Molecular dynamics (MD) simulations

The x‐ray structures of WT human TTR (PDB: 5CN3) and A120L (PDB: 8W2W) were used as a starting model for canonical ensemble MD simulations with explicit solvent using the AMBER ff14SB force field (Maier et al., 2015) and AMBER 16 software (Case et al., 2017) as previously described (Sun et al., 2024). The x‐ray crystallographic symmetry was used to create a tetrameric TTR model as an initial conformation using PyMOL (2.5.0) as described previously (Sun et al., 2017) and one 400‐ns MD simulation for each tetramer was run using previously described protocols (Sun et al., 2024). The mass‐weighted root‐mean‐square‐fluctuations (RMSF) per residue (Hunenberger et al., 1995) for backbone heavy atoms were extracted using CPPTRAJ (Roe and Cheatham 3rd, 2013), averaged over all four chains of the tetramer, and analyzed using MATLAB. The surface area of the F87 sidechain was estimated using VADAR (Willard et al., 2003) based on the WT TTR structure (PDB: 5CN3).

3. RESULTS

3.1. A120S is more stable than A120L but less stable than WT‐TTR

To measure the relative populations of monomer (M) and tetramer (T), we constructed variants of the C10S‐S85C‐BTFA (TTR^F) construct, which has distinct ¹⁹F chemical shifts for the monomeric and tetrameric species as the S85C‐BTFA label is close to the strong dimer interface (Sun et al., ^2018a). We introduced both the A120S and A120L mutations into TTR^F separately and denote the resulting constructs as A120S^F and A120L^F, respectively. Figure 2a shows that A120S^F forms a mixture of 88% tetramer (T) and 12% monomer (M) at 10 μM concentration. Fitting the populations of T and M from a series of A120S^F spectra over a range of concentrations (Figure 2b), we determined the K _d for the T ⇌ 4M equilibrium to be 6.5 × 10⁻¹⁹ M³ (Figure 2c), comparable to the K _d = 6.5 × 10⁻¹⁹ M³ for K80D^F (Ferguson et al., 2021) and much weaker than the K _d = 9 × 10⁻²⁵ M³ of WT human TTR (Hurshman Babbes et al., 2008). Because A120L^F aggregates above 10 μM, we were unable to directly determine its T/M K _d by ¹⁹F‐NMR. However, based on the population‐weighted ¹⁵N‐filtered translational diffusion constant of A120L at 100 μM (Figure S1A) (Sun et al., ^2018b), the K _d was estimated to be 4.1 × 10⁻¹³ M³ (see Section 2), 6 orders of magnitudes weaker than A120S. Based on this weak A120L K _d, the monomer population is predicted to be greater than 99% when the total concentration of A120L is below 10 μM, consistent with the observation that A120L^F is fully monomeric at 8 μM as shown by ¹⁹F‐NMR (Figure 2a) as well as by gel‐filtration analysis at 1 μM (Figure S1B,C). At 800 μM, the predicted translational diffusion constant of A120L is (5.3 ± 0.2) × 10⁻⁷ cm²/s, the same within error as the experimental value of (5.0 ± 0.1) × 10⁻⁷ cm²/s (Sun et al., ^2018b).

FIGURE 2.

(a) ¹⁹F‐NMR spectra of TTR^F‐based mutants at 298 K and pH 7.0 showing the ¹⁹F chemical shifts of the native tetramer (T), the alternative tetramer (T*), and the monomer (M). The concentration of A120L^F was 8 μM and the other constructs were at 10 μM. (b) ¹⁹F‐NMR concentration titration for A120S^F at 298 K and pH 7.0. Spectra were normalized by total peak areas. (c) The fitted apparent K _d = 6.5 × 10⁻¹⁹ M³ for the T ⇌ 4M equilibrium was determined by fitting the relative population of T and M from (b).

As indicated by the tetramer/monomer K _d, A120S is thermodynamically less stable than WT‐TTR. We next measured the acid‐mediated aggregation kinetics of A120S^F at pH 4.4 (Figure 3a,b) and analyzed the data using a three‐state T ⇌ M ⇌ Aggregates (A) model (Sun et al., ^2018a). The tetramer dissociation rate constants are greater for A120S^F than in the control TTR^F at 298 and 310 K (Table 1). Notably, the reverse tetramerization rate constant of A120S^F is seven‐fold slower than TTR^F at 298 K. Accordingly, we observed a higher level of the monomeric aggregation intermediate (M) in A120S^F than in TTR^F (Figure 3c). We have previously shown (Sun et al., ^2018a) that mutation of the F87 sidechain to Ala also enhances the kinetics of tetramer dissociation, slows reassembly, and increases the steady population of the aggregation intermediate M (Figure 2c, Table 1). Taken together, our data indicate that disruption of the hydrophobic interaction between the sidechains of F87 and A120, by mutation of either residue, destabilizes the tetramer and increases the aggregation propensity (Table 1).

FIGURE 3.

(a and b) ¹⁹F‐NMR aggregation assays at pH 4.4 for A120S^F at 310 K (a) and 298 K (b). The black lines denote the fits from using a three‐state aggregation kinetics model T ⇌ M ⇌ A where T, M, and A stand for tetramer, monomeric intermediate, and aggregates, respectively. (c) The steady‐state monomer concentration (M) of TTR^F, A120S^F, and F87A^F at 310 and 298 K. The data for TTR^F and F87A^F were replotted from Sun et al. (2018a) with uncertainties estimated as 1 standard deviation from 50 bootstrapped datasets.

TABLE 1.

¹⁹F‐NMR aggregation kinetics based on the T ⇌ M ⇌ A model.

Mutant	Temp. (K)	k 1 (h−1)	k −1 (h−1)	k 2 (h−1)	k −2 (h−1)	γ 2 (h−1)
TTRF	310	0.10 ± 0.01	0.75 ± 0.12	0.73 ± 0.03	0.03 ± 0.01	0.06 ± 0.01
TTRF	298	0.13 ± 0.02	0.63 ± 0.10	0.06 ± 0.01	0.01 ± 0.01	0.02 ± 0.01
A120SF	310	0.22 ± 0.03	0.54 ± 0.14	0.93 ± 0.09	0.07 ± 0.01	0.16 ± 0.01
A120SF	298	0.24 ± 0.04	0.06 ± 0.02	0.05 ± 0.01	0.02 ± 0.01	0.05 ± 0.01
F87AF	310	1.60 ± 0.16	0.18 ± 0.04	0.56 ± 0.01	0.09 ± 0.01	0.57 ± 0.02
F87AF	298	0.46 ± 0.02	0.04 ± 0.01	0.06 ± 0.01	0.03 ± 0.01	0.08 ± 0.01

Note: The results for TTR^F and F87A^F were taken from Sun et al. (2018a) for comparison with A120S^F. The slow relaxation rate constant γ ₂ describes the aggregation propensity. Errors are fitting uncertainties calculated as 1 standard deviation from 50 bootstrapped datasets.

3.2. The F87 sidechain is mispacked in the x‐ray structure of A120L but not A120S

To understand the molecular basis of the distinct F87‐A120 interactions in A120S and A120L, we next solved their x‐ray structures. The A120L and A120S constructs crystallized in space groups I222 and P2₁2₁2, with one and two TTR protomers in the asymmetric unit, respectively (Figure 4a). Despite being monomeric below 10 μM, A120L formed a tetramer at the high concentration (~500 μM) used for crystallization. The diffraction data were refined to 2.07 and 1.60 Å, respectively (Table S1). The mean Cα RMSD values for both x‐ray structures relative to a WT‐TTR structure (PDB: 5CN3) are less than 0.5 Å. This comparison shows that the mutations do not significantly perturb the tertiary structure of TTR tetramers (Figure S2A,B). The B‐factor profile of A120S closely resembles that of the WT human TTR (Hörnberg et al., 2000) whereas A120L displays elevated B factors throughout the EF region (Figure 4b).

FIGURE 4.

(a) X‐ray crystal structures of A120S chain A (green) and A120L (magenta) shown as cartoons, with chain B of A120S shown as a surface model in a strong dimer. Inset: Close‐up view of the structure, zoomed from the black box in (a), showing the distinct orientations of the F87 sidechain in A120L and A120S. (b) Comparison of mass‐weighted crystallographic B factors for backbone heavy atoms averaged for all subunits in one unit cell for A120L and A120S x‐ray structures, in comparison with the WT TTR (black, PDB: 5CN3). B factors were mass‐weighted per residue and averaged over all four chains of the tetramer. The overall elevation of B‐factors for A120L reflects the lower resolution of the structure (Table S1). (c) Sidechains of F87, F95, and L120 modeled into the electron density in the A120L dimer structure imposed by crystallographic symmetry. (d) F87 with the electron density map for the A120L monomer structure. The displayed maps are |2F _o − F _c| with σ = 1.00. TTR, transthyretin.

The difference in the orientation of the F87 sidechain between the WT and A120S TTR structures is subtle (Figure S2C) while in the A120L structure the F87 sidechain is displaced from its binding pocket and occupies a solvent‐exposed position (Figure 4c, Figure S2D). In the A120L structure, the sidechain of L120 occupies the hydrophobic pocket that accommodates the F87 sidechain in the WT and A120S structures (Figure S2C,D). The distinct orientations of the F87 sidechains in A120L and A120S are consistent with the greater volume of the Leu sidechain compared to Ser; occupancy of the binding pocket by the bulky Leu sidechain precludes docking of the F87 aromatic ring and explains the lower thermodynamic stability of A120L^F compared to A120S^F (Figure 2a). The electron density for the F87 sidechain in the A120L structure does not completely encompass the phenyl ring (Figure 4d). In addition, there is no electron density for the Cβ atom, suggesting that multiple conformations are sampled by the F87 sidechain of the A120L mutant, a model of the T* state. Consistent with this, enhanced nanosecond timescale fluctuations of the F87 sidechain about the χ1 and χ2 dihedral angles were observed in all‐atom MD simulations of the A120L tetramer, but not in simulations of WT‐TTR (Figure 5a–d).

FIGURE 5.

Nanosecond dynamics of the F87 χ1 (a and b) and χ2 (c and d) dihedral angles in canonical ensemble MD simulations for WT (a, c) and A120L tetramers (b, d). MD, molecular dynamics.

3.3. The T* population is conserved and consistent with displacement of a single F87 sidechain within the tetramer

We used ¹⁹F‐NMR to measure the population of the T* state in the pathogenic variants A25T, V30M, L55P, A81T, G83R, A97G, A120S, V122I, and the protective variant T119M (Table 2 and Figure S3). All mutations were introduced into the TTR^F construct and coupled to BTFA as described previously (Sun et al., ^2018a). The coupling sites are either S85C or S100C, both close to F87 in the strong dimer interface (Sun et al., 2022). The T* population was determined from the relative intensities of the T and T* peaks. Including the T* population (7%) previously reported for TTR^F (Sun et al., ^2018b), all T* populations are within a range of 4%–7%, corresponding to ΔG relative to the correctly packed tetramer T of ~1.5–1.9 kcal/mol at 298 K, based on the Boltzmann distribution. The buried surface area of F87 decreases from 120 Ų in the WT tetramer to 11 Ų in A120L, where the aromatic ring is displaced from its binding pocket and becomes exposed to solvent. Based on a coefficient of 15 cal/mol/Ų at 298 K (Vallone et al., 1998), the associated ΔG is ~1.6 kcal/mol. This comparison indicates that the measured T* population is a result of the Boltzmann distribution associated with the energetic difference linked to undocking of one of the four F87 sidechains from its binding pocket in the strong dimer interface of the tetramer.

TABLE 2.

Population of the T* state for TTR variants at 298 K and pH 7.0.

Mutant	T* (%)
TTRF	6.8 ± 1.3
A25TF	5.1 ± 0.7
V30MF	6.4 ± 0.7
L55PF	6.7 ± 0.2
A81TF	7.4 ± 0.5
G83RF	5.4 ± 2.1
A97GF	4.4 ± 2.3
T119MF	4.7 ± 0.7
A120SF	3.5 ± 2.2
V122IF	4.7 ± 0.6
C10S‐S100CF	4.1 ± 0.3
Mean	5.4 ± 1.3

Note: The superscript “F” denotes that the mutations were introduced into the C10S‐S85C‐BTFA (TTR^F) background or the BTFA labeling for C10S‐S100C. The T* population for TTR^F from Sun et al. (2018b) is included for comparison. Errors are fitting uncertainties calculated as 1 standard deviation from 50 bootstrapped datasets.

Abbreviation: TTR, transthyretin.

3.4. The T* state displays enhanced conformational fluctuations of interfacial residues

We have so far relied on one ¹⁹F probe, the CF₃ group of BTFA coupled to C10S‐S85C TTR, to report on the T* state. To expand the number of probes using backbone amide cross peaks, we mixed ¹⁵N‐labeled WT‐TTR with unlabeled A120L and recorded a series of HSQC spectra over 5 days at 298 K (Figure 6a). Aggregation does not occur under these conditions. We observed extensive intensity decay for a set of amide cross peaks, which could be fitted by a global rate constant (0.036 ± 0.003 h⁻¹) (Figure 6b, Figure S4). The population of the T* state, measured by ¹⁹F NMR (see below), increases as the molar ratio of A120L:WT is changed from 1:3 to 1:1. The increase in T* population is associated with amplitude loss of amide cross peaks (Figure 6c). Interfacial residues, particularly at the AB loop, and F and H β‐strands, show pronounced peak broadening together with residues in the EF helix and across the entire DAGH β‐sheet (Figure 7a,b).

FIGURE 6.

(a) ¹⁵N HSQC spectra of ¹⁵N‐labeled WT TTR before and immediately after mixing with unlabeled A120L at a 1:1 molar ratio. (b) Time‐dependent peak intensity changes of the three residues that show the largest loss of signal amplitude, normalized by the peak intensity of E127 at the first time point. The black solid lines denote single exponential fits. For comparison, the G6 amide shows nearly constant peak intensity (green data points). The cross peaks for these residues are identified by arrows in the HSQC spectrum in panel (a). (c) The percentage of cross peak amplitude loss after 5 days as a function of amino acid residue number at 1:1 (red) and 1:3 (blue) A120L:TTR molar ratios. The location of the secondary structure elements is indicated at the top of (c). HSQC, heteronuclear single quantum coherence; WT TTR, wild‐type transthyretin.

FIGURE 7.

(a) Location of residues that exhibit greater than 70% (red spheres) or 60%–70% (orange spheres) HSQC cross peak intensity loss upon mixing unlabeled A120L and ¹⁵N‐labeled WT TTR in a 1:1 ratio. The structure shown is that of a WT tetramer (PDB code 5CN3). (b) The DAGH β‐sheets in the strong dimer as viewed from the weak dimer interface. The residues that exhibit cross peak intensity loss in the HSQC spectrum of the 1:1 mixture are colored as in panel (a). HSQC, heteronuclear single quantum coherence; WT, wild‐type.

To deconvolute the populations of ¹⁵N‐labeled WT TTR in the A120L:WT heterotetramers in the two HSQC experiments (Figure 6c), we need to estimate free energy changes of the T/M equilibrium for each tetramer species. To extract ΔG values for formation of the 1:3, 2:2, and 3:1 A120L:WT heterotetramers ($T_{{\mathrm{WT}}_1\mathrm{A}120{\mathrm{L}}_3}$, $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$, and $T_{{\mathrm{WT}}_3\mathrm{A}120{\mathrm{L}}_1}$), we fitted the T* populations of 10 μM TTR^F as a function of A120L concentration (Figure 8a and also see Section 2). We found that the fitted ΔG values match well with stoichiometry‐weighted ΔG values based on the T/M ΔG values for WT and A120L TTR (Figure 8b), consistent with a simple statistical mixing. Based on the free energy changes, we estimated the population distribution for the various tetramers formed under the conditions of the HSQC mixing experiments (Table 3). Changing the mole ratio of A120L from 1:3 A120L:WT to 1:1 A120L:WT greatly increases the population of the 2:2 heterotetramer $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$ (51% vs. 19%) and reduces the population of the WT homotetramer $T_{\mathrm{WT}}$ from 54% to 18%. The conformational fluctuations are enhanced in $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$ compared to $T_{\mathrm{WT}}$. Compared to $T_{\mathrm{WT}}$, amide cross peaks of $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$ are associated with more pronounced exchange‐induced broadening, particularly for residues at the subunit interfaces (Figure 7a), thus leading to more signal loss in the 1:1 A120L:WT dataset than the 1:3 A120:WT dataset (Figure 6c).

FIGURE 8.

(a) Fits of the T* population of TTR^F (10 μM) measured by ¹⁹F‐NMR as a function of A120L concentration at 298 K and pH 7.0. Error bars are fitting uncertainties calculated as 1 standard deviation from 50 bootstrapped datasets. (b) Fitted T/M ΔG values for 1:3, 2:2, and 3:1 heterotetramers (y coordinates for blue circles, calculated from populations in panel a using Equation (8), as described in Section 2) match with the stoichiometry‐weighted ΔG (x coordinates of blue circles) calculated using the respective T/M ΔG of WT and A120L (red triangle) at 298 K. The WT T/M ΔG is from Hurshman Babbes et al. (2008). The solid line denotes a linear fit. WT, wild‐type.

TABLE 3.

Populations of homotetramer $T_{\mathrm{WT}}$ and heterotetramers $T_{{\mathrm{WT}}_3\mathrm{A}120{\mathrm{L}}_1}$, $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$, and $T_{{\mathrm{WT}}_1\mathrm{A}120{\mathrm{L}}_3}$ containing at least one ¹⁵N‐labeled WT subunit for the two HSQC mixing experiments.

Population	A120: WT = 200:600 μM	A120: WT = 200:200 μM
$T_{\mathrm{WT}}$	0.54	0.18
$M_{\mathrm{WT}}$	<0.001	<0.001
$T_{{\mathrm{WT}}_3\mathrm{A}120{\mathrm{L}}_1}$	0.27	0.26
$T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$	0.19	0.51
$T_{{\mathrm{WT}}_1\mathrm{A}120{\mathrm{L}}_3}$	<0.01	0.05

Note: The WT monomer population is minimal under these concentrations.

Abbreviations: HSQC, heteronuclear single quantum coherence; WT, wild‐type.

4. DISCUSSION

The low population of the alternatively packed T* state (~7%) in WT human TTR makes the determination of its molecular structure challenging. We have previously shown that the population of the T* state of TTR^F can be increased by mixing with A120L and that the less stable T* state is associated with the mispacking of the F87 sidechain (Sun et al., ^2018b). In the present work, we determined the x‐ray structure of the A120L tetramer as a model of the T* state. Due to steric clash with the bulky L120 sidechain, the F87 aromatic ring is displaced from its hydrophobic pocket in the neighboring subunit across the strong dimer interface of the tetramer and rotates into a solvent‐accessible position (Figure 4c). By contrast, the F87 sidechain remains docked in the A120S structure (Figure S2C), explaining the enhanced thermodynamic stability of A120S compared to A120L at pH 7.0 (Figure 2a). The B factors of A120S are very similar to those of WT TTR whereas F87 displays elevated B factors throughout the EF region that likely reflect enhanced dynamics due to loss of the F87 sidechain packing interactions. We also note that residues in the FG loop (residues 97–104) consistently show elevated B factors in both the A120L and A120S structures as well as in WT‐TTR (Figure 4b).

Displacement of F87 from its binding pocket, together with increased backbone flexibility in the EF region, allows the sidechain to sample multiple solvent‐exposed conformations. In the MD simulations for A120L, F87 samples multiple χ1 and χ2 dihedral angles (Figure 5b,d), suggesting that in the T* state, F87 can interconvert between these conformations on the nanoscale timescale. By contrast, the direct conversion between the native T and the alternative T* state is much slower for TTR^F (0.038 h⁻¹) (Sun et al., ^2018b) and is comparable to the subunit exchange rate constant of the WT‐TTR (0.03–0.04 h⁻¹) (Rappley et al., 2014; Wiseman et al., ²⁰⁰⁵). Consistently, the rate at which the amide cross peaks lose intensity when WT‐TTR is mixed with A120L (0.036 h⁻¹) is also within this range of subunit exchange rates (Figure 6b, Figure S4). The similarity in rate constants indicates that subunit exchange mediates the process by which A120L becomes incorporated into WT:A120L heterotetramers. Moreover, the loss of amide cross peak intensity upon mixing of ¹⁵N‐labeled WT TTR with unlabeled A120L shows that the T* state is associated with increased conformational dynamics. Not all residues lose cross peak intensity to the same extent and interfacial residues, in particular, lose much of their cross peak intensity during the 5 days of data acquisition, indicating enhanced conformational fluctuations associated with the destabilized tetramer interfaces (Figures 6c and 7a,b). The amide cross peak intensity loss increases as the molar ratio of A120L mixed with WT‐TTR increases (Figure 6c), in agreement with the titration assay for TTR^F with A120L by ¹⁹F‐NMR (Sun et al., ^2018b). Increasing the proportion of A120L from an A120L:WT ratio of 1:3 to 1:1 greatly increases the population of $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$ (Table 3) and causes a greater than average intensity loss for cross peaks of V20, G22, and S112 in the weak dimer interface (2.4–2.8‐fold intensity loss compared to the 1.8‐fold intensity loss averaged over all residues, Figure 6c). A greater than average intensity loss (2.5–4.4‐fold) is also observed for several residues in the CD, D, and DE regions (residues 52, 54, 57, 59, and 62), suggesting that destabilization of the strong dimer interface by displacement of the F87 side chain leads to enhanced conformational fluctuations in regions that are implicated in amyloid formation (Childers and Daggett, 2020; Eneqvist et al., ²⁰⁰⁰; Hörnberg et al., ²⁰⁰⁴; Olofsson et al., ²⁰⁰⁴; Serag et al., ²⁰⁰²; Srinivasan et al., ²⁰²⁰; Yang et al., ²⁰⁰⁵; Yang et al., ²⁰⁰⁶; Yee et al., ²⁰¹⁹). The amide cross peak of C10, which is hydrogen bonded through its side chain to the G57 amide NH in the WT TTR tetramer, gains intensity in the 1:1 A120L:WT spectrum, possibly due to disruption of the interactions with G57.

The interfacial residues that display enhanced conformational fluctuations and cross peak broadening in the T* state (Figure 7a,b), particularly in $T_{{\mathrm{WT}}_2\mathrm{A}120{\mathrm{L}}_2}$ (Table 3), have broadened resonances in the unstable and pathogenic A25T variant (Sun et al., 2024). These interfacial residues are also perturbed due to the incorporation of 6‐fluoro‐tryptophan at W79 in the EF helix, which reduces tetramer stability (Sun et al., 2017), and many undergo millisecond timescale conformational fluctuations revealed by NMR relaxation dispersion studies of WT‐TTR and the V30M, L55P, and V122I variants (Leach et al., 2018). This comparison suggests that increased dynamics associated with perturbation of the protomer interfaces are coupled to the enhanced aggregation propensity of the T* state. The relevant dynamics are likely on the millisecond timescale, not on the fast nanosecond timescale probed in the MD simulations. Indeed, most of the residues that exhibit cross peak broadening in the T* state undergo millisecond time scale conformational fluctuations in monomeric TTR variants (Leach et al., 2024; Lim et al., ²⁰¹³), highlighting the importance of correctly packed subunit interfaces for stabilization of the TTR tetramer.

The observation that the population of the T* state is relatively conserved (~4%–7%) in WT human TTR and several pathogenic variants (Table 2) provides new insights into tetramer assembly. We have shown here and elsewhere (Sun et al., ^2018b) that TTR assembles to form both a correctly packed (T) and a mispacked tetramer (T*) in which an F87 aromatic ring is displaced from its binding pocket in the neighboring subunit across the strong dimer interface. The population of T* is consistent with a Boltzmann distribution over two tetrameric states (T and T*) that differ in free energy by ~1.6 kcal/mol at 298 K, the free energy associated with displacement and solvent exposure of a single F87 sidechain. This free energy change is identical to the ΔΔG between the T and T* states at 298 K for a destabilized A25T mutant determined by van't Hoff analysis (Sun et al., 2024). Compared to a free energy change of 32.8 kcal/mol for WT TTR to fully dissociate from tetramer to monomers at 298 K (Hurshman Babbes et al., 2008), the free energy difference between the T and T* states is relatively small. We envisage that, during the process of tetramer formation, a small population of molecules assemble with a misplaced F87 sidechain, that is, form the T* state. Interconversion between the T and T* states is slow (0.038 h⁻¹ in TTR^F) (Sun et al., ^2018b) so that, once formed, the mispacked T* tetramer is kinetically trapped. Over long time periods, TTR will reach a thermodynamic equilibrium in which there is a Boltzmann distribution over the T and T* states, and likely even higher energy states in which more than one F87 phenyl ring is displaced from the strong dimer interface. However, the population of such postulated higher energy states is too small to be observable in our experiments.

In conclusion, we have used a combination of solution NMR, x‐ray crystallography, and MD simulations to elucidate the structural consequences of F87 sidechain mispacking in the T* state, which is less stable than the native T state. Population of the T* tetramer is a natural consequence of the small energy difference between the native tetramer and one in which a single F87 aromatic ring is undocked from the strong dimer interface, resulting in a Boltzmann distribution of the T and T* states. The enhanced dynamics at interfacial residues likely predispose the T* TTR tetramers to dissociate more readily and subsequently enter the aggregation pathway.

AUTHOR CONTRIBUTIONS

Xun Sun: Conceptualization; investigation; methodology; validation; visualization; writing – original draft; writing – review and editing. James A. Ferguson: Investigation. Ke Yang: Investigation. Robyn L. Stanfield: Investigation. H. Jane Dyson: Writing – review and editing; visualization; data curation. Peter E. Wright: Conceptualization; funding acquisition; writing – review and editing; validation; project administration; supervision.

FUNDING INFORMATION

This work was supported by National Institutes of Health Grants DK124211 (P.E.W.) and GM131693 (H.J.D.). X.S. acknowledges past fellowship support from American Heart Association grants #17POST32810003 and #20POST35050060. The Berkeley Center for Structural Biology is supported in part by the Howard Hughes Medical Institute. The Advanced Light Source is a Department of Energy Office of Science User Facility under Contract No. DE‐AC02‐05CH11231.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Table S1. X‐ray data collection and refinement statistics for A120S and A120L.

Figure S1. SEC and DOSY characterization of A120L.

Figure S2. X‐ray structure overlays for A120L and A120S with the WT‐TTR.

Figure S3. ¹⁹F‐NMR spectra of BTFA‐labeled TTR mutants at 298 K and pH 7.0.

Figure S4. Raw kinetic datasets for the A120L:WT real‐time mixing experiment.

Sun X, Ferguson JA, Yang K, Stanfield RL, Dyson HJ, Wright PE. Mispacking of the F87 sidechain drives aggregation‐promoting conformational fluctuations in the subunit interfaces of the transthyretin tetramer. Protein Science. 2024;33(9):e5101. 10.1002/pro.5101

DATA AVAILABILITY STATEMENT

The x‐ray structures of A120L and A120S have been deposited in the PDB with accession codes as 8W2W and 8W1N, respectively.