A molecular basis for tetramer destabilization and aggregation of transthyretin Ala97Ser

Abstract

Transthyretin (TTR)‐related amyloidosis (ATTR) is a syndrome of diseases characterized by the extracellular deposition of fibrillar materials containing TTR variants. Ala97Ser (A97S) is the major mutation reported in Taiwanese ATTR patients. Here, we combine atomic resolution structural information together with the biochemical data to demonstrate that substitution of polar Ser for a small hydrophobic side chain of Ala at residue 97 of TTR largely influences the local packing density of the FG‐loop, thus leading to the conformational instability of native tetramer, the increased monomeric species, and thus the enhanced amyloidogenicity of apo‐A97S. Based on calorimetric studies, the tetramer destabilization of A97S can be substantially altered by interacting with native stabilizers via similarly energetic patterns compared to that of wild‐type (WT) TTR; however, stabilizer binding partially rearranges the networks of hydrogen bonding in TTR variants while FG‐loops of tetrameric A97S still remain relatively flexible. Moreover, TTR in complexed with holo‐retinol binding protein 4 is slightly influenced by the structural and dynamic changes of FG‐loop caused by A97S substitution with an approximately five‐fold difference in binding affinity. Collectively, our findings suggest that the amyloidogenic A97S mutation destabilizes TTR by increasing the flexibility of the FG‐loop in the monomer, thus modulating the rate of amyloid fibrillization.

Short abstract

PDB Code(s): 7Y6J, 7YBR, 7YCQ and 8HY4;

1. INTRODUCTION

Transthyretin (TTR) is responsible for transporting thyroid hormone thyroxine (T₄) to target cells and associating with retinol‐binding proteins, which carry retinol to maintain the plasma level of retinol (Nilsson et al., 1975; Steinhoff et al., ²⁰²¹; Zanotti & Berni, ²⁰⁰⁴). In the native state, TTR assembles into a 55‐kDa tetramer with a dimer of dimers quaternary structure, comprising two T₄ binding sites. Currently, more than 150 autosomal dominant variants of the TTR gene are described. Most pathogenic TTR variants have been identified with decreased stability and increased dissociation from tetramers into monomers, resulting in senile systemic amyloidosis, familial amyloid cardiomyopathy, and familial amyloid polyneuropathy (FAP). Of all variants, Val30Met (V30M) with two types of onset patterns (early‐ and late‐onset) is the most common mutation involved in hereditary transthyretin amyloidosis (ATTR) neuropathy worldwide (Hamilton et al., 1993; Plante‐Bordeneuve & Said, ²⁰¹¹). Leu55Pro (L55P) exhibits a highly unstable tetramer variant, reflecting the early‐onset aggressive diffuse amyloidosis, severe cardiac, and neurologic dysfunctions (Hammarstrom et al., 2002; Lashuel et al., ^{1998, 1999}). Ala25Thr (A25T) is known for causing leptomeningeal amyloidosis (Sekijima et al., 2003). Val122Ile (V122I) variant is associated with ATTR cardiomyopathy, resulting from myocardial deposition of misfolded TTR (Jiang et al., 2001). Furthermore, Thr119Met (T119M), Arg104His (R104H), and Ala108Val (A108V) are dissociation‐resistant mutations that perform protective effects on FAP by raising the kinetic barrier of tetramer dissociation (Johnson et al., 2005; Sant'Anna et al., ²⁰¹⁷; Sekijima et al., ²⁰⁰⁶; Terazaki et al., ¹⁹⁹⁹).

Previously studies have reported that Ala97Ser (A97S) is a hotspot TTR variant identified in Taiwanese FAP patients by genetic analysis (Hsieh, 2011; Hsu et al., ²⁰¹⁸; Lachmann et al., ²⁰⁰⁰; Lai et al., ²⁰¹⁵; Liu et al., ²⁰⁰⁸) while an increasing number of cases have been reported in Southern China (Du et al., 2021), Malaysia (Low et al., 2019), and Thailand (Pasutharnchat et al., 2021). In contrast with V30M, FAP patients with A97S variant have a late disease onset with rapid disease progression (Chao, Liao, et al., 2019) and also show a mixed polyneuropathy‐cardiopathy phenotype. However, there are three cases diagnosed with early‐onset polyneuropathy in Thailand (Pasutharnchat et al., 2021). The great majority of A97S‐TTR FAP patients may present with peripheral neuropathy, along with a variety of autonomic symptoms. Skin denervation is commonly found and possibly regarded as a biomarker for progression stage of A97S‐TTR FAP (Chao, Hsueh, et al., 2019; Yang et al., ²⁰¹⁰). Moreover, the cardiomyopathy syndromes are found in A97S variant patients with polyneuropathy (Lai et al., 2020). About half of the A97S patients show dysphagia symptoms and some also develop carpal tunnel syndrome at earlier onset age while dysphagia is not common in the late‐onset V30M (Hsueh et al., 2021). The autopsy of an A97S patient has provided the extensive evidence of amyloid pathology in the tongue and larynx, suggesting that swallowing function could be influenced by the amyloid deposits (Hsueh et al., 2021). Furthermore, the experiments of cellular secretion, cytotoxicity assay, and transgenic flies engineered with human A97S‐TTR have been performed to help understand the cellular mechanism of amyloidogenic A97S. The results show A97S mutation has shortened the lifespan of Drosophila with late‐onset fast neurodegeneration and has caused intermediate cytotoxicity compared to V30M (Ibrahim et al., 2020; Kan et al., ²⁰¹⁸). However, the structure of A97S‐TTR has not been solved yet. It is not clear how amyloidogenic A97S variant affects the tetramer instability and triggers the fibril formation.

Therapeutic approaches targeting on the process of amyloidogenicity have been developed for pathogenic variants‐related ATTR, including liver transplantation, native TTR stabilizers, gene modifiers, and antisense oligonucleotide therapies (Bezerra et al., 2020; Cristobal Gutierrez et al., ²⁰²⁰; Ihne et al., ²⁰²⁰; Johnson et al., ²⁰⁰⁸). However, late‐onset ATTR patients are not eligible for liver transplantation (Suhr, 2003). It has been shown that native‐state stabilizers are promising drugs to treat TTR amyloidosis by binding to TTR dimer–dimer interface and elevating the energy barrier for tetramer destabilization, which is kinetically coupled with the aggregation equilibria (Yokoyama & Mizuguchi, 2020). Taken some small‐molecule drugs, for example, diflunisal, which is a nonsteroidal anti‐inflammatory agent, binds to TTR T₄‐binding sites and inhibits FAP progression in clinical trials (Adamski‐Werner et al., 2004; Berk et al., ²⁰¹³). Tafamidis is a TTR structure‐based designed compound and has been approved to treat early‐stage FAP in Europe and Japan (Merkies, 2013; Zhao et al., ²⁰¹⁹). AG10 is still being evaluated with regard to clinical outcomes (Judge et al., 2019; Penchala et al., ²⁰¹³). Epigallocatechin‐3‐gallate (EGCG) prominently reduces the rate of amyloid fibrillization in vitro and decreases the intensity of fibril in nondiabetic mice (Almeida & Saraiva, 2012). Tolcapone, a catechol‐O‐methyltransferase inhibitor, is authorized to treat Parkinson's disease in the United States and Europe (Artusi et al., 2021; Leegwater‐Kim & Waters, ²⁰⁰⁷). Notably, tolcapone can kinetically stabilize TTR tetramer, reduce TTR aggregation, and prevent TTR‐induced cardiotoxicity more effectively than tafamidis by occupying the T4‐binding sites located at the interface of TTR dimer–dimer (Pinheiro et al., 2021; Sant'Anna et al., ²⁰¹⁶). Although previous work has demonstrated the destabilization of A97S‐TTR by chemical and thermal denaturation at neutral pH (Liu et al., 2019), the underlying molecular basis has remained elusive due to the lack of structural information of A97S‐TTR and the quantitative details of A97S aggregation pathway at mildly acidic pH. Our study fills in this gap by providing for the crystal structures of A97S‐TTR in the apo‐ and stabilizer‐bound states. Combining structural and mutational analyses with biophysical and biochemical assays, our results reveal that A97S substitution causes a dramatic effect on the structure and flexibility of FG‐loop, thus switching the dissociation equilibrium of tetramer, and triggering the formation of amyloid fibril.

2. RESULTS

2.1. The FG‐loop is highly flexible in A97S variant

We determined the cryogenic crystal structure of apo‐A97S at 1.38 Å resolution (Table S1, PDB ID: 7Y6J) to gain structural insights. Compared with the structure of apo wild‐type (WT) TTR (derived from protein data bank, ID: 2QGB), the overall structure of apo‐A97S adopted extremely similar the tetrameric assembly with the root mean square deviation (RMSD) value of 0.174 Å for alignment of the main‐chain atoms of 266 equivalent residue pairs from the core regions, excluding the loop regions from superposition (Figure 1a). The secondary structure of A97S was comprised of eight beta‐strands and a short alpha‐helix, which were β strands A to H and EF helix (Figures S1a, b). Residue 97 is located at the end of β strand F close to the FG‐loop spanning residues 98–103. DSSP (hydrogen bond estimation algorithm) analysis showed that the secondary structure elements of A97S variant were highly similar to that of the WT. The methyl‐group side chain of Ala‐97 formed hydrophobic contacts with Phe‐64, Tyr‐69, Phe‐95, and Tyr‐105. Although A97S substitution introduced a hydroxyl group on the side chain of C_β atom, the beta‐protons of Ser‐97 still maintained these hydrophobic interactions with a slight rotation of the side chain of Phe‐64 in chain A (Figure 1b). Notably, the electron density map of the FG‐loop was apparently broken due to the hydroxyl side chain of A97S in chain A while that in chain A of WT can be interpreted and refined well (Hornberg et al., 2000) (Figure S1c–f). We further examined the B‐factors, which indicated the fluctuation of atoms around their equilibrium positions caused by thermal motion and positional disorder, for both structures of A97S and WT in the apo state (Figures S1a and S2). It suggested that the point mutation A97S significantly increased the flexibility of FG‐loop, which was obviously revealed by the decreased quality of the electron density map, and the prominently raised B‐factors in this region (Figures S1 and S2).

FIGURE 1.

Structural comparison between A97S‐TTR and WT‐TTR in the apo state. (a) Superposition of apo A97S‐TTR (bright orange, ID: 7Y6J) and WT‐TTR (cyan, ID: 2QGB) with a RMSD value of 0.174 Å for aligned 266 equivalent residue pairs of the core regions. Tetrameric assembly of TTR represented by a cartoon diagram with four chains labeled as A, B, symmetry‐related A′ and B′, respectively. Poor‐defined FG‐loop (spanning residues 98–103) in chain A of A97S structure is presented with a dashed line. (b) Zoom‐in viewing the highlighted boxes display Ala‐97/Ser‐97 forms the hydrophobic contacts with residues Phe‐64, Tyr‐69, Phe‐95, and Tyr‐105 in chain A (top) and chain B′ (bottom), respectively. Structures are shown by cartoon with 50% of transparency, and the side chains involved in hydrophobic contacts with Ala‐97/Ser‐97 are presented by sticks with a radius set to 0.15. Residue 97 and the FG‐ loop are pointed by black arrow. (c), (d) Intramolecular H‐bond networks of FG‐loop in chain A of apo‐structures of WT (c, ID: 2QGB) and A97S (d, ID: 7Y6J). Transparent cartoon is used to clearly visualize the hydrogen‐bonding residues presented as a stick. The residues 97–103, which include the point mutation residue 97 and FG‐loop spanning residues 98–103, are colored in cyan and bright orange for WT and A97S, respectively. H‐bonds are indicated as dashed green lines. Nine hydrogen bonds (H‐bonds) are identified in the FG‐loop of chain A of apo‐WT while only one H‐bond is observed with that of apo‐A97S (Table S2). (e), (f) The networks of intramolecular H‐bonding in the FG‐loop regions of chain B of apo‐ WT (e) and A97S (f) structures are shown with dashed lines colored in green. Backbones and side chains of the H‐bonding residues are presented as sticks. Twelve H‐bonds are recognized in the FG‐loop of chain B of apo‐WT while nine H‐bonds are formed in that of apo‐A97S (Table S2).

Moreover, we found that there were nine hydrogen bonds (H‐bonds) identified in the FG‐loop of chain A of apo‐WT by superposition of 16 reported structures of WT‐TTR in the apo state (Hornberg et al., 2000) (Figures 1c, S3, and Table S2). The carboxamide group of Asn‐98 side chain formed three H‐bonds with the backbone amide of Arg‐103, and the backbone carbonyls of Phe‐64 and Val‐65, respectively. The backbone of Asn‐98 formed two H‐bonds with the side chain phenolic hydroxyl of Tyr‐105 and the backbone amide of Gly‐101, respectively. An intra‐residue H‐bond was observed between the amide proton and side chain carbonyl oxygen of Asp‐99. The side chain of Arg‐103 formed three H‐bonds with the backbone carbonyls of Ser‐100 and Pro‐102, respectively. Contrarily, in chain A of apo‐A97S, only one H‐bond was observed and formed by the side chain carbonyl of Asn‐98 with the side chain of Glu‐66 (Figure 1d and Table S2), indicating that introducing the hydroxyl sidechain of A97S dramatically abolished hydrogen‐bonding interactions and thus pronouncedly led to the raised flexibility of FG‐loop in chain A.

Furthermore, there were twelve H‐bonds identified in the FG‐loop of chain B in apo‐WT (Figure 1e and Table S2). The backbone of Asn‐98 formed two H‐bonds with the side chain of Tyr‐105; the side chain carbonyl oxygen of Asn‐98 formed three H‐bonds with Ser‐100 and the side chain carbonyl of Glu‐66, respectively. The backbone of Asp‐99 formed three H‐bonds with the side chains of Arg‐103 and Glu‐66, respectively. The backbone amide of Arg‐103 formed two H‐bonds with the backbone carbonyl of Thr‐123 and the side chain carbonyl of Asn‐124, respectively; the side chain of Arg‐103 formed two H‐bonds with the backbone carbonyls of Ser‐100 and Gly‐101, respectively. In contrast to WT, there were nine H‐bonds identified in chain B of apo‐A97S, and four out of nine H‐bonds were identified in chain B of apo‐WT, including that two H‐bonds were formed by the backbone of Asn‐98 with phenolic hydroxyl of Tyr‐105, and two H‐bonds were formed by the side chains of Arg‐103 and Glu‐66 with the backbones of Ser‐100 and Asp‐99, respectively (Figure 1f and Table S2). Five out of nine H‐bonds were mainly contributed by residues Asn‐98, Asp‐99, Ser‐100, Gly‐101, and Arg‐103 as listed in Table S2. Collectively, A97S mutation also partially reduced H‐bonding interactions and caused the flexibility change of FG‐loop in chain B.

A97S substitution remarkably resulted in repositioning of both main‐ and side chain of Asn‐98, thus dramatically impacting the H‐bonding networks and the stability of the FG‐loops of both chains A and B. We further examined the H‐bond networks in the interfaces between chains A, A′, B, and B′ for the apo‐form structures of both A97S and WT. Two H‐bonds formed by the backbone carbonyl of Gly‐22 and backbone amide of Val‐122 were identified at the interfaces of chains A and A′ as well as chains B and B′ for both A97S and WT structures. Moreover, we identified 18 H‐bonds at the interfaces between chains A and B of both TTR variants. Sixteen out of eighteen H‐bonds of A97S were identical to those of WT (Table S3), suggesting that A97S substitution almost adopted the similar interfaces between chains A and B.

2.2. Conformational stability of TTR A97S variant

The pathogenic mechanism of TTR variants has been thought to be involved in either their tetramer stability or the misfolded monomeric species (Adams et al., 2019). To characterize whether the A97S mutation introduced the tetramer destabilization, we performed acid treatment, Congo‐red (CR) staining, turbidity assay, pH‐jump experiment, SDS‐trapping (S‐Trap) assay, and calcium treatment. First, we induced aggregation of TTR variants with acidic buffer (pH 4.0) and monitored the insoluble aggregates in the absence and presence of stabilizers. Within a 4‐day incubation period, the concentrations of soluble TTR variants declined along with the incubation time of acid treatment (Figure 2a–f). The results revealed that the insoluble aggregate formations of A97S and V30M were much faster than those of WT and R104H in the absence of stabilizers (Figure 2a). Although the tetramer instability of A97S was extremely similar to that of V30M in the apo state, A97S demonstrated much better resistivity against acid treatment than V30M in the presence of T4 (Figure 2a, b) and tolcapone (Figure 2c, d) at various concentrations. However, A97S and V30M had almost the same stability in the presence of diflunisal (Figure 2e, f). Moreover, the fibril formations of TTR variants have been supposed to take place in the acidic environment of the lysosome during protein turnover (Hammarstrom et al., 2002; Lashuel et al., ¹⁹⁹⁹). Therefore, the amyloidogenicity for A97S was evaluated with CR staining, which was used to identify amyloid fibrils. The result demonstrated that the fibrillization of A97S and V30M was much more significant than that of both WT and R104H (Figure 2g) while the amyloid content of A97S exhibited quite similar reactivity of CR to that of V30M at acidic pH.

FIGURE 2.

Characterization of tetramer stability of A97S‐TTR by acid‐induced aggregation assay and Congo‐red binding assay. (a), (b) TTR variants (final tetramer concentration ~ 163 μM) were pre‐treated with final 1% DMSO or nature ligand T₄ at concentrations of 81.5 μM (a) and 163 μM (b), respectively, and were incubated in acidic buffer (pH 4.0) at 37°C for 1, 2, 3, and 4 days, respectively. After centrifugation for separation of aggregates, the concentrations of soluble TTR variants in solutions were determined by the absorbance of 280 nm. The changes in residual TTR variants were indicated by the ratios of calculated soluble TTR concentrations at specific time point over 163 μM, and plotted against incubation time. WT‐, A97S‐, V30M‐, and R104H‐TTR groups are colored red, orange, cyan, and purple, respectively. The DMSO‐control groups are represented by solid lines; the T₄‐treated groups are denoted by dashed lines. The soluble A97S and V30M are much less than WT and R104H in the apo state while T₄ binding largely reduces the insoluble TTR variants in the acidic condition. (c), (d) Pre‐treatment of four TTR variants with 81.5 μM (c) and 163 μM (d) of the small‐molecule tolcapone were induced to form fibrils by incubating in an acidic condition (about pH = 4.0) lasting for 1–4 days. The tolcapone‐treated groups are denoted by dashed lines. WT‐, A97S‐, V30M‐, and R104H‐TTR groups are colored red, orange, cyan, and purple, respectively. The soluble TTR variants are prominently increased in the presence of tolcapone while A97S shows much better resistivity against acid treatment than V30M in the presence of T4 and tolcapone. (e), (f) The relative residual levels of soluble TTR pre‐incubated with NSAID diflunisal at two different concentrations 81.5 μM (e) and 163 μM (f), respectively. The diflunisal‐treated groups are denoted by dashed lines. WT‐, A97S‐, V30M‐, and R104H‐TTR groups are colored red, orange, cyan, and purple, respectively. A97S and V30M show almost the same stability in the presence of diflunisal. (g) Quantification of acid‐induced fibrils of TTR variants by using Congo‐red (CR) dye. WT, A97S, V30M, and R104H are represented by bars colored red, orange, cyan, and purple, respectively. Error bars indicate the standard deviations (n = 3). Both WT and R104H have less fibril formation compared with A97S and V30M; the amyloid content of A97S exhibits quite similar reactivity of CR to that of V30M at acidic pH.

Previous studies have shown that the tetrameric WT‐TTR is quite stable from pH 7 to 5 but not under pH 5 (Yang et al., 2003) and therefore acid‐mediated aggregation has been widely applied in most studies of TTR variants. Indeed, the acid‐treated A97S‐ and WT‐TTR samples were inspected by the micrographs of transmission electron microscopy (TEM), indicating both A97S‐ and WT‐TTR at pH 4.0 had protofibril formation (Figure S4a, b). It was of interest to investigate how the conformation of A97S‐TTR could be affected by pH fluctuations. Here, we characterized pH‐mediated amyloidogenicity of A97S‐ and WT‐TTR by monitoring the turbidity at 330 nm along time (Figure 3a, b), revealing that A97S greatly developed obvious turbidity (from pH 5 to 4) as compared with WT. Both A97S‐ and WT‐TTR samples were then examined with TEM (Figure 3c, d), demonstrating that A97S formed evident prefibrillar aggregates at pH 5. We also incubated samples at pH 5.25 (Figure S4c, d) and found that A97S appeared large amorphous aggregates mixed with thin silk‐like deposits, which was highly similar to acid‐induced aggregation of V122I as previously reported (Robinson & Reixach, 2014). In contrast to WT‐TTR, A97S had a remarkable potential tendency to form amyloid fibrils under pH 5.25. Moreover, the seminative SDS‐PAGE experiments were further utilized to evaluate pH‐mediated structural changes of A97S, revealing that pH alterations easily caused instability of A97S within 3 h while WT‐TTR only showed increased monomeric population under pH 5 (Figure 3e) (Lashuel et al., 1999).

FIGURE 3.

The pH effect on fibrils formation and monomer species of A97S. (a), (b) Samples of 3.6 μM A97S‐ (a) and WT‐TTR (b) in citrate–phosphate buffer pH ranging from pH 4.0 to 7.0 were incubated at 37°C for 1, 2, 3, 7, 8, 9, and 10 days, and the turbidity was monitored at 330 nm. Error bars indicate the standard deviations of three replicate data. In the range of pH 5.0 to 4.0, A97S develops much more remarkable turbidity compared with WT. (c), (d) Electron micrographs of A97S (c) and WT (d) under pH 5.0 condition of citrate–phosphate buffer. Scale bar = 100 nm. At pH 5.0, A97S shows evident prefibrillar aggregates. (e) Seminative PAGE for analyzing conformational changes of WT and A97S incubated at 37°C for 1 h (top) and 3 h (bottom) under different pH values of citrate–phosphate buffer. Two major bands of dimer and monomer are, respectively, denoted as D and M at the right sides of gels. The fractions of monomer are calculated by the intensity ratios of monomer band over the sum of both monomeric and dimeric bands at specific pH and plotted as a function of pH value. A97S and WT are represented as gray and black bars, respectively. The pH shifts easily caused instability of A97S within 3 h while WT‐TTR only shows increased monomeric population under pH 5.0.

Furthermore, we used pH‐jump experiments (from pH 4 to 7) to examine the refolding populations of TTR variants in the absence and presence of diflunisal and tolcapone (Figure S5a–c). After removing the insoluble aggregates by high‐speed centrifugation, the gel filtration analyses showed that there were no soluble aggregates and the major populations of TTR variants were tetrameric at pH 7 (Figure S5). In contrast to WT‐TTR, about 70%–75% of A97S‐ and V30M‐TTR molecules populated the tetrameric conformations while A97S had increased monomeric species in the apo state (Figure S5d). Due to stabilizer tolcapone with absorbance at 280 nm, werespectively presented TTR variants in the presence of diflunisal and tolcapone, indicating that the folded population of tetrameric A97S was about 82%–84% relative to that of WT when V30M had just 50%–60% of folded tetramer (Figure S5d‐f). These results suggested that stabilizer binding to TTR variants caused the refolding of A97S to be much more efficient than that of V30M but total folded A97S was still less than well‐folded WT.

We also examined the stability of A97S‐TTR at physiological pH using S‐Trap experiments, in which SDS was used to impel the unfolding of stable proteins and calcium‐mediated destabilization (Cantarutti et al., 2022; Palmieri Lde et al., ²⁰¹⁰; Wieczorek et al., ²⁰²¹). With the increased incubation time at 85°C, we found that monomeric A97S‐ and WT‐TTR, respectively, became the predominant species after 120 and 300 s (Figure 4a–c), suggesting that WT‐TTR was much more heat‐resistant than A97S‐TTR in the presence of SDS. Furthermore, previous reports have revealed that Ca²⁺ could interact with TTR with a pretty weak affinity, but still lead to the changes of conformational flexibility on TTR variants.(Cantarutti et al., 2022; Palmieri Lde et al., ²⁰¹⁰; Wieczorek et al., ²⁰²¹) In the presence of increased Ca²⁺ concentrations from 0 to 100 mM, we found that monomeric species of A97S‐TTR were gradually raised up to 40% while that of WT were still kept at ~5%–10% in a 16‐h incubation (Figure 4d, e), suggesting that the Ca²⁺ effect on A97S was substantially stronger than that on WT proteins.

FIGURE 4.

Conformational stability of A97S under SDS denaturation and calcium treatment. (a), (b) S‐trap assay for investigating the stability of A97S (a) and WT (b) under the action of SDS denaturation. WT‐ and A97S‐TTR undergo unfolding by heating with sample loading dye containing 1% SDS at 85°C for 0–600 s, followed by running 11% acrylamide gels. D and M indicate dimeric and monomeric bands, respectively. (c) Bar depicting the unfolding ratios of TTR variants by determining intensity ratios of monomer bands relative to the sum of dimer and monomer bands. Gray and black bars respectively correspond to A97S and WT. Monomeric A97S‐ and WT‐TTR respectively become the predominant species after 120 and 300 s suggesting that A97S is much more heat‐sensitive than WT in the presence of SDS. (d), (e) Destabilizing effect of Ca²⁺ on the structural stability of TTR. WT and A97S (5 μM) in Tris buffer pre‐incubated with 0–100 mM calcium chloride at 4°C for 16 h and analyzed by seminative PAGE. Two visible bands are respectively presented as dimer (D) and monomer (M). The fractions of TTR monomer are quantified by calculating the relative intensity ratios of monomer bands and the sum of dimer and monomer bands. In the presence of increased Ca²⁺ concentrations from 0 to 100 mM with a 16‐h incubation, monomeric WT still keeps at ~5%–10% but that of A97S‐TTR is gradually raised up to 40%, indicating that the effect of Ca²⁺ on A97S is stronger than that on WT protein.

2.3. Characterization of stabilizers tolcapone and diflunisal binding energetics to A97S‐TTR

Furthermore, we performed Isothermal Titration Calorimetry (ITC) to explore the binding of stabilizers to TTR variants. In the case of tolcapone binding to WT, we obtained the average values for Kd₁ and Kd₂ were 5.93 nM and 29.1 nM with negative cooperativity (Figure 5a and Table 1); the data were consistent with previously published results with favorable enthalpy (Sant'Anna et al., 2016). Compared with WT, tolcapone bound to A97S with Kd₁ = 1.49 nM and Kd₂ = 64.3 nM, respectively, revealing that A97S showed similar binding pattern with negative cooperativity (Figure 5b, c, and Table 1). The isotherms of diflunisal binding to TTR variants displayed that there were two binding events with very close affinity and the average Kd values for A97S and WT were 389 nM and 574 nM, respectively (Figure 5d‐f and Table 1), suggesting that A97S‐ and WT‐TTR had a similar binding affinity toward stabilizer diflunisal. From the ITC studies, we concluded that the binding affinities of stabilizers tolcapone and diflunisal to A97S were similar to those of WT, thus explaining that the tetrameric population of A97S in the presence of stabilizers was apparently increased in the acid‐mediated aggregation assay and pH‐jump experiments.

FIGURE 5.

Interactions of WT‐TTR and A97S‐TTR with TTR stabilizers assessed by ITC. (a), (b) Titration of 100 μM tolcapone against 5 μM WT‐ (a) and A97S‐ (b) TTR in the FPLC buffer (pH 7.0) containing final 2.5% DMSO at 25°C as shown in the raw data of thermograms (top). Data points of binding isotherms (bottom) are fitted by two‐site model with Kd₁ = 6.11 nM and Kd₂ = 26.0 nM for WT; Kd₁ = 1.67 nM and Kd₂ = 144 nM for A97S, respectively. (c) Bar graphs of thermodynamic parameters for two ligand‐binding sites of WT‐ and A97S‐TTR with tolcapone by fitting the titration profiles. ΔG, ΔH, ΔS, and T represent the changes in Gibbs free energy, enthalpy, entropy, and absolute temperature (K). Error bars present the standard deviations of enthalpies derived from curve fitting. The binding events of tolcapone to both A97S and WT are exothermic and contributed by strongly favorable enthalpy. (d), (e) Thermograms (top) and binding isotherms (bottom) showing the titration of 100 μM diflunisal into 5 μM WT (d) and A97S (e), respectively. Using the Malvern MicroCal PEAQ‐ITC Analysis Software, the Kd value is obtained by fitting with single‐site binding model. The dissociation constants are 292 $\pm$ 56.8 nM and 490 $\pm$ 99.0 nM for the diflunisal binding to A97S and WT, respectively. A97S‐ and WT‐TTR have a similar binding affinity toward stabilizer diflunisal. (f) Bar graphs of thermodynamic parameters for the diflunisal binding to WT and A97S. The standard deviations of enthalpy are shown with error bars. Both diflunisal binding cases are driven by favorable enthalpies with negative values of ΔH. The binding affinities of stabilizers tolcapone and diflunisal to A97S are similar to those of WT.

TABLE 1.

The binding affinities and thermodynamic components of TTR variants for stabilizers.

Sample cell/syringe	Sample number	Kd1 (nM) a	ΔG 1 (kcal mol−1)	ΔH 1 (kcal mol−1) a	−TΔS 1 (kcal mol−1)	Kd2 (nM) a	ΔG 2 (kcal mol−1)	ΔH 2 (kcal mol−1) a	−TΔS 2 (kcal mol−1)
WT/tolcapone
	No.1	6.11 ± 0.22	−11.17	−11.20 ± 0.28	0.03	26.0 ± 7.67	−10.36	−14.90 ± 0.61	4.54
	No.2	6.25 ± 1.58	−11.22	−13.80 ± 1.20	2.58	50.2 ± 6.65	−9.92	−17.60 ± 0.60	7.68
	No.3	5.44 ± 0.55	−11.27	−12.90 ± 0.90	1.63	11.1 ± 1.59	−10.83	−13.20 ± 0.80	2.37
Average b		5.93 ± 0.43	−11.22 ± 0.05	−12.63 ± 1.32	1.41 ± 1.29	29.1 ± 19.7	−10.37 ± 0.46	−15.23 ± 2.22	4.86 ± 2.70
A97S/tolcapone
	No.1	1.67 ± 0.13	−12.02	−10.20 ± 0.30	‐1.82	144 ± 17.50	−9.43	−12.30 ± 0.40	2.87
	No.2	1.17 ± 0.23	−12.18	−8.51 ± 0.28	‐3.67	45.50 ± 2.69	−10.03	−10.20 ± 0.27	0.17
	No.3	1.64 ± 0.05	−11.98	−12.00 ± 0.14	0.02	33.40 ± 0.04	−10.21	−15.00 ± 0.17	4.79
Average b		1.49 ± 0.28	−12.06 ± 0.11	−9.68 ± 1.75	‐1.82 ± 1.84	64.30 ± 43.50	−9.89 ± 0.41	−12.50 ± 2.51	2.61 ± 2.32
WT/diflunisal
	No.1	490 ± 99.0	−8.61	−11.80 ± 0.50	3.19	‐	‐	‐	‐
	No.2	720 ± 99.3	−8.38	−14.30 ± 0.46	5.92	‐	‐	‐	‐
	No.3	511 ± 68.6	−8.58	−11.80 ± 0.32	3.22	‐	‐	‐	‐
Average b		574 ± 127	−‐8.52 ± 0.13	−12.60 ± 1.44	4.11 ± 1.58	‐	‐	‐	‐
A97S/diflunisal						‐	‐	‐	‐
	No.1	292 ± 56.8	−8.91	−9.45 ± 0.32	0.54	‐	‐	‐	‐
	No.2	424 ± 68.1	−8.69	−11.30 ± 0.35	2.61	‐	‐	‐	‐
	No.3	450 ± 87.8	−8.66	−11.20 ± 0.44	2.54	‐	‐	‐	‐
Average b		389 ± 84.7	−8.75 ± 0.14	−10.65 ± 1.04	1.90 ± 1.18	‐	‐	‐	‐

^a Standard deviations (SDs) are obtained from curve fitting of each titration by the Malvern MicroCal PEAQ‐ITC Analysis Software.

^b All average values are the means ± SDs of triplicate data.

2.4. Crystal structures of A97S‐TTR in complexed with stabilizers

We further determined the structures of A97S in complexed with tolcapone and diflunisal at 1.71 Å and 1.99 Å resolution, respectively, (PDB codes: 7YBR and 7YCQ, Table S1). The 4‐methyl phenyl ring of tolcapone penetratingly positioned into the inner cavity of the T₄‐binding site, making hydrophobic contacts with residues Ala‐108, Leu‐110, Ser‐117, and Thr‐119 of the two symmetrical T₄‐halogen‐binding pockets HBP2–2′ and HBP3–3′ (Figure S6a). Additionally, the central carbonyl group of tolcapone formed one H‐bond with the side chain of Thr‐119, and the 3,4‐dihydroxy‐5‐nitrophenyl ring of tolcapone made contacts with residues Lys‐15, Leu‐17, Thr‐106, and Ala‐108 of HBP1–1′ and HBP2–2′. The electrostatic interactions contributed by Glu‐54, Lys‐15, and tolcapone limited the entrance of solvent into the HBP pockets and thus restricted water rearrangements (Figure S6a). The crystal structure of A97S/tolcapone was similar to that of WT/tolcapone with superposed Cα RMSD of 0.237 Å of aligned 236 atoms of the core structure. Furthermore, diflunisal binding to A97S also had a similar mode like that of WT (Figure S6b) with the RMSD from the superposition being 0.331 Å over all equivalent main‐chain atom pairs of the core structure (Adamski‐Werner et al., 2004). The results indicated that both tolcapone and diflunisal binding patterns to A97S were very similar to those of WT (Figure S6a, b).

Moreover, we analyzed the intramolecular hydrogen bonding and electron density maps of the FG‐loops of both A97S‐TTR and WT‐TTR in the presence of tolcapone (Figures 6a‐d, S6c, d, and Table S4). There were ten H‐bonds observed in chain A of WT‐TTR in the tolcapone‐bound state. Seven out of ten H‐bonds existed in chain A of apo‐WT‐TTR, and the three other different H‐bonds were newly formed (Figure 6a and Table S4). Contrarily, three hydrogen bonds were detected in the FG‐loop in chain A of A97S/tolcapone structure, including that one H‐bond was formed by the backbone carbonyl of Asn‐98 with the side chain of Glu‐66 (Figure 6b and Table S4), and the other two H‐bonds were generated by the side chain oxygen of Asp‐99 with its backbone amide and phenolic hydroxyl of Tyr‐105, respectively. These results suggested that tolcapone binding partially rearranged the network of intramolecular H‐bonds in the FG‐loop of chain A of WT while FG‐loops of tetrameric A97S still kept relatively flexible. For the structure of chain B of WT/tolcapone complex, eight H‐bonds were identified, including that two out of eight H‐bonds were identical to the FG‐loop in chain B of apo‐WT, and the six other different H‐bonds were newly formed in the FG‐loop in chain B of WT‐TTR/tolcapone complex (Figure 6c and Table S4 ). In contrast, there were five H‐bonds in the FG‐loop in chain B of A97S/tolcapone, including that three out of five H‐bonds, were identical to the FG‐loop in chain B of apo‐A97S; the two other H‐bonds were respectively formed by the side chain of Asn‐98 with side chain carbonyl of Glu‐66 and the side chain of Arg‐103 with the side chain of Tyr‐105 (Figure 6d and Table S4). Moreover, the similar results were observed in the structure of A97S/diflunisal complex (Figures 6e–h, S6e, f, and Table S5), suggesting that stabilizers tolcapone and diflunisal binding to both A97S‐ and WT‐TTR caused the rearrangement of the intramolecular H‐bond networks in the FG‐loops in chains A and B of both WT and A97S, but the FG‐loops of tetrameric A97S were still more flexible than those of WT.

FIGURE 6.

The H‐bonding networks of FG‐loops in the structures of A97S and WT in complex with stabilizers. (a), (b) The intramolecular H‐bonds of FG‐loop in chain A of WT (a, cyan) and A97S (b, violet) in complex with tolcapone. Dashed lines in green indicate H‐bonds. Hydrogen‐bonding residues are displayed as sticks with transparent cartoons. WT/tolcapone complex is derived from protein databank (PDB code: 4D7B). Due to low quality of the electron density map in FG‐loop, residues 100–102 in chain A of A97S/tolcapone structure are presented with a dashed violet line. In the presence of tolcapone, ten H‐bonds, in which Asn‐98 forms four H‐bonds with residues Phe‐64, Gly‐101, Arg‐103, and Tyr‐105, are observed in the FG‐loop of chain A of WT in the tolcapone‐bound state while three hydrogen bonds, in which Asn‐98 forms one H‐bond with residue Glu‐66, are detected in that of A97S (listed in Table S4). (c), (d) Cartoon and stick models of tolcapone‐bound WT (c) and A97S (d) showing the hydrogen‐bonding networks in the FG‐loop regions of chain B. There are eight H‐bonds, in which Asn‐98 forms four H‐bonds with residues Phe‐64, Gly‐101, Arg‐103, and Tyr‐105, identified in the structure of chain B of WT/tolcapone complex while five H‐bonds, in which Asn‐98 forms three H‐bonds with residues Glu‐66, Arg‐103, and Tyr‐105, are observed in that of A97S/tolcapone (listed in Table S4). (e), (f) Structural representation for the hydrogen‐bonding formation of FG‐loop residues of chain A in WT/diflunisal (e shown in pale green; PDB ID: 6E70) and A97S/diflunisal (f shown in orange) complex structures. H‐bonding residues are shown as sticks and connected by dashed green lines. Due to low quality of the electron density map in FG‐loop, a cartoon dashed line colored orange is used to present the region (D99–G102) in chain A of A97S/diflunisal complex. All hydrogen bonds are listed in Table S5. In the structure of chain A of WT/diflunisal complex, there are eight H‐bonds, in which Asn‐98 forms four H‐bonds with residues Glu‐66, Ser‐100, and Tyr‐105, identified while one H‐bond, in which Asn‐98 interacts with residue Ser‐97, is observed in that of A97S/diflunisal. (g), (h) Close‐up view of the intramolecular H‐bonding pairs with the residues of FG‐loop in chain B of diflunisal‐bound TTR complex of WT (g) and A97S (h). There are nine H‐bonds, in which Asn‐98 forms five H‐bonds with residues Glu‐66, Val‐65, Gly‐101, Arg‐103, and Tyr‐105, identified in the structure of chain B of WT/diflunisal complex while eight H‐bonds, in which Asn‐98 forms four H‐bonds with residues Glu‐66, Ser‐97, Gly‐101, and Tyr‐105, are observed in that of A97S/diflunisal. All hydrogen bonds are listed in Table S5, suggesting that stabilizers tolcapone and diflunisal binding to both A97S‐ and WT‐TTR cause the rearrangement of the intramolecular H‐bond networks in the FG‐loops of chains A and B of both WT and A97S, but the FG‐loops of tetrameric A97S are still more flexible than those of WT.

2.5. A97S mutation induces slight perturbation of the interactions between TTR and holo‐retinol binding proteins

Human TTR also plays a leading role in carrying holo‐retinol binding proteins in plasma and cerebrospinal fluid (Fex et al., 1979; White & Kelly, ²⁰⁰¹; Zanotti et al., ²⁰⁰⁸). Based on the crystal structure of TTR in complex with the holo‐retinol binding protein 4 (holo‐RBP4), the amino acid residues of the FG‐loop contribute to the interaction of TTR to holo‐RBP4 (Figure S7) (Monaco et al., 1995). To investigate how A97S affected the interaction of TTR with holo‐RBP4, we used nuclear magnetic resonance (NMR) spectroscopy and surface plasmon resonance (SPR) experiments to examine the binding event and measure the affinity. Upon addition of various concentrations of unlabeled holo‐RBP4, dramatic variations in the 2D spectral features of both A97S and WT were detected and the resonances underwent a prominent decrease in intensity (Figure 7a, b), where both A97S and WT were isotopically enriched. The results indicated that holo‐RBP4 interacted with both A97S and WT, and the linewidth of the bound‐state peaks became broadened and undetectable in the presence of binding events (Figure 7a, b). By assessing the attenuation of the NMR signal intensity, we concluded that the effects were slightly pronounced for WT, suggesting that A97S did not significantly perturb the overall structure of TTR complexed with holo‐RBP4. We further performed SPR experiments to determine the binding affinity of A97S for holo‐RBP4 (Figures 7c, d, S8, and Table 2), and the calculated average K _D values for RBP4 binding to WT and A97S were 0.50 ± 0.19 μM and 3.55 ± 0.88 μM, respectively, revealing that single mutation A97S caused an approximately five‐fold decrease in binding affinity of TTR for holo‐RBP4 and slightly weakened the holo‐RBP4/TTR complex interactions.

FIGURE 7.

The molecular interactions of TTR variants with holo‐RBP4 were analyzed by NMR and SPR. (a), (b) Overlay of 2D ¹H‐¹⁵ N TROSY‐HSQC spectra of 50 μM WT (a) and A97S (b) in the absence (colored in black) and presence of unlabeled holo‐RBP4 (25, 50, 75, and 125 μM; colored in blue, green, orange, and red, respectively) in the buffer containing 5% DMSO‐d₆ and D₂O recorded at 310 K. In the presence of unlabeled holo‐RBP4, the 2D spectral features of both A97S and WT undergo a prominent decrease in peak intensity, where A97S and WT are isotopically enriched. The holo‐RBP4 prominently interacts with both A97S and WT, respectively. (c), (d) Similar levels of WT‐TTR and A97S‐TTR immobilized onto the surface of channels 4 and 2, respectively, of CM5 sensor chip. A series of concentrations of holo‐RBP4 (0–1000 nM) simultaneously flowed through four channels. SPR sensorgrams (left panel) showing the binding responses of holo‐RBP4 to WT (c) and A97S (d) against time. Steady‐state fitting curves against the various concentrations of holo‐RBP4 are presented at the right panel and used to calculate the affinities of holo‐RBP4 with TTR variants (K _D = 0.47 ± 0.02 μM and 1.71 ± 0.08 μM for WT and A97S, respectively). Insets drew for the responses of concentrations ranging from 2.0 to 62.5 nM holo‐RBP4. About a five‐fold decrease in binding affinity of TTR for holo‐RBP4 is due to mutation A97S.

TABLE 2.

Affinity measurements of TTR variants interacting with holo‐RBP4 by SPR.

TTR variant	Ligand	Analyte	K D (μM)
WT	holo‐RBP4	TTR
No.1			0.41 ± 0.08
No.2			0.37 ± 0.07
No.3			0.72 ± 0.07
Average			0.50 ± 0.19
A97S	holo‐RBP4	TTR
No.1			3.13 ± 0.41
No.2			4.56 ± 0.88
No.3			2.97 ± 0.64
Average			3.55 ± 0.88

Note: All average values are the means ± standard deviations (SDs) of triplicate data.

2.6. Discussion

A97S variant is majorly identified in Taiwanese males and females with late‐onset neuropathy as well as cardiopathy (Lai et al., 2020). Compared with WT, A97S has worse resistivity against acid treatment and shows a higher propensity for amyloid fibril formation, indicating that the tetramer instability of A97S is similar to other amyloidogenic mutants, such as V30M (Figure 2). Indeed, the tetramer destabilization of A97S and V30M in the apo state demonstrates quite a similar tendency by acid‐mediated denaturation although A97S has slightly better resistivity against chemical and thermal denaturation than V30M at neutral pH (Liu et al., 2019). Interestingly, compared with V30M, A97S is more effectively stabilized by native‐state stabilizers, such as T4, tolcapone, and tafamidis (Liu et al., 2019; Sant'Anna et al., ²⁰¹⁶). By comparing the expression and purification of TTR variants in Escherichia coli (E. coli) system, the yield of A97S and WT is pronouncedly better than that of V30M, which develops more soluble and insoluble aggregates than A97S and WT, suggesting that V30M is more difficult to form correct folding than A97S and WT. These data might help explain that V30M is early‐onset variant contrary to the late‐onset A97S while V30M and A97S show close stability in acid environments in vitro. However, the rising number of late‐onset V30M cases has been reported, but we cannot clearly explain why V30M patients show different age of disease onset (Pinto et al., 2019; Waddington‐Cruz et al., ²⁰²¹). It has been hypothesized that the age‐ and symptom‐related variability among TTR variants may be ascribed to the environmental and genetic factors (Pinto et al., 2019; Waddington‐Cruz et al., ²⁰²¹), thus resulting in differentials in age of onset and clinical phenotypes of V30M patients.

Some TTR variants cannot rapidly form fibrils under physiological conditions but are sensitive to acidic pH. Therefore, acid‐mediated aggregation has been applied in most studies of TTR variants. To investigate whether pH fluctuations affect the FG‐loop of A97S, we also have determined apo A97S structure at pH 7.6 (PDB ID: 8HY4, Table S1, S6, and Figure S9) compared with that at slightly acidic pH (5.3–5.5), revealing that the highly flexible FG‐loop of A97S is inherent and pH‐independent in the range of pH 7 to 5. Moreover, we have included ¹H‐¹⁵ N TROSY HSQC NMR spectra of WT and A97S recorded at pH 7 to 5 (Figure S10a and b), and the observed cross‐peaks are from the major state populations of TTR variants, which adopt the tetrameric conformations, while the minor state populations with monomeric or intermediate species cannot be directly detected. We find that the chemical shift difference between WT and A97S at pH 7 is highly similar to that at pH 5 (Figure S10c and d). Therefore, the stability of A97S‐ and WT‐TTR at pH 5 is examined by recording NMR spectra at 47°C along time. The intensity of apo A97S declined over time when compared with that of WT, suggesting the major conformation of WT is more stable than that of A97S at pH 5 (Figure S11). Furthermore, we also characterize the tetramer destabilization of A97S by S‐Trap assay and calcium treatment at neutral pH (7.3), suggesting that the effects of SDS and Ca²⁺on A97S are substantially stronger than those on WT protein at physiological pH (Figure 4). Collectively, A97S mutation destabilizes TTR by increasing the flexibility of the FG‐loop in the monomer, thus changing the rate of TTR fibril formation.

Compared with tetramer destabilization and aggregation induced by the increased FG‐loop flexibility of A97S substitution, several TTR variants have been reported to perturb the flexibility of BC‐, CD‐, and DE‐loop regions, resulting in decreased dissociation barrier of tetrameric assembly (Dasari et al., 2020; Esperante et al., ²⁰²¹; Klimtchuk et al., ²⁰¹⁸; Lim et al., ²⁰¹³; Mangione et al., ²⁰¹⁴). Interestingly, R34 and K35 variants make a prominent effect at the BC‐loop (residues 36–40) locally, leading to the destabilization of the dimer–dimer interface allosterically and therefore the connection between the flexible degree of structural conformers and the tendency toward amyloid formation (Esperante et al., 2021). The stability of CD‐loop (residues 49–53) is under the influence of destabilizing mutations, such as S52P, G53A, and an unusual duplication mutation (Glu‐51_Ser‐52 dup), bringing about the formation and accumulation of amyloid fibrils (Dasari et al., 2020; Klimtchuk et al., ²⁰¹⁸; Mangione et al., ²⁰¹⁴). Substitutions on residues of DE‐loop (residues 56–66) for positively charged amino acids (L58H, L58R, and T59K) likely evoke the exposure of hydrogen bonding networks, thus resulting in partial destabilization and promoting transient dissociation (Takeuchi et al., 2007). Overall, the increased flexibility of BC‐, CD‐, DE‐ and FG‐loop regions is remarkably correlated with the aggregation‐prone degree of disease‐associated TTR substitutions.

The binding of native‐state stabilizer to TTR variants greatly increases stability of tetrameric assembly against acid‐mediated dissociation that promotes amyloid fibril formation. Based on the structures of both A97S and WT in complex with tolcapone and diflunisal, we conclude that the FG‐loop of A97S still exhibits inherent conformational flexibility (Figure S6) while the extensive networks of intramolecular H‐bonding in the FG‐loop of WT are partially organized due to the binding event (Figure 6, Tables S4 and S5). The calorimetrically measured large enthalpy entirely drives the binding of tolcapone to A97S and WT (Figure 5 and Table 1), indicating the specific interactions of tolcapone to A97S. However, the flexibility of A97S FG‐loop in the apo and stabilizer‐bound states is relatively similar, thus being advantageous for preventing unfavorable entropy changes from affecting the binding affinity between A97S and stabilizers.

Previous studies reported that the binding affinity between human holo‐RBP4 and WT‐TTR is about 0.35 μM, which is quite close to our SPR data (K _d ∼ 0.35 μM compared with 0.50 μM, respectively) (Zanotti et al., 2008). V20S and I84A variants of TTR seriously impair the protein–protein interactions while both mutations D99A and S100E located at the FG‐loop of TTR result in a 20‐fold decrease in the binding affinity of TTR for holo‐RBP4, confirming that the amino acid residues of the FG‐loop participate in the macromolecular interactions between TTR and holo‐RBP4. By analyzing the complex structure of holo‐RBP4‐TTR, (Monaco et al., 1995; Zanotti et al., ²⁰⁰⁸) we find that Ala‐97 does not directly mediate the interactions with holo‐RBP4, but mutation A97S elicits the dynamic changes on the FG‐loop, thus bringing about a five‐fold difference in binding affinity for holo‐RBP4.

In conclusion, we report crystal structure of A97S‐TTR determined at 1.38 $\mathrm{\AA }$ resolution, revealing that this amyloidogenic point mutation shares almost identical core structure to WT‐TTR except that the flexibility of the FG‐loops in both chains A and B is largely affected by a change of nonpolar Ala to a polar Ser at position 97. Asn‐98 plays an essential role in stabilization of the FG‐loops by mediating the H‐bonding networks in both chains A and B. A97S substitution significantly exerts influence on positioning of both main‐ and side chain of Asn‐98, therefore, leading to inherent flexibility of the FG‐loops of A97S. In contrast to WT‐TTR, A97S mutation shows decreased conformational stability and increased propensity to form monomer, thus resulting in the enhanced amyloidogenicity in its apo state. As measured by calorimetry, the modes of both tolcapone and diflunisal binding to A97S‐TTR are similar to those to WT, and the tetrameric population of A97S‐TTR is increased in the presence of native‐state stabilizers, suggesting that A97S mutation do not interfere with T4 binding in the physiological condition. Moreover, A97S mutation introduces a little influence on the RBP4‐TTR complex with an approximately five‐fold decrease in binding affinity. Collectively, our study reveals that tetramer dissociation of A97S‐TTR is altered by the dramatic alteration of the FG‐loop, giving rise to destabilization of native tetrameric structure and thus promoting amyloidogenesis.

3. MATERIALS AND METHODS

3.1. Expression and purification of recombinant human TTR variants

Recombinant human A97S‐TTR was cloned into pET21a plasmid (termed pET21‐A97S henceforth), containing a C‐terminal tag with residues LEHHHHHH. WT‐TTR and two variants (V30M and R104H) were generated by the QuikChange® kit (Stratagene). The plasmids encoding TTR variants were transformed into BL21 (DE3) strain, and the transformed cells were cultured in Luria Bertani (LB) medium at 37°C in the presence of ampicillin. The expressions of recombinant proteins were induced at 25°C overnight with addition of 0.5 mM isopropyl β‐D‐1‐thiogalactopyranoside (IPTG) at OD₆₀₀ ~ 0.7–0.8. These E. coli cells were harvested by centrifugation at 5000 $\times$ g for 45 min at 4°C, and the pellets were resuspended in lysis buffer (25 mM NaPi, pH 8.5, 500 mM NaCl, 5 mM β‐mercaptoethanol [β‐ME], 5% glycerol) and lysed by EmulsiFlex‐C3 high‐pressure homogenizer (Avestin). The supernatants were collected after centrifugation of cell lysates at 50,000 $\times$ g and purified by Ni Sepharose high‐performance affinity resin (GE Healthcare). The nickel resin was thoroughly washed with lysis buffer followed by 5, 10, 20, and 30 mM imidazole in lysis buffer; the recombinant proteins were eluted out of the column by using 400 mM imidazole in lysis buffer. These eluted proteins were concentrated and further purified via HiLoad 16/600 Superdex 200 column (GE Healthcare) equilibrated in FPLC buffer (50 mM KPi, pH 7.0, 100 mM KCl, 1 mM Ethylenediaminetetraacetic acid [EDTA]). Tetrameric fractions of TTR variants were checked by running 15% SDS‐PAGE. Tetrameric protein concentrations were calculated by detecting A₂₈₀ using a nanophotometer NP 80 (Implen) with the extinction coefficient (ε) of tetramer = 18,450 $\times$ 4 = 73,800 M⁻¹ cm⁻¹.

3.2. Preparation of recombinant human holo‐RBP4

The construction of pET28m plasmid containing human gene encoding mature form of RBP4 (from residue 19 to residue 201; pET28m‐RBP4) was entrusted to the First Core Laboratory, National Taiwan University College of Medicine. BL21 (DE3) of E. coli system was used to overexpress and purify recombinant RBP4 by following published protocol (Wang et al., 1993). After homogenization and centrifugation, RBP4 protein was mainly insoluble in pellet fraction. Denaturing buffer containing 8 M urea in lysis buffer was applied to resuspend and denature RBP4 in the inclusion bodies. Room temperature (RT) Ni‐affinity chromatography was utilized to purify crude extract of RBP4 in denaturing buffer. The denatured RBP4 was eluted, concentrated, and then pipetted into a pre‐cooling steel tank containing 50 mM Tris buffer (pH 8.5), 1 mM reduced glutathione (GSH), 0.1 mM oxidized glutathione, 0.2 mM retinol, 10 mM EDTA, and 2.5% DMSO (Xie et al., 1998). After overnight refolding in a cold room, the holo‐hRBP4 protein solution was further purified with HiLoad 16/600 Superdex 75 pg (GE Healthcare). The peak fractions of holo‐RBP4 were collected, and protein concentration was determined according to the absorbance of 280 nm with the extinction coefficient (ε) = 34,295 M⁻¹ cm⁻¹.

3.3. Crystallization and structure determination

A97S‐TTR protein was concentrated to 10 mg/mL and crystallized at 4°C by the hanging‐drop vapor diffusion method using 1 μL of protein solution and 1 μL of precipitation solution (10 mM citrate buffer, pH 5.3–5.5, 1.5–2.0 M ammonium sulfate) against 200 μL of reservoir. Single 3D crystals suitable for data collection would appear in 2 days. To obtain A97S‐TTR/diflunisal complex crystals, apo‐crystals of A97S‐TTR were soaked into a 2 μL drop containing 10 mM citrate buffer, pH 5.5, 1.5 M ammonium sulfate with diflunisal at a final concentration of 800 μM at 4°C overnight. For A97S‐TTR/tolcapone co‐crystals, 13 mg/mL A97S‐TTR was mixed with tolcapone in a molar ratio of 1:11, preincubated at RT for 2 h, and further co‐crystallized via hanging‐drop vapor diffusion under 100 mM HEPES, pH 7.0, 200 mM calcium chloride, 20% PEG 400 at 4°C (Sant'Anna et al., 2016). The apo‐form crystals of A97S at neutral pH (pH 7.6) were also obtained by growing at 4°C in the hanging droplets containing 1:1 volume of A97S protein solution (10 mg/mL in 20 mM Tris, 150 mM NaCl, pH 8.0) and mother liquor (0.1 M HEPES, pH 7.6, 38% PEG 400, and 0.4 M CaCl₂). All crystals were cryo‐protected with 20% glycerol in mother liquid and frozen by liquid nitrogen at 100 K. X‐ray diffraction data of apo‐A97S‐TTR, A97S‐TTR/diflunisal complex, and A97S‐TTR/tolcapone complex were collected at beamline BL13B1, 13C1, and 15A1 of TLS of National Synchrotron Radiation Research Center in Hsinchu (Taiwan). The data were processed by HKL‐2000, and the phases were solved by molecular replacement based on the deposited WT‐TTR model (ID: 2QGB) as starting template. Atomic models were improved in Coot and further refined using Phenix. eLBOW and LigandFit programs in Phenix were used to generate ligand restraints files and fit ligands to the difference map.

3.4. Structural analysis

The model of apo‐structure of A97S‐TTR was uploaded to STRIDE Web interface (http://webclu.bio.wzw.tum.de/cgi-bin/stride/stridecgi.py) for secondary structure assignment (Frishman & Argos, 1995).

PDBePISA (https://www.ebi.ac.uk/pdbe/pisa/) was utilized for interaction analysis between chains A, A′, B, and B′ interfaces of both WT‐ and A97S‐TTR structures.(Krissinel & Henrick, 2007) Hydrogen Bond Calculation (http://cib.cf.ocha.ac.jp/bitool/HBOND/) platform was applied to assess the intramolecular hydrogen bond networks of FG‐ loop in both chains A and B of WT‐ and A97S‐TTR in the apo‐ and stabilizer‐bound states. The coordinate files of apo‐WT‐TTR (ID: 2QGB), WT‐TTR/tolcapone (ID: 4D7B), and WT‐TTR/diflunisal (ID: 6E70) structures were fetched from the database of Protein Data Bank.

3.5. Acid‐induced aggregation assay

TTR aggregation was induced by acidic incubation as previously described with a modified protocol (Groenning et al., 2015; Reixach et al., ²⁰⁰⁴). The purified WT and variants (A97S, V30M, and R104H) of TTR were diluted to 181 μM with FPLC buffer, and pre‐treated with final 1% dimethyl sulfoxide (DMSO) as control or molar ratios of 1:0.5 or 1:1 of TTR stabilizers, including T₄, tolcapone, and diflunisal. The solutions were further diluted with the volume ratio 9:1 with acidic buffer (1 M acetic acid, 2 M NaCl) to adjust pH value to 4.0 at a final protein concentration of 163 μM and placed in the incubator at 37°C for 1, 2, 3, and 4 days. After 1–4‐day acid‐incubation time, the samples were centrifugated to precipitate aggregates at 13,200 rpm for 15 min at 4°C, and soluble TTR concentrations of the supernatants were determined by the absorbance of 280 nm. The relative residual level of soluble TTR was represented by the ratios of [TTR]t/[TTR]t ₀, where [TTR]t is TTR concentration at the specific acid‐incubation time, and [TTR]t ₀ is the starting constant protein concentration.

3.6. CR binding assay

To test amyloid fibril formation of TTR variants trigged by acid environment, amyloid‐specific CR was utilized to detect the level of amyloid as a described method with minor modification (White & Kelly, 2001). Preparation of TTR amyloid fibrils was performed by following the protocol of acid‐induced aggregation assay described above. TTR variants (3.6 μM) in buffer, which contained 0.1 M acetic acid and 0.2 M NaCl, were incubated at 37°C for 3 days. The samples were resuspended by vortexing and 10 μM CR dye in FPLC buffer was added into the suspension of TTR variants (50 μL). Spectrophotometer DU730 was operated to scan the spectra from 400 to 600 nm, and quantification of fibril was represented as μmoles of CR bound fibrils over a liter of amyloid suspension, and calculated as the following equation: [(A₅₄₀/25295) − (A₄₇₇/46306)] × 1000,000 (Lai et al., 1996).

3.7. Transmission electron microscopy

In order to observe the formed fibrils under low pH conditions, the suspensions of WT‐TTR and A97S‐TTR were collected from acid‐induced aggregation assay (pH 4.0) and turbidity assay (pH 5.0). For acquisition of EM data at pH 5.25, 18 μM WT and A97S were pre‐incubated in citrate–phosphate buffer (pH 5.25) at 37°C for 6 days. The samples (15 μL) were applied onto a glow‐discharged copper 200‐mesh grid coated with carbon/formvar film. After washing with distilled water and negative staining with 1% uranyl acetate, the grids were visualized using an FEI Tecnai T12 electron microscope (ThermoFisher).

3.8. Turbidity assay

WT‐ and A97S‐TTR proteins (3.6 μM) were prepared in citrate–phosphate buffer containing 100 mM KCl and 1 mM EDTA at pH 7.0, 6.5, 6.0, 5.5, 5.0, 4.5, and 4.0, respectively, and then incubated at 37°C for 1, 2, 3, 7, 8, 9, and 10 days in a 96‐well microplate. The samples were vortexed for 5 s, and the turbidity at 330 nm was measured on the Spectra Max i3x multimode microplate reader (Molecular Devices).

3.9. Evaluating pH‐mediated quaternary structural changes

Two microliters of TTR stock solutions (400 μM) in the FPLC buffer (pH 7.0) were diluted with 198 μL of citrate–phosphate buffer (pH 7.0–4.0), and incubated at 37°C for 1 and 3 h, respectively.

The samples (20 μL) were mixed with 4 μL 6 × loading buffer (0.35 M Tris at pH 6.8, 30% glycerol, 20 mM β‐ME, 0.012% bromophenol blue [BPB]) in the absence of SDS. The mixtures (8 μL) without additional boiling were loaded into 11% acrylamide gel and ran at a constant current of 100 mA for analysis (Robinson & Reixach, 2014). The gels were soaked in Easy Blue‐Plus CBB reagent (EBL Biotechnology) for staining overnight and destained with water with rocking at 60 rpm at RT for 2 h. The intensities of monomer and dimer bands were quantified by image processing in NIH‐developed ImageJ program. The fraction of monomer at a specific pH value was determined by the intensity ratios of monomer band over the sum of dimeric and monomeric bands and plotted against pH values.

3.10. pH‐jump assay

In the absence and presence of a 1 molar ratio of TTR stabilizers, the supernatants of TTR variants in acid‐induced aggregation buffer with one‐day incubation were collected and diluted with FPLC buffer (pH 7.0) to neutralize pH from amyloidogenic pH 4.0 to pH 7.0. The neutralized samples were centrifuged at 13,200 rpm until no precipitation and subsequently loaded into Superdex 200 increase 10/300 on AKTA purifier 10 system (GE Healthcare). The areas of tetrameric peaks were integrated by internal UNICORN program. Relative tetramer ratios were quantified by integrating the areas of tetrameric TTR variants and divided by those of the WT group.

3.11. S‐Trap assay

SDS‐traping assay was used to investigate resistance of TTR variants under denaturing environment (Wieczorek et al., 2021). Total 5 μM protein solutions of WT‐ and A97S‐TTR in Tris buffer (50 mM Tris, 100 mM NaCl, pH 7.3) were prepared and incubated at RT for 30 min. Twenty microliters of aliquots were added with five microliters of preheated (95°C) 5 × SDS loading buffer dye consisting of 300 mM Tris (pH 8.8), 30% glycerol, 5% SDS, and 0.05% BPB. The samples underwent thermal denaturation at 85°C for various time points (0, 15, 30, 60, 120, 180, 300, and 600 s), and cooled in ice bath for 5 s after each time point for stopping unfolding reactions. Further RT spinning down was performed via mini centrifuge for 5 s, and 8 μL samples were subsequently analyzed by running seminative PAGE of 11% acrylamide/bis‐acrylamide (29:1) gels containing 0.1% SDS. The gels were stained overnight with EBL Easy Blue‐Plus CBB, and quantification of the intensities of monomer bands was carried out by the ImageJ software (NIH). The unfolding ratios of TTR variants by determining intensity ratios of monomer bands relative to the sum of dimer and monomer bands.

3.12. Evaluating calcium‐mediated quaternary structural changes

To avoid co‐precipitation of calcium with phosphate of FPLC buffer, TTR protein solutions were exchanged into the same Tris buffer of S‐Trap assay by spin column. TTR variants (5 μM) were incubated with calcium chloride (0, 2, 5, 10, 20, 50, and 100 mM) at 4°C for 16 h, and analyzed the destabilizing effect of a calcium ion by running seminative PAGE (Wieczorek et al., 2021). After centrifugation at RT for 5 s with a mini bench‐top centrifuge, 5 μL of the 5 × SDS loading buffer dye was added into the tubes. A total of 8 μL of the samples without heating were loaded into the wells of 11% acrylamide gel, and EBL Easy Blue‐Plus CBB dye was utilized for staining overnight. The fraction of monomeric TTR variants at a specific  concentration of calcium chloride was determined by the intensity ratios of monomer band over the sum of dimeric and monomeric bands, and plotted against different concentrations of calcium chloride.

3.13. Isothermal titration calorimetry (ITC) measurements

ITC experiments were performed to analyze the interactions between stabilizers and TTR variants by iTC200 machine (MicroCal). Stabilizers tolcapone and diflunisal were diluted to 100 μM in FPLC buffer containing 2.5% DMSO, and loaded into the injection syringe (Sant'Anna et al., 2016). Stabilizers (100 μM) were titrated into TTR variants (5 μM). The titration experiments were programmed with a 0.5 μL initial injection followed by 19 injections of 2 μL, and underwent with 120 s duration for equilibration at 25°C. The raw experimental data of thermograms were used to integrate the thermal change of differential power of each injection peak, and the data points were analyzed by Malvern MicroCal PEAQ‐ITC Analysis Software. Data points of tolcapone binding isotherms were fitted by two‐site model while single‐site binding model was applied to that of diflunisal binding.

3.14. Nuclear magnetic resonance (NMR) spectroscopy experiments

Isotopically labeled A97S‐ and WT‐TTR samples for NMR experiments were prepared using the minimal (M9) medium. For characterization of the interactions between TTR variants and holo‐RBP4, 50 μM [U‐¹⁵N]‐labeled protonated TTR variants in the FPLC buffer containing 5% DMSO‐d6 and 5% D₂O were titrated with 25, 50, 75, and 125 μM unlabeled protonated holo‐RBP4, and the 2D ¹H‐¹⁵N TROSY‐HSQC spectra were recorded on Bruker 800‐MHz spectrometers (Bruker BioSpin, Karlsruhe, Germany) at 310 K. ¹⁵N‐labeled TTR variants were exchanged to different pH buffer conditions by the zeba™ 7 K MWCO spin desalting column (Thermo), and the NMR samples (50 μM TTR variants) were prepared and supplemented with final 5% D₂O. All two‐dimensional ¹H‐¹⁵N TROSY‐HSQC NMR spectra were acquired at 320 K on Bruker 800‐MHz spectrometers (Bruker BioSpin, Karlsruhe, Germany). The sequence‐specific backbone resonance assignments of WT‐ and A97S‐TTR were derived from previously published assignments (BMRB ID: 27515, 27575, and 27576), and adjusted to the centers of peaks at different pH conditions in our study (Leach et al., 2018; Liu et al., ²⁰¹⁹). The chemical shift perturbation for differences between pH 5.0 and pH 7.0 were calculated by the following equation:  $∆{\delta }_{\left(\mathrm{HN},\mathrm{N}\right)}$ = $\sqrt{{\left({\omega }_{\mathrm{HN}}∆{\delta }_{\mathrm{HN}}\right)}^2+{\left({\omega }_{\mathrm{N}}∆{\delta }_{\mathrm{N}}\right)}^2}$, where ${\omega }_{\mathrm{HN}}$=1 and ${\omega }_{\mathrm{N}}$=0.154.

3.15. Surface plasmon resonance

All SPR measurements were conducted on a Biacore T200 apparatus (GE Healthcare) with constant temperature control at 25°C. The ligand solutions of 10 μg/mL WT‐ and A97S‐TTR were, respectively, prepared in 10 mM sodium acetate buffer at pH 5.0. They were covalently immobilized onto the CM5 sensory chip surface of flow channel 2 (A97S‐TTR) and channel 4 (WT‐TTR) via the suggested amine coupling procedure. The final immobilized levels of ligands (R _L) were achieved to 1300–1700 RU. The running buffer contained 10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.003% Tween‐20, 1% DMSO, and 5 μM retinol. Buffer exchange of analyte holo‐RBP4 was freshly prepared by Thermo zeba™ spin desalting column (7 K MWCO, 2 mL). A series of increasing analyte concentrations ranging from 0 to 1000 nM were sequentially diluted in the running buffer and programmed to simultaneously flow over four channels at a flow rate of 30 μL/min. Each running cycle was set with 120 s for contact time and 180 s for dissociation, and regeneration of the chip surface was fulfilled by loading 4 M MgCl₂. Channels 1 and 3 were defined as blank cells for A97S‐TTR and WT‐TTR, respectively, and subtracted from the result responses of channels 2 and 4, respectively. The sensorgrams for surface‐bound holo‐RBP4 with various concentrations were plotted by the resonance units against time. The Biacore T200 evaluation software was utilized to execute steady‐state analysis for determination of the binding affinities of WT‐TTR and A97S‐TTR in complex with holo‐RBP4. For reverse SPR analysis using holo‐RBP4 as ligand, 25 μg/mL holo‐RBP4 was prepared in 10 mM sodium acetate buffer at pH 5.0, and covalently immobilized onto a new CM5 sensory chip surface of channel 4 via the suggested amine coupling procedure. The final immobilized level of ligand was achieved at 994.3 RU. Sample preparation, buffer condition, and experimental parameters were carried out as previously mentioned above.

AUTHOR CONTRIBUTIONS

Yi‐Shiang Wang: Data curation (lead); formal analysis (lead); methodology (lead); writing – original draft (lead); writing – review and editing (lead). Chun‐Hsiang Huang: Data curation (lead); formal analysis (equal). Gunn‐Guang Liou: Data curation (lead); formal analysis (equal); writing – review and editing (supporting). Hsueh‐Wen Hsueh: Conceptualization (supporting); resources (lead). Chi‐Ting Liang: Data curation (lead); formal analysis (supporting). Hsi‐Ching Tseng: Data curation (lead); resources (supporting). Shing‐Jong Huang: Data curation (lead); formal analysis (supporting); resources (supporting). Chi‐Chao Chao: Conceptualization (lead); data curation (supporting); resources (lead). Sung‐Tsang Hsieh: Conceptualization (lead); investigation (supporting); project administration (supporting); resources (lead); writing – original draft (supporting); writing – review and editing (supporting). SHIOU‐RU TZENG: Conceptualization (lead); data curation (lead); formal analysis (supporting); funding acquisition (lead); investigation (lead); methodology (supporting); project administration (lead); resources (lead); supervision (lead); validation (lead); writing – original draft (lead); writing – review and editing (lead).

Supporting information

Data S1: Supporting Information

Wang Y‐S, Huang C‐H, Liou G‐G, Hsueh H‐W, Liang C‐T, Tseng H‐C, et al. A molecular basis for tetramer destabilization and aggregation of transthyretin Ala97Ser . Protein Science. 2023;32(4):e4610. 10.1002/pro.4610