Computational Insights into the Aggregation Mechanism of Human Calcitonin

Abstract

Human calcitonin (hCT) is a peptide hormone that regulates calcium homeostasis, but its abnormal aggregation can disrupt physiological functions and increase the risk of medullary thyroid carcinoma. To elucidate the mechanisms underlying hCT aggregation, we investigated the self-assembly dynamics of hCT segments (hCT_1–14, hCT_15–25, and hCT_26–32) and the folding and dimerization of full-length hCT_1–32 through microsecond atomistic discrete molecular dynamics (DMD) simulations. Our results revealed that hCT_1–14 and hCT_26–32 predominantly existed as isolated monomers with transient small-sized oligomers, indicating weak aggregation tendencies. In contrast, hCT_15–25 exhibited robust aggregation capability, forming stable β-sheet aggregates independently. Full-length hCT_1–32 monomers displayed dynamic helical structures, with dimerization decreasing helix content and enhancing β-sheet formation. The transition to β-sheets in full-length hCT_1–32 correlated with the loss of helical structure in the hCT_15–25 region. Conformations with high helical content in hCT_15–25 corresponded to significantly reduced β-sheet structures across the peptide, underscoring the importance of helical stability in preventing β-sheet conversion. Thus, the development of amyloid-resistant hCT analogues should focus on enhancing helical stability in this crucial region. Overall, our study not only elucidates the aggregation mechanism of hCT but also identifies a critical target for designing drug inhibitors to prevent hCT aggregation.

Graphical Abstract:

1. Introduction.

Human calcitonin (hCT) is a 32-amino acid hormone peptide primarily secreted by the parafollicular cells (C cells) of the thyroid gland[1, 2]. It plays a crucial role in calcium homeostasis by regulating blood levels of calcium and phosphate[3]. The primary physiological function of hCT is to inhibit osteoclast activity, thereby reducing bone resorption and promoting calcium deposition in the bones, which helps maintain bone density and prevent conditions like osteoporosis[4, 5]. During pregnancy and lactation, hCT levels increase to regulate calcium ion levels in the mother’s bloodstream, ensuring an adequate supply for fetal development and milk production[6, 7]. These functions have significant implications in the clinical treatment of bone disorders such as osteoporosis and Paget’s disease[8, 9], particularly in postmenopausal women[10], by increasing bone density and reducing the risk of fractures. However, the efficacy of hCT-based treatments is hampered by its intrinsic tendency to aggregate in solution[11], leading to structural reformatting and sequestration, which diminishes the bioavailability and effectiveness of the therapies[9, 12]. Additionally, abnormal aggregation of hCT can result in the formation of cytotoxic aggregates, causing abnormal cell death[13]. Furthermore, amyloid deposits of hCT have been identified in individuals with medullary thyroid carcinoma (MTC), suggesting a possible association between MTC and hCT aggregation[14]. Therefore, the clinical application of hCT has been discontinued by the Food and Drug Administration (FDA)[15, 16].

The therapeutic potential of hCT is hindered by its strong propensity to form amyloid fibrils[11, 15], akin to peptides implicated in degenerative diseases[17, 18] such as amyloid-β (Aβ) and tau in Alzheimer’s disease (AD)[19–21], α-synuclein in Parkinson’s disease (PD)[22, 23], and human islet amyloid polypeptide (hIAPP) in type II diabetes (T2D)[24, 25]. Salmon calcitonin (sCT), though possessing lower aggregation propensity, shares only 50% homology with hCT and is thus clinically used[26, 27]. However, sCT exhibits reduced potency and necessitates frequent injections due to its short half-life (~1 h)[28, 29]. Additionally, sCT administration may trigger immunogenic reactions like anorexia and vomiting and potentially lead to clinical resistance from antibody development in patients[30, 31]. Interestingly, prior research has revealed that hCT demonstrates significantly greater potency than sCT in the absence of fibrillation[32]. To address this, one approach involves designing non-amyloidogenic hCT analogues that closely resemble hCT in terms of physicochemical properties and sequence, aiming to preserve therapeutic efficacy while mitigating amyloid fibril formation[27]. Several hCT analogues with reduced aggregation propensity have been developed through targeted substitutions of amyloidogenic residues. Examples include phCT[27], incorporating substitutions Y12L, N17H, A26N, I27T, and A31T; TL-hCT[33], featuring changes Y12L, F16L, and F19L; and DM-hCT[34, 35], with double mutations Y12L and N17H. Another strategy entails developing biocompatible inhibitors to counter hCT amyloidosis[36]. Previous studies have identified small molecules such as EGCG[36], cucurbit[7]uril[28], magnolol[16], and honokiol[16] as potential inhibitors capable of suppressing hCT aggregation and amyloid deposition. Despite advancements in the design of hCT analogues and amyloid inhibitors, clinical use remains limited due to the incomplete understanding of the aggregation mechanism of hCT, particularly the critical residues that drive aggregation and the process by which they convert into β-sheet structures.

The monomers of hCT are highly dynamic, featuring a central helix region, a loop structure at the N-terminus, and a random coil C-terminus, as determined by Circular Dichroism (CD) and Nuclear Magnetic Resonance (NMR) spectroscopy[27, 36]. The fibrillization of hCT follows a two-step process: hCT monomers first accumulate into helical oligomers and subsequently undergo a conformational conversion towards β-sheet structures[1, 12, 37, 38]. For instance, during aggregation, a helix-to-β-sheet conversion is observed around the central region of ¹⁵DFNKF¹⁹[33, 37, 38]. This segment readily forms fibrils, highlighting its pivotal role in hCT aggregation and suggesting it as the primary amyloidogenic core[39, 40]. Additionally, β-sheet formation is also found around residues G10, F22, and A26 in hCT aggregates[33]. Substitutions such as Y12L, F16L, and F19L in TL-hCT[33], and Y12L and N17H in DM-hCT[35], suppress hCT aggregation, indicating that regions beyond the central segment, including Y12, also participate in the aggregation process[41, 42]. For example, nitration of Y12[13] or phosphorylation of T13[43] decreases the aggregation propensity of hCT. This suggests a second amyloidogenic core, residues ⁶TCMLGT¹¹, although the relevance of this fragment remains debated[44]. The phCT analogue, with substitutions Y12L, N17H, A26N, I27T, and A31T, exhibits a much weaker aggregation tendency, indicating that the C-terminus also plays a role in hCT fibrillization[27]. Our prior computational studies suggest that the formation of β-sheets within hCT aggregates involves residues 8–31, with a notable presence around residues 16–25[41, 42, 45]. Despite extensive research on hCT aggregation, the critical residues driving hCT self-assembly into β-sheet aggregates and the synergistic effects between different regions on their conformational dynamics remain incompletely understood. This knowledge is crucial for developing amyloid-resistant hCT analogs for clinical use.

To systematically investigate the aggregation mechanism of hCT at the molecular level, we conducted multiple independent microsecond atomistic discrete molecular dynamics (DMD) simulations for the self-assembly of hCT_1–14, hCT_15–25, and hCT_26–32 segments, as well as the folding and dimerization of the full-length hCT_1–32. As is well known, the formation of β-sheet regions typically plays a critical role in driving amyloid peptide aggregation[22, 46–48]. Therefore, the full-length hCT_1–32 peptide was divided into three segments (hCT_1–14, hCT_15–25, and hCT_26–32) at residues corresponding to β-sheet breaker points according to prior hCT conformation studies[41, 42, 45]. Our results showed that the aggregation tendency of hCT_1–14 and hCT_26–32 was extremely weak, primarily existing as isolated unstructured monomers and only transiently forming small-sized oligomers. In contrast, the hCT_15–25 segment exhibited significant aggregation capability and readily self-assembled into stable β-sheet aggregates. Full-length hCT_1–32 monomers mainly adopted unstructured conformations along with dynamic helix structures around residues 4–9 and 14–19, although the helical tendency of the former was much weaker. Dimerization of full-length hCT_1–32 led to a decrease in helicity around residues 14–19 and an enhancement of β-sheet formation for residues 8–13, 18–24, and 26–31. The formation of β-sheets in the entire hCT_1–32 peptide correlated with the loss of the helical structure in the central region. Conformations with high helical content around hCT_15–25 exhibited extremely weak β-sheet structures throughout the entire peptide, indicating that maintaining the stability of the central helix around hCT_15–25 could prevent the full-length hCT_1–32 peptide from converting to β-sheets. Therefore, designing amyloid-resistant hCT analogues and amyloid aggregation inhibitors for hCT should focus on enhancing the helical stability around hCT_15–25, as this region not only exhibited the strongest aggregation tendency but also significantly influenced the overall peptide structure.

2. Materials and Methods

2.1. Molecular Systems.

The amino acid sequence of full-length hCT (referred to as hCT_1–32) employed in our simulation is ¹CGNLSTCMLG ¹¹TYTQDFNKFH ²¹TFPQTAIGVG ³¹AP. Notably, residues C1 and C7 are linked by an intra-peptide disulfide bond, a characteristic that is conserved in all known calcitonin sequences[12]. To identify the specific roles of different regions in driving hCT amyloid aggregation, the full-length hCT was divided into three segments: hCT_1–14, hCT_15–25, and hCT_26–32. This division is based on our previous computational simulations, which indicated that the local β-sheet structures end around residues Q14 and T25 in both hCT monomers and oligomers[41, 42, 45]. The initial structure of hCT was derived using PyMol mutagenesis based on the structure of phCT (PDB id: 2jxz)[27], given the unavailability of an experimentally characterized structure of hCT. Subsequently, the initial structure of each segment was extracted from the constructed full-length hCT_1–32 monomer. For every segment, forty independent DMD simulations were conducted, with up to ten simulated peptides per simulation, each extending over 1000.0 ns. To mitigate potential biases arising from identical initial states, ten peptides were randomly positioned within a 9.0 nm cubic simulation box, ensuring diverse inter-peptide distances and orientations for each run. Additionally, the minimum distance between any pair of peptides in each initial state was maintained at no less than 1.5 nm, surpassing the cutoff for non-bonded interactions.

Besides examining the self-assembly conformational dynamics of each segment, the folding and dimerization dynamics of one and two isolated full-length hCT_1–32 peptides were also scrutinized to elucidate the impact of inter-segment interactions on the structure and aggregation of hCT. The conserved intra-peptide C1-C7 disulfide bond, found in all known calcitonin family proteins[12], was explicitly modeled in both the segment self-assembly and full-length folding and dimerization simulations of hCT. For every molecular system, forty independent DMD simulations were conducted, each spanning 1000.0 ns and commencing with different initial velocities. To ensure thorough conformational sampling and minimize potential bias from identical initial states, we generated 40 different initial configurations for the dimerization simulations of hCT, each with an intermolecular distance greater than 1.5 nm. Specifically, two hCT molecules were randomly placed within a 9.0 nm cubic box, with their orientations and translations set using randomly generated parameters. This approach provided a diverse range of starting structures, thereby enhancing the reliability of the dimerization simulations. All simulation details are comprehensively summarized in Table 1.

Table 1.

Details of each molecular system, encompassing self-assembly simulations of each segment and folding and dimerization simulations of hCT monomer and dimer (System), the number of simulated peptides in each molecular system (Num. Pep.), the total number of residues in each corresponding system (Num. Res.), the dimension size of each cubic simulation box (Box), the duration of each independent DMD simulation (Time), the total number of independent DMD simulations conducted for each molecular system (DMD Run), and the cumulative total simulation time for each corresponding system (Total Time).

System	Num. Pep.	Num. Res.	Box (nm)	Time (μs)	DMD Run	Total Time (μs)
Seg-1: hCT1-14	10	140	9.0	1.0	40	40.0
Seg-2: hCT15–25	10	110	9.0	1.0	40	40.0
Seg-3: hCT26–32	10	70	9.0	1.0	40	40.0
Full-length: hCT1-32	1	32	7.1	1.0	40	40.0
	2	64	9.0	1.0	40	40.0

2.2. Discrete Molecular Dynamics (DMD) Simulations.

All simulations were conducted at 300 K using the atomistic DMD algorithm[49, 50] coupled with the Medusa force field[51], which is known for its accurate predictions of protein stability and ligand binding affinity changes resulting from mutations[51–53]. Previous computational studies successfully replicated the experimentally observed self-assembly of hCT into β-sheet-rich structures and the amyloid-resistant behavior of hCT analogues (including DM-hCT, TL-hCT, F16L-hCT, F19L-hCT, and phCT), affirming the suitability of the Medusa force field for studying hCT conformational dynamics[41, 42, 45]. Similar to conventional MD force fields, the Medusa force field encompasses both bonded (covalent bonds, bond angles, dihedrals) and nonbonded (van der Waals, solvation, hydrogen bond, and electrostatic terms) interactions. However, DMD differs from traditional MD in its use of step functions for interatomic interaction potential functions instead of continuous ones. Van der Waals (VDW) parameters are derived from the CHARMM force field[54], while solvation energy is computed via the Lazaridis and Karplus effective energy function[55]. Hydrogen bonds are explicitly modeled using a reaction-like algorithm[52], and electrostatic interactions between charged atoms are determined using the Debye–Hückel approximation with a Debye length of approximately 10 Å.

The key distinction between DMD and conventional MD lies in how interaction potentials are represented. Instead of using continuous potential functions, DMD employs stepwise functions to approximate interatomic interactions. As a result, the system’s dynamics are governed by collision events, where atoms encounter an energy step and adjust their velocities based on conservation laws. By updating only the two atoms involved in each collision, predicting their subsequent interactions with neighboring atoms, and identifying the next collision using efficient sorting algorithms, DMD significantly boosts sampling efficiency. This method eliminates the need for frequent calculations of forces and accelerations, which are typically performed every 1–2 fs in traditional MD. When the step size in DMD is sufficiently small, the stepwise function approximates a continuous potential, making the outcomes of DMD comparable to those of conventional MD. Over time scales exceeding a few pico-seconds, DMD dynamics align with traditional MD, with differences primarily occurring at shorter, sub-picosecond time scales—specifically the intervals between successive collisions where a change in potential energy occurs. The DMD software is accessible to academic researchers through Molecules In Action, LLC (https://www.moleculesinaction.com). In our simulations, mass, time, length, and energy units were normalized to 1 Da, approximately 50 fs, 1 Å, and 1 kcal/mol, respectively. The validation of the Medusa force field, along with the EEF1 implicit solvation model, has been extensively demonstrated through benchmarking studies involving the aggregation of functional amyloid suckerin peptides[56] and pathological hIAPP peptides[57]. Because of its rapid computational speed and improved sampling efficiency, DMD has become widely used for investigating protein folding and aggregation dynamics, both within our group[46, 58–60] and in other research communities[61–63].

2.3. Analysis Methods.

The secondary structure was determined through the DSSP (Dictionary of Secondary Structure of Proteins) method[64]. Contacts were defined when the minimum distance between heavy atoms of two non-consecutive residues was within 0.55 nm. Peptides were considered part of the same oligomer if they were connected by at least one intermolecular contact, with the oligomer size being the total number of peptides involved[65]. Hydrogen bonds were identified by a distance of ≤3.5 Å between backbone N and O atoms and an N–H···O angle of ≥120°[66]. β-sheets were formed if at least two consecutive residues from each β-strand were connected by more than two backbone hydrogen bonds. A β-sheet oligomer was characterized by multiple β-sheets connected by at least one heavy atom contact pair, with its size determined by the total number of β-strands[67]. To construct a two-dimensional (2D) free energy surface, the probability P(x, y) of a conformational state with parameters x and y was calculated, and the free energy surface was derived as −RT ln P(x, y)[68].

3. Results and Discussion.

3.1. Fragments of hCT_1–14 Exhibited Extremely Weak Aggregation Capability, Predominantly Existing as Isolated Unstructured Monomers.

To investigate the aggregation capability and conformational dynamics of hCT_1–14, we conducted forty independent 1.0 μs DMD simulations of the self-assembly of ten hCT_1–14 segments. The self-assembly dynamics were monitored by examining the time evolution of the radius of gyration (Rg), the total number of inter-peptide backbone hydrogen bonds and heavy atom contacts, the largest oligomer size, and the content of each secondary structure (Figures 1a and S1a-c). The self-assemblies of hCT_1–14 exhibited an Rg value greater than 3.5 nm, similar to the initial fully isolated state, and only formed a transient small number of inter-peptide hydrogen bonds and contacts, indicating a relatively weak aggregation tendency. The frequent fluctuation of oligomer size, mostly ranging from 2–5, further confirmed that the self-aggregates of hCT_1–14 were unstable and easily dissociated. The time evolution of each secondary structure content indicated that during the self-assembly process, hCT_1–14 predominantly adopted unstructured conformations (i.e., random coil and bend structures), consistent with the simulation snapshots.

Figure 1. Self-assembly dynamics and conformational analysis of simulations involving ten hCT_1–14 fragments.

(a) The self-assembly dynamics of ten hCT_1–14 fragments is monitored by the time evolution of the radius of gyration (Rg, first column), the total number of inter-peptide backbone hydrogen bonds and heavy atom contacts (second column), the size of the largest oligomer (third column), and the content of each secondary structure (unstructured, helix, and β-sheet conformations, fourth column). Snapshots are presented every 250 ns below. One trajectory is randomly selected from a simulation pool of forty independent DMD runs. (b) The aggregation free energy landscape plotted as a function of the oligomer size against the average number of β-sheet residues per chain within each aggregate of hCT_1–14. Conformations for states labeled 1–6 on the surface of the free energy landscape are presented on the right. Data from all forty independent 1000 ns simulations are used for the analysis to capture the entire aggregation process. (c) The time evolution of the largest oligomer size for the ten hCT_1–14 fragments aggregated in each independent trajectory. (d) The mass-weighted oligomer size probability distribution for the self-assemblies of hCT_1–14 during the last 500 ns. (e) The average secondary structure propensity for the self-assemblies of hCT_1–14 during the last 500 ns.

The aggregation free energy landscape, also known as the potential mean force (PMF), for the self-assembly of ten hCT_1–14 segments was calculated as a function of oligomer size against the average number of residues adopting β-sheet per chain within the corresponding oligomer (Figure 1b). To capture the entire self-assembly process, all 40 independent 1.0 μs DMD trajectories were used for the PMF analysis. The results revealed only one global free energy basin, with the corresponding states exhibiting oligomer sizes of 1–3 and an average of nearly zero β-sheet residues per chain. This suggests that small-size aggregates formed by 1–3 hCT_1–14 segments without significant β-sheet conformation were the predominant species. Oligomers with sizes greater than 5 or with an average number of β-sheet residues greater than 1.0 exhibited high free energy values, indicating these conformations were unfavorable and rare. The time evolution of the largest oligomer size in each independent DMD simulation showed that ten hCT_1–14 segments only self-assembled into small, dynamic oligomers (sizes 1–5) and exhibited frequent association and dissociation (Figure 1c).

Ensemble average values of the radius of gyration, total number of inter-chain backbone hydrogen bonds and heavy atom contacts, largest oligomer size, and each secondary structure content over 40 independent trajectories exhibited no significant fluctuations during the last 500 ns (Figure S1d), suggesting that convergence was achieved for the self-assembly simulations of ten hCT_1–14 segments. Therefore, the last 500 ns of saturated simulation data were used to analyze the mass-weighted oligomer size probability distribution and average secondary structure propensity for the aggregates of hCT_1–14 (Figure 1c&d). On average, isolated monomers were the predominant species with a population of 40.1% (Figure 1c). Oligomeric states of hCT_1–14 populated only as dimers, trimers, and tetramers, with populations of 22.2%, 14.4%, and 9.6%, respectively. Oligomers formed by more than 5 peptides were extremely rare, mostly less than 5.0%. The secondary structure content showed that hCT_1–14 mostly adopted unstructured conformations (~80.9%), while structured conformations such as helix (~9.2%) and β-sheet (~2.9%) were relatively weakly populated (Figure 1e). Overall, our results suggested that the aggregation tendency of hCT_1–14 was too weak to form stable aggregates, and hCT_1–14 predominantly adopted an unstructured monomer state.

3.2. The hCT_15–25 Segment Exhibited Significant Aggregation Tendency and Readily Self-assembled into β-sheet-rich Aggregates.

Unlike hCT_1–14, which predominantly existed as isolated monomers, ten isolated hCT_15–25 fragments readily aggregated into a stable, β-sheet-rich oligomer composed of all ten peptides. This was evidenced by their self-assembly dynamics analysis (Figures 2a&S2a-c). Specifically, the ten isolated hCT_15–25 fragments spontaneously and rapidly aggregated into compact states with an Rg value of approximately 1.25 nm, significantly smaller than the initial fully isolated state. This aggregation was accompanied by the formation of numerous inter-peptide backbone hydrogen bonds and heavy atom contacts, with minimal fluctuations. The time evolution of oligomer size and secondary structure content indicated that all ten peptides spontaneously self-assembled into a single oligomer, predominantly composed of β-sheet structures. The radius of gyration, the total number of backbone hydrogen bonds and contacts, the largest oligomer size, and the secondary structure content all reached and maintained saturated values with minimal fluctuations. These results suggested that the β-sheet-rich aggregates formed by hCT_15–25 were relatively stable and that equilibrium was well achieved in the self-assembly simulation. The ensemble average of these conformational parameters over forty independent trajectories also showed very small changes during the last 500 ns (Figure S2d), indicating that the convergence for the self-assembly simulation of hCT_15–25 was reasonably achieved. The ensemble average conformational analysis revealed that the largest oligomer was consistently composed of ten peptides, with an average β-sheet content of approximately 30%. This indicated that the hCT_15–25 segment had a significant self-assembly capability in forming stable, β-sheet-rich aggregates.

Figure 2. Self-assembly dynamics and conformational analysis of simulations involving ten hCT_15–25 fragments.

(a) Monitoring the self-assembly dynamics of ten hCT_15–25 fragments via the time evolution of the radius of gyration (Rg, first column), the total number of inter-peptide backbone hydrogen bonds and heavy atom contacts (second column), the size of the largest oligomer (third column), and the content of each secondary structure (unstructured, helix, and β-sheet conformations, fourth column). Snapshots are presented every 250 ns below. One trajectory is randomly selected from a simulation pool of forty independent DMD runs. (b) Aggregation free energy landscape plotted as a function of the oligomer size against the average number of β-sheet residues per chain within each aggregate of hCT_15–25. Conformations for states labeled 1–6 on the surface of the free energy landscape are presented on the right. Data from all forty independent 1000 ns simulations are used for the analysis to capture the entire aggregation process. (c) Time evolution of the largest oligomer size for the ten hCT_15–25 fragments aggregated in each independent trajectory. (d&e) Probability distributions of mass-weighted oligomer size (d) and β-sheet oligomer size (e) for the self-assemblies of hCT_15–25 during the last 500 ns.

The aggregation free energy landscape for the self-assembly of hCT_15–25 was analyzed using all 40 independent 1.0 μs DMD simulations to capture the entire aggregation process. The free energy landscape of hCT_15–25 revealed a narrow and deep energy well, with the corresponding conformations having an oligomer size of 10 peptides and an average of 1–5 β-sheet residues per chain (Figure 2b). Oligomers with a size less than 9, present in the early aggregation state, exhibited very high free energy, suggesting such conformational states were very transient. The time evolution of the oligomer size within each independent DMD trajectory showed that all ten hCT_15–25 peptides rapidly aggregated into a 10-peptide stable oligomer within the first 200 ns without frequent dissociation and association (Figure 2c). The mass-weighted oligomer size probability distribution showed that oligomers with a size of 10, the total number of simulated peptides, accounted for 97.5% during the saturated state of the last 500 ns of simulations (Figure 2d). The β-sheet oligomer size distribution, corresponding to the number of β-strand peptides within each oligomer, showed that 6–9 peptides adopting β-strand formations were commonly observed in the self-assemblies of hCT_15–25 aggregates (Figure 2e). Overall, the self-assembly dynamics and conformational analysis suggested that hCT_15–25 had a strong tendency to form stable β-sheet aggregates.

The average secondary structure content for the hCT_15–25 self-assemblies showed that hCT predominantly assumed unstructured and β-sheet conformations with propensities of ~62.7% and 31.5%, respectively (Figure 3a). The average secondary structure propensity for each residue revealed that β-sheet structures were primarily formed by the central residues K18, F19, H20, T21, and F22, with probabilities ranging from ~35.1%−62.5% (Figure 3b). Residue-pairwise contact frequency analysis indicated that interactions among aromatic residues (including F16, F19, H20, and F22) drove the self-assembly of hCT_15–25, forming in-register or out-register parallel and antiparallel β-sheet structures (Figure 3c). The in-register parallel β-sheet, characterized by the maximum number of aromatic and hydrophobic residue-pairwise contacts, was the most prevalent configuration.

Figure 3. Structural analysis of the self-assemblies formed by hCT_15–25 fragments.

(a) Average secondary structure propensity for the self-assemblies of hCT_15–25 during the last 500 ns. (b) Average propensity of each residue from hCT_15–25 to assume unstructured, β-sheet, helix, and turn conformations within their aggregates during the last 500 ns simulations. (c) Residue-pairwise contact frequencies among heavy atoms within the self-assemblies of hCT_15–25. Representative extended parallel and anti-parallel β-sheet motifs, exhibiting high structural contact pattern frequency on the contact map, are also presented on the right.

The central aromatic residues F19, H20, and F22 exhibited the strongest β-sheet propensities of 53.0%, 61.3%, and 45.4%, respectively, and played critical roles in maintaining β-sheet formations of hCT_15–25. These residues were predominantly involved in inter-peptide interactions within the β-sheet structures. Notably, F19 demonstrated a significantly greater capability for forming inter-peptide β-sheet structures compared to F16, which aligned with previous NMR measurements. For instance, substituting F19 with leucine (F19L), but not F16 with leucine (F16L), significantly delayed the primary nucleation of hCT in vitro[33]. The spontaneous aggregation of the hCT_15–25 segment into β-sheet aggregates also corroborated previous experimental studies, as the formation of amyloid fibrils by hCT_15–19 had been extensively documented[33, 37, 38].

3.3. The hCT_26–32 Fragments Displayed Weak Aggregation Tendency, Primarily Adopting Isolated Unstructured Monomers.

Despite hCT_26–32 being rich in hydrophobic residues, the dynamic self-assembly simulations indicated a relatively weak aggregation capability for this segment (Figures 4a and S3a-c). Specifically, the radius of gyration of ten hCT_26–32 peptides fluctuated between 3.0 and 4.7 nm throughout the simulations. Moreover, the total number of backbone hydrogen bonds and heavy atom contacts exhibited transient, low numbers with frequent fluctuations. The time evolution of oligomer size revealed that hCT_26–32 formed predominantly small oligomers, mostly less than 4 peptides, characterized by frequent association and dissociation events. Unstructured conformations were consistently present at 85%−100%, while structured conformations such as β-sheet and helix were rarely observed. Snapshots from the simulations consistently depicted hCT_26–32 maintaining an isolated and unstructured monomeric state.

Figure 4. Self-assembly dynamics and conformational analysis of simulations involving ten hCT_26–32 fragments.

(a) The self-assembly dynamics of ten hCT_26–32 fragments are monitored through the time evolution of the radius of gyration (Rg, first column), the total number of inter-peptide backbone hydrogen bonds and heavy atom contacts (second column), the size of the largest oligomer (third column), and the content of each secondary structure (unstructured, helix, and β-sheet conformations, fourth column). Snapshots are presented every 250 ns below. One trajectory is randomly selected from a simulation pool of forty independent DMD runs. (b) The aggregation free energy landscape is plotted as a function of the oligomer size against the average number of β-sheet residues per chain within each aggregate of hCT_26–32. Conformations for states labeled 1–6 on the surface of the free energy landscape are presented on the right. Data from all forty independent 1000 ns simulations are used for the analysis to capture the entire aggregation process. (c) The time evolution of the largest oligomer size for the ten hCT_26–32 fragments aggregated in each independent trajectory. (d) The mass-weighted oligomer size probability distribution for the self-assemblies of hCT_26–32 during the last 500 ns. (e) The average secondary structure propensity for the self-assemblies of hCT_26–32 during the last 500 ns.

The aggregation free energy landscape of hCT_26–32 exhibited a single global energy basin with corresponding conformational states comprising 1–3 peptides but lacking β-sheet formations (Figure 4b). Aggregates featuring β-sheet formations or formed by more than three peptides showed significantly high free energy, indicating such conformations were unfavorable for hCT_26–32. The time evolution of the largest oligomer size for hCT_26–32 in each DMD simulation demonstrated that ten peptides could only transiently aggregate into dimers and trimers, with larger aggregates rarely observed (Figure 4c). The time evolution of ensemble-averaged structural parameters exhibited a large radius of gyration (~3.9 nm), a small number of inter-peptide backbone hydrogen bonds (~0.8), heavy atom contacts (~18), the largest oligomer size (~2.4), and a high content of unstructured conformations (~93.2%), without significant fluctuations (Figure S3d) during the last 500 ns. This suggested the convergence of the self-assembly simulations, confirming that hCT_26–32 lacked the capability to form β-sheet aggregates. Equilibrium conformational analysis further confirmed these findings (Figure 4d&e). Aggregates composed of 1–3 hCT_26–32 peptides were the predominant species, with the probabilities of monomers, dimers, and trimers being ~68.6%, 20.4%, and 7.2%, respectively (Figure 4d). The unstructured conformations of hCT_26–32 reached up to 93.2%, while the β-sheet content was less than 2% (Figure 4e). Overall, the self-assembly simulations indicated that this segment preferred to assume unstructured monomer states rather than forming stable aggregates.

3.4. Full-Length hCT_1–32 Monomer Predominantly Assumed Dynamic Helical Conformations with a Small Population of Transient β-Sheets, While Dimerization Promoted a Helix-to-β-Sheet Conversion.

Apart from investigating the self-assembly conformational dynamics of each segment, the folding and dimerization dynamics of full-length hCT_1–32 peptides were also studied to uncover the impact of inter-segment interactions on the structure and aggregation of hCT_1–32. Both folding and dimerization conformational dynamics of hCT_1–32 were extensively investigated through 40 independent 1.0 μs DMD simulations. The folding dynamic simulations showed that hCT_1–32 primarily assumed dynamic helical conformations, especially for the central residues 14–19 (Figure 5a). Additionally, transient β-hairpin formations by residues 20–31 were observed. Upon dimerization, the stability of the β-sheet formations, which were very transient in hCT_1–32 monomers, was enhanced, accompanied by a decrease in helical conformations, as monitored by the time evolution of each secondary structure and inter-peptide backbone hydrogen bonds and heavy atom contacts (Figure 5b).

Figure 5. Conformational dynamics analysis for the folding and dimerization of full-length hCT_1–32.

(a) The folding dynamics of hCT are monitored through the time evolution of the secondary structure for each residue of the full-length hCT_1–32. Snapshots are presented every 250 ns. (b) The conformational dynamics for the dimerization of hCT_1–32 are tracked by the time evolution of the secondary structure for each residue and the total number of inter-peptide backbone hydrogen bonds and heavy atom contacts. Snapshots are shown every 250 ns. (c) The average content of each secondary structure for hCT_1–32 in monomeric and dimeric states. (d) The average propensity of each residue to assume helix and β-sheet structures for hCT in the one- and two-peptide simulations. Only the last 500 ns from forty independent DMD simulations are used for the average secondary structure analysis to avoid potential bias from the initial state.

The equilibrium and convergence assessment was conducted by examining the time evolution of several structural parameters, including the radius of gyration, the total number of backbone hydrogen bonds and heavy atom contacts, along with the content of each secondary structure (Figures S4 and S5). All these parameters indicated that the simulations of both systems reached equilibrium during the last 500 ns. Thus, the saturated simulation data from this period was used for the conformational analysis of hCT_1–32 monomer and dimer. Although the initial states of each independent DMD run for the dimerization of hCT_1–32 varied in terms of relative intermolecular translocation and orientations, frequent structural transitions were observed in each simulation trajectory. This indicated that none of the simulations were trapped in local conformational energy basins (Figure S6a), and sufficient sampling was achieved. Furthermore, the average secondary structure content for each independent trajectory during last 500 ns showed similar corresponding values (Figure S6b), suggesting that our dimerization simulations effectively minimized any bias from the initial states. The observed conformational transition toward β-sheets in all independent DMD trajectories suggested that this transition was driven by the biophysical properties of hCT, rather than by the initial states.

The average secondary structure content analysis showed that both the hCT_1–32 monomer and dimer primarily assumed unstructured conformations, with propensities of ~68.6% and 65.2%, respectively (Figure 5c). The structured conformations of the hCT_1–32 monomer included ~15.3% helix and ~4.4% β-sheet. The predominance of unstructured conformations along with partially dynamic helices in the hCT_1–32 monomer was consistent with prior CD and NMR experimental measurements[27, 36]. Dimerization of hCT_1–32 resulted in a decrease in helix content accompanied by an enhancement in β-sheet content, with the hCT_1–32 dimer displaying ~13.9% helix and ~8.9% β-sheet.

To better understand the conformational characteristics, we further analyzed the average secondary structure propensity of each residue within hCT_1–32 monomers and dimers (Figure 5d). The hCT_1–32 monomer exhibited two helical regions around residues 4–9 and 14–19, with average probabilities of ~21.4% and ~53.8%, respectively. The hCT_1–32 dimer also displayed helical conformations in the same regions. The helical propensity of residues 4–9 in the hCT dimer was similar to that in the monomer, but the average helical propensity of residues 14–19 decreased to ~44.7%, significantly lower than the ~53.8% observed in the monomer. Transient β-sheets in the hCT_1–32 monomer were mainly formed by residues 8–13, 18–24, and 26–31, with average propensities of ~4.1%, 7.9%, and 8.4%, respectively (Figure 5d). Similar β-sheet regions were also observed in hCT_1–32 dimers, but with a much greater β-sheet tendency. For instance, the average β-sheet probabilities of residues 8–13, 18–24, and 26–31 were ~7.6%, 16.4%, and 14.8% in the hCT_1–32 dimer, approximately twice those in the hCT_1–32 monomer. Overall, these results indicated that the aggregation of hCT_1–32 led to a decrease in central helical formations and an enhancement of β-sheet conformations for residues 8–13, 18–24, and 26–31.

3.5. Residues 15–25 Triggered hCT Conversion to β-Sheet Formations via Aromatic and Hydrophobic Interactions.

The self-assembly dynamics and conformational analysis suggested that only the hCT_15–25 segment spontaneously assembled into β-sheet aggregates, whereas the formation of β-sheet assemblies from hCT_1–14 and hCT_26–32 segments was extremely weak (Figures 1–4). However, residues from hCT_1–14 and hCT_26–32 could also contribute to β-sheet formations in both full-length hCT_1–32 monomers and dimers (Figure 5). These results indicated that hCT_15–25 may have assisted hCT_1–14 and hCT_26–32 through their residue-pairwise interactions. Therefore, the frequency of residue-pairwise contact frequency in hCT_1–32 monomers and dimers was further analyzed (Figure 6) using the simulation data from the last 500 ns of 40 independent DMD trajectories.

Figure 6. Identifying critical residue-pairwise interactions in full-length hCT_1–32 monomer and dimer structures.

(a) The frequency of intra-peptide residue-pairwise contacts formed between heavy atoms within the hCT_1–32 monomer. Representative structured contact patterns and their corresponding structures, selected based on contact frequency, are labeled and presented as 1–4 below. (b) Intra-chain (lower diagonal) and inter-chain (upper diagonal) residue-wise contact frequencies of heavy atoms for the hCT_1–32 dimers. Representative structured motifs exhibiting high contact pattern frequency, labeled 1–8 on the contact map, are also presented. (c) The average number of intra-peptide and inter-peptide contacts for each residue within hCT_1–32 dimers. Only the last 500 ns of saturated simulation data from forty independent DMD trajectories are used for this analysis.

Consistent with the secondary structure analysis, two helical contact patterns along the diagonal were observed around residues 4–11 and 15–22 within hCT_1–32 monomers and dimers, which also agreed with prior NMR measurements[33, 38] (snapshots 1&2 in Figure 6a&b). Moreover, the helical contact pattern formed by residues 15–22 displayed a higher contact frequency than that formed by residues 4–11, indicating that the helical structures formed by residues 15–22 were more stable. In addition to the helical contact pattern, hCT_1–32 exhibited a β-hairpin contact pattern formed by residues 18–30 through hydrophobic residue-pair interactions, such as F22-A26 and F19-V29 (snapshot 3 in Figure 6a). Furthermore, the hydrophobic interactions of M8-A26 and L9-I25 facilitated the capping of hCT_8–13 at the β-sheet elongation edge of the β-hairpin region, although the average frequency of this interaction was relatively weak (snapshot 4 in Figure 6a).

A similar intra-peptide β-hairpin contact pattern formed by residues 18–30 was also observed in the hCT_1–32 dimer (snapshot 3 in Figure 6b). Interestingly, various inter-peptide parallel β-sheets (including hCT_8–12 vs hCT_26–30 and hCT_17–22 vs hCT_14–19) and anti-parallel β-sheets (including hCT_8–12 vs hCT_19–23, hCT_26–32 vs hCT_19–23, hCT_19–25 vs hCT_17–23, and hCT_26–31 vs hCT_26–30) were identified, primarily stabilized by interactions among hydrophobic and aromatic residues (snapshots 4–9 in Figure 6b). The average number of inter-peptide contacts revealed that residues F16, F19, and F22 exhibited significantly higher values than other regions, indicating that this central region played a critical role in driving hCT aggregation. This was because the hCT_15–25 segment could self-aggregate into inter-peptide β-sheets due to its strong self-assembly capability. Additionally, interactions between the central hydrophobic and aromatic residues F16, F19, and F22 with N-terminal residues M8, L9, and Y12, or C-terminal residues A26, I27, V29, and A30, facilitated the formation of β-sheets for hCT_8–13 and hCT_26–30. Given that the self-assembly capability of hCT_1–14 and hCT_26–32 to form inter-peptide β-sheets was extremely weak, the presence of hCT_15–25 promoted the formation of inter-peptide β-sheets by facilitating interactions between N-terminal residues 8–13 and C-terminal residues 26–30 with F16, F19, and F22. Thus, our results suggested that the aggregation of full-length hCT_1–32 was primarily triggered by the central hCT_15–25 region, which leads to the formation of β-sheet aggregates by residues 8–13, 15–25, and 26–30.

3.6. Protecting the Stability of the Central Helix Structure as a Promising Strategy for Developing Amyloid-Resistant hCT Analogues.

Due to its significant aggregation capability and its ability to adopt both helical and β-sheet conformations within full-length hCT_1–32 monomers and dimers, hCT_15–25 played a crucial role in the overall structural dynamics of the hCT_1–32 peptide. To understand this impact, the stability of the central helical conformations of hCT_15–25 was further analyzed using a conformational free energy landscape based on data from the last 500 ns of forty independent DMD simulations of hCT_1–32 dimers (Figure 7a&b). This landscape was calculated to correlate the average helical content of hCT_15–25 with the corresponding helix-to-β-sheet ratio of the entire hCT_1–32 dimers. The hCT_1–32 dimers revealed three distinct energy basins. Within these basins, dimers featuring highly populated helical structures around hCT_15–25 also exhibited predominant helical conformations throughout the entire peptide (e.g., snapshot 1 in Figure 7). Conversely, dimers with prevalent β-sheet conformations across the entire hCT_1–32 peptide corresponded to hCT_15–25 adopting non-helical conformations (e.g., snapshots 3&4 in Figure 7). Additionally, moderate helical populations in hCT_15–25 were associated with dimers displaying intermediate levels of both helix and β-sheet formations (e.g., snapshot 2 in Figure 7). Overall, the conformational free energy landscape suggested that maintaining the stability of the central helix around hCT_15–25 would prevent the entire hCT_1–32 peptide from converting to β-sheet structures. Thus, further amyloid-resistant hCT analogues should focus on improving the helical stability around hCT_15–25.

Figure 7. The correlation between the helical formation of hCT_15–25 and the corresponding helix-to-β-sheet ratio of the entire hCT_1–32 dimers.

(a&b) The conformational free energy landscape is generated as a function of the average helix ratio of hCT_15–25 against the content of helix (a) and β-sheet (b) for the entire peptide within hCT_1–32 dimers. Conformational states around the free energy basins corresponding to the most, moderate, and least populated helical formations of hCT_15–25, which also exhibit the least, moderate, and most populated β-sheet formations for the entire hCT_1–32 dimers, are labeled 1–4 on the free energy surface, with the corresponding snapshots presented on the right. To avoid potential bias from the initial states, only the last 500 ns from forty independent trajectories are used for this analysis.

Indeed, the inhibition mechanism of prior amyloid-resistant analogues also supported that the central aromatic region played a critical role in driving the aggregation of full-length hCT_1–32. For instance, our previous simulation results showed that the reason for the reduced aggregation tendency of TL-hCT compared to the wild type was that the replacement of Y12L, F16L, and F19L significantly enhanced the stability of the central helix[41, 45]. Experimentally, it was observed that replacement of F19L but not F16L strongly delayed the primary aggregation of hCT[33]. This was computationally explained by the fact that F19L enhanced the central helix’s stability, whereas F16L did not[45]. The DM-hCT, comprising Y12L and N17H double mutations, exhibited weaker aggregation than hCT[35, 42]. This could mainly be attributed to the Y12L substitution, which disrupted the aromatic interaction with the central region, resulting in the central region being more helical than the wild type hCT[42]. Additionally, the β-sheet formed by the C-terminal hydrophobic residues interacting with the central helical region in the wild-type hCT monomer and dimer was frequently observed. The substitutions Y12L, N17H, A26N, I27T, and A31T in phCT destroyed the aromatic interaction of Y12 and the hydrophobic interaction of hCT_26–31 with the central aromatic regions[27]. As a result, phCT also exhibited more helical conformations and weaker aggregation than hCT[41]. Despite distinct mutations in TL-hCT, DM-hCT, and phCT, all of these hCT analogues computationally exhibited stronger helical stability around the central region and experimentally demonstrated a weaker aggregation tendency than wild-type hCT[41, 42]. These findings confirmed that amyloid-resistant hCT analogues should target the improvement of the central helical stability, as this region not only exhibited the strongest aggregation tendency but also significantly affected the overall peptide structure in silico.

Interestingly, prior NMR assays revealed that EGCG effectively inhibited hCT from forming oligomeric and fibrillar aggregates by binding to the central aromatic residues, primarily within hCT_15–25[36], and this aligned with our findings that hCT_15–25 played a pivotal role in driving hCT aggregation. The hCT_15–25 segment exhibited the primary prion aggregation potential, spontaneously self-assembling into well-ordered cross-β aggregates. This region also drove the aggregation of full-length hCT, promoting the transition to β-sheet structures. In contrast, although the hCT_1–14 and hCT_26–32 segments contributed partially to β-sheet formation within full-length hCT aggregates, they displayed minimal aggregation tendencies and primarily remained in an isolated monomeric state. These findings suggested that future efforts to design amyloid-resistant hCT analogs or amyloid inhibitors should focus on the hCT_15–25 region. This could be achieved by rationally designing specific mutants or inhibitors that bind to this region, enhancing helical stability and preventing self-assembly. While the design of such mutants required further computational and experimental validation, our comprehensive investigation into the self-assembly of hCT fragments, along with the folding and dimerization of full-length hCT, offered valuable insights and a promising direction for the development of amyloid-resistant hCT analogs and inhibitors.

4. Conclusions

The intrinsically abnormal aggregation of hCT disrupts its physiological activity, increases the risk of MTC, and hampers its clinical use for bone-related diseases. To systematically investigate the aggregation mechanism of hCT at the molecular level, we conducted multiple independent microsecond atomistic DMD simulations. These simulations included the individual self-assembly of hCT_1–14, hCT_15–25, and hCT_26–32, as well as the folding and dimerization of the full-length hCT_1–32. Our results suggested that only the hCT_15–25 segment could independently self-assemble into stable β-sheet aggregates, while hCT_1–14 and hCT_26–32 primarily formed unstructured monomers and transient small-sized oligomers. Full-length hCT_1–32 monomers mainly exhibited dynamic unstructured conformations with partial helical regions around residues 4–9 and 14–19. Dimerization of full-length hCT_1–32 led to a helical-to-β-sheet conversion around residues 14–19 and enhanced β-sheet conformations for residues 8–13, 18–24, and 26–31. Residue-pairwise contact frequency and conformational free energy landscape analyses revealed that the formation of β-sheets in the entire hCT_1–32 peptide correlated with the loss of the helical structure in the central region. Maintaining the stability of the central helix around hCT_15–25 was found to prevent the entire hCT_1–32 peptide from converting to β-sheet structures. Therefore, designing amyloid-resistant hCT analogues should focus on improving the helical stability around hCT_15–25, as this region not only exhibited the strongest aggregation tendency but also significantly influenced the overall peptide structure.

Supplementary Material

Appendix A. Supplementary data

Supplementary figures

Data availability

Data will be made available on request.