A Buried Glutamate in the Cross-β Core Renders β-endorphin Fibrils Reversible

Abstract

Functional amyloids are abundant in living organisms from prokaryotes to eukaryotes playing diverse biological functions. In contrast to the irreversible aggregation of most known pathological amyloids, we postulate that naturally-occurring functional amyloids are reversible under the evolutionary pressure to be able to module the fibrillization process, reuse the composite peptides, or perform their biological functions. β-endorphin, an endogenous opioid peptide hormone, forms such kind of reversible amyloid fibrils in secretory granules for efficient storage and returns to the functional state of monomers upon release into the blood. The environment change between the low pH in secretory granules and the neutral pH in extracellular spaces is believed to drive the reversible fibrillization of β-endorphin. Here, we investigate the critical role of a buried glutamate, Glu8, in the pH-responsive disassembly of β-endorphin fibrils using all-atom molecular dynamics simulations along with structure-based pKa prediction. The fibril was stable at pH 5.5 or lower with all the buried Glu8 residues protonated and neutrally charged. After switching to the neutral pH, Glu8 residues of peptides at the outer layers of the ordered fibrils became deprotonated due to partial solvent exposure, causing sheet-to-coil conformational changes and subsequent exposure of adjacent Glu8 residues in the inner chains. Via iterative deprotonation of Glu8 and induced structural disruption, all Glu8 residues would be progressively deprotonated. Electrostatic repulsion between deprotonated Glu8 residues along with their high solvation tendency disrupted the hydrogen bonding between the β1 strands and increased the solvent exposure of those otherwise buried residues in the cross-β core. Overall, our computational study reveals that strategic positioning of ionizable residues into the cross-β core is a potential approach to design reversible amyloid fibrils as pH-responsive smart bio-nanomaterials.

Graphical Abstract

The pH-responsive disaggregation of β-endorphin amyloids. The experimentally determined β-endorphin fibril structure is stable at pH 5.5 or lower. At neutral pH, the fibril structure is destabilized by a progressive deprotonation of a buried Glu8.

Introduction

The self-assembly of proteins into nanosized fibrils has attracted extensive attention due to the association with numerous human degenerative diseases^1–5, physiological functions^6–9, and potential applications in biomedicine and bio-nanotechnology^10–12. Besides many commonly-known amyloid diseases¹³ caused by the pathological aggregation of amyloid peptides (e.g., Aβ in Alzheimer’s disease¹⁴, αS in Parkinson’s disease¹⁵, hIAPP in type-2 diabetes¹⁶, and Tau in multiple neurodegenerative diseases¹⁷), functional amyloids (e.g., curli¹⁸, suckerin^19–21, blood fibrin²², and β-endorphin^{23, 24}) have been discovered in all kingdoms of life, including organisms ranging from bacteria to human playing diverse functional roles²⁵. One recently discovered functional amyloid in human is formed by β-endorphin in the central and peripheral nervous systems^{23, 26}. β-endorphin is a 31-residue endogenous opioid neuropeptide responsible for the regulation of pain perception and stress response^27–29. The peptides synthesized in the cells are stored in secretory granules in a highly compact amyloid state for efficient storage until a signal triggers their release into the blood or extracellular space^{27, 30, 31}.

One possible feature separating disease-related amyloids^{1–5, 32} from functional amyloids might be fibril stability or fibrillization reversibility, since functional amyloids likely experience the evolutionary pressure of reusing or recycling peptides forming these aggregates but pathological amyloids do not. Most of the disease-related aggregations are irreversible and the corresponding fibrils are stable even under highly denaturing conditions^33–35, while peptide hormone fibrils (e.g., β-endorphin) are often required to dissociate into monomers when the cellular environment changes⁶. For example, prior experiments showed that 31 out of 42 randomly selected peptide hormones could assemble into amyloid fibrils in the presence of heparin at granule-relevant pH 5.5^{24, 27}. All of these pre-formed hormone fibrils would disaggregate into monomers at pH 6.0 or 7.4²⁷. Since the dissociation process of β-endorphin fibrils is much faster than the synthesis rate, the reversible aggregation enables rapid responses to the stimulus – i.e., release enough amounts of the hormone/neuropeptide into the blood in a short period of time^{23, 30, 36}. The pH-responsive fibrillization was also observed in other functional amyloids^{37, 38}. For example, the suckerin amyloids, formed by the squid ring teeth protein under neutral pH, could dissociate into the unstructured form in acid environment due to the protonation of histidine residues^{21, 39, 40}. Owing to the high biocompatibility, low immunogenicity, pH-responsiveness, and reversibility of fibrillization, functional amyloids have been explored for potential applications in biomedicine and bio-nanotechnology^{41, 42}. Therefore, uncovering the molecular mechanism for the pH-responsive regulation of the hormone/neuropeptide aggregation will not only help understand the secretory pathway of the peptide hormones^{6, 43, 44} but also aid in the future design of peptide-based stimuli-responsive smart materials.

The aggregation and disaggregation of β-endorphin strongly depend on the chemical environment (including pH, phosphate, and cofactors)^{6, 23, 24}. The β-endorphin fibril structure under pH 5.5 determined by solid-state NMR (ssNMR) is similar to most known fibril structures⁴⁵, featuring a cross-β core with the β-strands of each chain aligned parallel in-register and perpendicular to the fibril axis^{14, 23, 46}. Each chain has three β-strands formed by residues 4–10 (β1), 14–19 (β2), and 21–24 (β3) (Fig. 1). Sidechains buried inside the fibril core partition into one hydrophobic (i.e., Phe4, Phe18, Ala21, Ile23, and Ala26) and one hydrophilic (i.e., Thr6, Glu8, Ser10, and Thr16) clusters. As the only residue with ionizable sidechain in the core, Glu8 has been postulated to contribute to the pH-responsive reversibility of β-endorphin fibril²⁷. The pH change during the exocytosis of the β-endorphin from the acidic secretory granule with pH ~5.5 to the physiological pH of ~7.4 in the blood could induce the change of Glu8 protonation state, which subsequently modulates the stability of β-endorphin fibril. However, since the sidechain pKa of glutamate in solution is ~4.5 which is outside the above pH ranges, the molecular mechanism for the reversible β-endorphin fibril remains to be uncovered.

Fig. 1. Experimentally determined β-endorphin fibril structure.

a) The sequence of β-endorphin used in our simulation. Three β-strands forming the fibrils are highlighted in blue, green, and red, respectively. b) The reconstructed β-endorphin proto-fibril structure comprised of 10 peptides based on prior ssNMR determined fibril structure (PDB ID: 6TUB) is shown in both top and side views. Structural details of a single chain in β-endorphin proto-fibril is also shown, highlighting a hydrophobic cluster formed by residues F4, F18, A21, I23, and A26 (indicated by the purple circle); and also a hydrophilic cluster in the core consisted of residues T6, E8, S10, and T16 (the cyan circle). c) The computationally predicted pKa values of Glu8 in each chain based on the experimental structure of the β-endorphin fibril. d) The free energy of the β-endorphin fibril structure as a function of the pH values.

Here, we applied all-atom molecular dynamics (MD) simulations with explicit solvent to systemically investigate the role of Glu8 in the pH-dependent stability of a model β-endorphin fibril comprised of 10 peptides. The protonation states of 10 Glu8 residues were estimated according to the structure-based pKa calculation^47–49. The experimentally determined β-endorphin fibril structure²⁷ was stable at the secretory granule-relevant pH ~5.5, where only the two exposed Glu8 residues located at two fibril ends were deprotonated with their pKa values lower than 5.0 and all the buried Glu8s of inner chains in β-endorphin fibril remained protonated and neutrally charged with pKa above 5.5. Specifically, by using the structural ensemble sampled in MD simulations at pH 5.5 Glu8 residues in the second outer layers were estimated to have pKa below 7.0 due to partial solvent exposure while all other inner Glu residues have pKa values above 7.0. After switching to pH 7.4, Glu8 residues in both first and second outer layers were thus deprotonated, which induced a sheet-to-coil transition in the corresponding β1 strands and subsequently exposing adjacent Glu8 in the inner layers. Via iterative Glu8 deprotonation and induced structural disruption, all Glu8 residues would be progressively deprotonated. Due to electrostatic repulsions between the negatively charged Glu8 residues as well as the increased solvation, the cross-β core of the fibril was disrupted with increased solvent accessible surface area of the originally buried hydrophobic residues. Although the complete fibril disaggregation and monomer release caused by the deprotonation of Glu8 was not observed in our simulations due to the high free energy barrier expected to overcome from the aggregated state to monomers, our computational study captured the early events of pH-induced fibril disaggregation^{27, 50} and revealed the molecular mechanism for the pH-dependent stability of β-endorphin fibrils, important for the efficient storage and rapid release of the analgesic hormone. Moreover, our study suggests that strategic positioning of ionizable residues into the cross-β core might be an attractive approach to design reversible amyloid fibrils as pH-responsive smart bio-nanomaterials.

Materials and methods

The Simulated Molecular Systems.

The β-endorphin peptide consists of 31 residues (the primary sequence in Fig. 1a). Based on the prior solid-state NMR determined data (PDB: 6TUB),²⁷ a 10-peptide fibril structure was constructed using our in-house code as the initial structure for simulation (Fig. 1b). The protonated Glu8 is buried in the fibril core (Fig. 1b). To probe whether the Glu8 plays a key role in the disaggregation of β-endorphin fibril, two additional systems with E8Q and E8L mutations, where the negatively charged glutamate was replaced by either the polar glutamine or the hydrophobic leucine, were also simulated.

Molecular Dynamics Simulations.

All the MD simulations were performed in the isothermal–isobaric (NPT) ensemble using Gromacs-2020.3 software⁵¹ with the AMBER99SB-ILDN force field⁵². For each system, the 10-peptide β-endorphin fibril was placed in the center of a rectangular box of 10.0 × 10.0 × 10.0 nm³ and solvated with TIP3P water⁵³. Counterions, including Na⁺ and Cl⁻, were added to neutralize the system and mimic a physiological salt concentration (0.15 mM). The bond length of peptides and water molecules was constrained respectively using the LINCS⁵⁴ and SETTLE⁵⁵ algorithms, allowing an integration time step of 2 fs. Nonbonded pair lists were updated every twenty integration steps. The protein, counterion and water groups were separately coupled to an external heat bath at 310 K with a coupling constant of 0.1 ps using a velocity rescaling coupling method⁵⁶. The pressure was kept around 1 bar applying the Parrinello–Rahman algorithm with a coupling time constant of 1.0 ps⁵⁷. Electrostatic interactions were treated with the particle mesh Ewald (PME) method⁵⁸ with a real space cutoff of 1.4 nm. The van der Waals interactions were calculated using a cutoff of 1.4 nm. Periodic boundary conditions were applied in all the simulations.

Analysis Methods.

All the data analysis using our in-house-developed codes and the tools implemented in Gromacs-2020.3 software package^{51, 59}. Secondary structure analyses were performed using the dictionary secondary structure of protein (DSSP) method⁶⁰. The structural stability of the β-endorphin proto-filament was examined by the time evolution of a backbone-root-mean-square-derivation (backbone-RMSD) corresponding to the optimized initial structure. The pKa prediction was used a fast empirical approach pKa predictors, PROPKA 3.0⁴⁹, which is based on better physical description of the desolvation and dielectric response for the protein^{47, 48}.

Results and discussion

The fibril structure is favorable for β-endorphin under the acidic environment.

To investigate the effect of pH environment changes on the protonation state of Glu8 residues and subsequently the fibril stability, we first calculated the pKa values of Glu8 in the fibril (Fig. 1c) using PROPKA 3.0, a rapid empirical pKa prediction tool⁴⁷. Different from solvent-exposed glutamate in solution, which has the side-chain pKa value around 4.35 as measured by NMR experiment⁶¹, the estimated pKa values of buried Glu8 residues in β-endorphin fibril were much higher than the secretory granule-relevant pH of 5.5^{27, 62} while the pKa values of two exposed Glu8 residues located at the elongation ends were lower than 5. Hence, all buried Glu8 residues except the two exposed ones are expected to be protonated (i.e., uncharged) in the fibril conformation at the pH of 5.5. Such pKa shifts of ionizable groups upon embedded in the hydrophobic protein cores have been well established by prior computational and experimental studies^63–68. The free energy of the β-endorphin fibril as a function of pH values was also predicted using PROPKA 3.0⁴⁷ (Fig. 1d). There is no obvious free energy change when the pH value ranges from 0 to 5 followed by a significant increase once the pH is larger than 6. Hence, the pKa analysis suggests that the experimentally determined fibril structure is favorable at the acidic environment with a pH value less than 6, but becomes unfavorable once it is released to the blood circulation with a pH of ~7.4.

To evaluate the stability of β-endorphin fibril under the acidic environment, two molecular systems – one with all Glu8 protonated corresponding to pH environments lower than 4.0 and the other with two only exposed Glu8 deprotonated (corresponding to 20% deprotonation) at pH around 5.5 were simulated. Here, Glu8 residues from the two outer layer chains at two fibril elongation ends were deprotonated according to the pKa prediction (chains 1 & 10 in Fig. 1c.) For each molecular system, two 1000-ns independent simulations were performed (Fig. 2). The time evolution of backbone root-mean square deviation (RMSD) of the whole β-endorphin fibril was saturated around 0.40 nm with a small fluctuation (Fig. 2a&d) for both systems. The time evolution of RMSD per chain was also analyzed (Fig. 2b&e). Except for two outer chains located at two fibril ends which had RMSD larger than 0.4 nm after reaching equilibrium, the RMSD of most inner chains in the fibril was less than 0.2 nm. Examination of the final representative snapshots from each trajectory revealed that the fibril structure was well retained without much structural destabilization (Fig. 2c&f). Deprotonation of the two exposed Glu8 only slightly increased the average RMSD of the β-endorphin fibril from 0.39 nm to 0.40 nm (Fig. 2g). The secondary structure analysis (Fig. 2h&i) revealed that the overall β-sheet content was ~47%, and the three β-strands were well maintained without much destruction, which was further confirmed by the RMSD analysis of each β-strand (Fig. S1).

Fig. 2. The structural stability analysis of β-endorphin fibrils under acid environment.

a) The time evolution of the backbone RMSD of the whole fibril, (b) the backbone RMSD of each chain in the fibril, and (c) the final structures in each of the two independent simulations are presented for the β-endorphin with all Glu8 residues protonated (noted as 0% deprotonation). Each independent MD simulation lasted 1000 ns. (d-f) The same results are shown for simulations with two Glu8 residues in the outer layer chains deprotonated (denoted as 20% deprotonation). Chains with Glu8 deprotonated were colored in pink. (g) The probability density distribution as a function of RMSD. Only structures in the last 200 ns simulations were used for the analysis. (h) The average propensities of each secondary structure (including random coil, β-sheet, helix, bend and turn) and (i) the probability of each residue to form β-sheet in each simulated system. (j) The averaged pKa prediction for Glu8 residues in each chain based on conformation ensembles sampled in the last 50 ns.

Since pKa is known to be sensitive to structural changes, we recomputed the average Glu8 pKa values using the equilibrium conformations of β-endorphin fibril during the last 50 ns for each of two molecular systems (Fig. 2j). Both systems had similar pKa profiles and the pKa values of all buried Glu8 residues in the inner chains were still higher than 5.5, suggesting no further changes of protonation states compared to the initial state. Overall, our simulation results confirmed that the experimentally determined β-endorphin fibril²⁷ was stable under an acid environment with a pH of 5.5 or lower.

Progressive deprotonation of Glu8 at neutral pH triggers β-endorphin fibril disruption.

Using the equilibrium structural ensemble derived from MD simulations, Glu8 residues in the second outer layer chains (chains 2 & 9, Fig. 2j) had reduced pKa values below 7 compared to the predicted values using the static fibril structure (Fig. 1c). Therefore, upon entering the neutral pH environment, the β-endorphin fibril stable at pH 5.5 with the only outer layer Glu8 deprotonated is expected to have Glu8 residues in the second outer layer also deprotonated. We next simulated the β-endorphin fibril with Glu8 residues in both first and second outer layers deprotonated (i.e., Glu8 in chains 1, 2, 9, and 10; denoted as 40% deprotonation). The time evolution of fibril backbone RMSD (Fig. 3a) reached equilibrium around 0.44 nm, having a larger deviation from the initial state compared to systems with 0% and 20% Glu8 deprotonated (Fig. 2g). Final structures in both independent simulations featured partially disrupted fibril structures (Fig. 3g), especially the first β-strands in the four outer layer peptides with the corresponding Glu8 deprotonated (Fig. S2a–c). This observation resulted from electrostatic repulsion interactions of Glu8 in adjacent peptides in the two outer layers. Indeed, the time evolution of RMSD per chain showed that chains with deprotonated Glu8 featured higher RMSD than that of systems with 0% and 20% outer Glu8 residues deprotonated (Fig. 3dand Fig. 2b&e).

Fig. 3. The structural stability analysis of β-endorphin fibrils after switching to neutral pH.

a-c) The time evolution of backbone RMSD of the β-endorphin fibril with Glu8 residues deprotonated ratio of 40%, 60, and 100% (denoted as 40%, 60%, and 100%), respectively. For each molecular system, two independent 1000-ns MD simulations were performed. d-f) The backbone RMSD of each chain in the fibril in each simulated system. g-i) The final structures out of each simulation from every system. Chains with corresponding Glu8 deprotonated were colored pink. j) The probability distribution function of backbone RMSD with respect to the initial fibril structure. Only structures from the last 200 ns simulations were used for the analysis. k) The average propensities of each secondary structure (including random coil, β-sheet, helix, bend and turn) and (l) the probability of each residue to form β-sheet in each simulated system. (m) The averaged pKa prediction for Glu8 residues in each chain based on conformation ensembles sampled in the last 50 ns.

With structural changes accompanying the deprotonation of Glu8 residues in the first two outer layer peptides, we re-calculated the pKa values of Glu8 using the structural ensemble from the equilibrated MD simulations. Compared to initial pKa values above 8 for adjacent Glu8 residues in the 3rd outer layer (e.g., chains 3 & 8 in Fig. 2j), the pKa values decreased close to ~7.0 (Fig. 3m), indicating significantly increased probability of these two glutamates to be deprotonated at the neutral pH. To test whether this observation led to an iterative deprotonation of adjacent Glu8 and subsequent structural disruption, we also performed MD simulations with first three outer layer Glu8 residues protonated (i.e., chains 1–3 and 8–10, or 60% deprotonated). As expected, the β-endorphin fibril structure was further destabilized with the backbone RMSD saturated around ~0.5 nm and increased β-sheet loss in the β1 region (Fig. 3b,j,l and Fig. S2). For example, the average β-sheet probability decreased to ~40% in β1 by structural conversion into random coil. The β-sheets formed by β2 and β3 were still retained (Fig.S2a–d). Moreover, the recalculated pKa values of all Glu8 residues in chains 1–3 and 8–10 were ~4.0–5.5 after reaching equilibrium, suggesting that the sampled structural ensemble with all these Glu8 residues protonated was the thermodynamically stable state at neutral pH (Fig. 3m). The pKa values of other inner Glu8 residues also significantly reduced. Together, our simulation results suggested an iterative deprotonation of Glu8 residues and structural disruptions of the β-endorphin fibril at neutral pH until all Glu8s become deprotonated and negatively charged. The progressive deprotonation of Glu8 residues and structural disruptions agree with the prior experimental study of pH-triggered release of β-endorphin, where the disaggregation rate of the sonicated β-endorphin fibrils was much faster than the non-sonicated long fibril²⁷. Although a complete dissociation of the preformed fibril into monomers was not observed in our all-atom MD simulations, the observed cascades suggested a cooperative dissociation of fibrils. Physiologically, the β-endorphin usually reached the maximal level in the blood within 20~60 min upon stimulation by physical exercise^{69, 70}, which was much faster than the average duration of protein synthesis in eukaryotes (2−3 h)²³. Further studies using either multiscale simulations or enhanced samplings are necessary to reach long timescales and overcome the large energy barriers associated with fibril dissociation.

Next, we also studied the fibril stability with all of the Glu8 deprotonated (i.e., 100% deprotonation). The back RMSD with respect to initial ideal fibril structure reached ~ 0.60 nm after 1000 ns in both independent MD simulations (Fig. 3c). The structural deviation was much larger than simulations with fractions of Glu8 deprotonated (Figs. 2g, 3j). Importantly, the slowly increasing trends of overall RMSD in both simulations also suggested that the equilibrium was yet to be achieved due to a large energy barrier expected for the complete disaggregation of the fibril. Such a trend of increasing fibril structure disruption was also obvious in the time evolution of backbone RMSD per chain (Fig. 3f). Interestingly, the largest structural deviation was in the middle of the fibril, chain 6. The representative final snapshots further confirmed that the fibril structure was significantly destabilized (Fig. 3i). The calculation of RMSD of each of three β-strands revealed that the β1 was significantly destroyed, although β2 and β3 still retained some of β-sheet structures (Fig. S2). The analysis of secondary structures (Fig. 3k) showed ~10% β-sheet content converted into random coils compared to fibril simulation in the acid environment (Fig. 2h). The plot of averaged β-sheet and random coil propensity per residue demonstrated the β-sheet structure in β1 region was significantly lost due to the deprotonation of Glu8 (Fig. 3l and Fig. S2). Using structures in the last 50 ns, the pKa of all Glu8 decreased less than 5.0 due to the structural disruption of the β1 region, conforming that the state with fully deprotonated Glu8 was the thermodynamically state at neutral pH (Fig. 3m)

The deprotonated Glu8 residues disrupted the hydrophilic core of β-endorphin fibril.

In the fibril core (Fig. 1b), interactions between residues in the hydrophobic cluster areexpected to stabilize the fibril while the hydrophobic cluster helps solvate the Glu8 and determines the reversibility of the fibril. To characterize the solvent accessibility of these residues, the averaged solvent accessible surface area (SASA) of each these residues in simulations of different molecular systems was also analyzed (Figs. 4 and S3). The pKa of Glu8 residues correlated well with their SASA (Fig. 4a,b). Under acid conditions (pH ≤ 5.5), all the inner Glu8 residues buried with SASA equal to 0 were protonated, while the two outer Glu8 residues with largest SASA are deprotonated. At neutral pH environment, partial solvent exposure of adjacent Glu8 with small increase of SASA from 0 resulted into significant decrease of pKa from above 7.0 to lower values, and thus, deprotonation. Eventually, all the Glu8 residues became solvent exposed and correspondingly deprotonated due to the significant structural disruption of β1 region (Fig. 4c) caused by the electrostatic repulsion of charged Glu8s and the corresponding high solvation propensity. As a result, the solvent exposure of other buried residues in β1 – Phe4, Thr6, and Ser10 – increased with increasing Glu8 deprotonation ratio (Fig. S3a). Although the β2 sheet wasn’t destroyed at neutral pH in our simulations (Fig. 3l), SASA of residues Leu14, Thr16 Phe18 in β2 also increased. For β3, the deprotonation of Glu8 only rendered Ala26 more exposed, but didn’t alter the SASA of Ala21 and Ile23 from the region (Fig. S3c). Hence, the SASA analysis confirmed that the deprotonation of Glu8 destructed the cross-β core by exposing the buried residues in the hydrophilic and hydrophobic clusters.

Figure. 4. The conformational analysis caused by progressive deprotonation of Glu8.

a) The pKa of Glu8 in each chain predicted using equilibrated conformations in simulations with increasing deprotonation. b) The average solvent accessible surface area (SASA) of Glu8 in each chain based on the equilibrated conformations. c) The average β-sheet propensity of each residue from β-endorphin peptide. d) The representative final structures of β-endorphin fibril with the Glu8 deprotonated ratio of 0%, 60%, and 100%. Glu8 residues are shown in spheres.

The E8Q and E8L mutants didn’t alter the structure but lost the pH-responsive reversibility of β-endorphin fibrils.

Prior experiments showed that the disassembly of β-endorphin fibrils triggered by pH change from 5.5 to 7.4 or higher was significantly inhibited when the Glu8 mutated to non-ionizable residues – polar Gln or hydrophobic Leu (denoted as E8Q and E8L, respectively)²⁷. We also performed MD simulations of the two mutant fibrils (Fig. 5) to evaluate the effects of single-point mutations, E8Q and E8L, on the fibril stability. Both mutant fibril structures were well maintained during 1000-ns simulations with RMSD fluctuating around ~0.4 nm (Fig. 5a), which was similar to the wild-type simulation results at a pH value of 5.5 or lower (Fig. 2). The average SASA computed using the equilibrated conformations revealed that the mutated residues in the inner chains of the corresponding mutant fibrils remained buried (Fig. 5b,c). Since there are no ionizable residues in the fibril within the pH range of 5.5–7.4, the simulation results suggest that the mutant fibrils remain stable over the pH change from acidic 5.5 to neutral 7.4. In other words, the E8Q and E8L mutations didn’t alter the structure but lost the pH-responsive reversibility of β-endorphin fibril. The results confirmed that ionizable residues buried inside of fibril played important role in the reversible aggregation of β-endorphin.

Fig. 5. The structural stability analysis of E8Q and E8L β-endorphin mutant fibrils.

a) The time evolution of backbone RMSD of two mutant fibrils. For each type of mutant fibrils, a 1000-ns MD simulation was performed. b) The average solvent accessible surface area (SASA) of residue Q8 and L8 in the last 200 ns of corresponding simulations. c) The final structures of the mutant fibrils in MD simulations. Gln8 and Leu8 in each chain are shown in spheres.

Positioning ionizable residues inside the cross-β core as a common strategy to render pH-responsive reversible fibrils.

It has been well established that charged amino acids like glutamate in globular proteins are predominantly found on the solvent-exposed surfaces, and only occasionally buried in the hydrophobic core paring with opposite charges (i.e., forming salt-bridges)^71–74. With the expense of reduced protein stability, unpaired glutamates in protein cores are evolutionally selected to play important functional roles, such as catalysis and pH-dependent conformational changes. Similar to protein folding, the self-assembly of amyloid fibrils is also driven by hydrophobic interactions and in the case of β-endorphin the strategic positioning of an unpaired glutamate in the core renders its fibrils pH-responsive and reversible by reducing the fibril stability. After entering the circulation with neutral pH 7.4, β-endorphin fibrils formed at low pH for efficient storage release the peptide hormone as intrinsically disordered monomers containing partial helices⁷⁵ (Fig. 6a left panel). Such environment-dependent reversible fibrillization is especially important for naturally-occurring functional amyloids so that the fibrillization process can be modulated by changing environment such as pH and the composite proteins or peptides can be reused. For example, suckerin, a squid ring teeth protein, could self-assemble into well-ordered fibrils in the neutral pH environment and the pre-formed fibrils disaggregate in the acidic conditions^{39, 76}. Our previous computational study demonstrated that the protonation of histidine residues buried in the cross-β core is important for destabilizing the fibril at low pH⁴⁰ (Fig. 6a rightpanel). Another example is pre-melanosomal protein Pmel17^{6, 37, 38}. The aggregation of Pmel17 repeats only occurred at the conditions of pH lower than 5, and the preformed fibrils rapidly dissociated under neutral pH solution condition^{37, 77}. Although Pmel17 fibril structure is still unknown, each Pmel17 repeat contained at least one glutamate (shown in Fig. 6b), and thus, the corresponding fibrils likely have unpaired glutamates buried in the cross-β core, also making the fibrils pH-responsive and reversible. Hence, positioning ionizable residues inside the cross-β core serves a common strategy to render pH-responsive and reversible fibrils for naturally-occurring functional amyloid.

Fig. 6. The schematic diagram illustrates that burying ionizable amino acids as a common strategy to render pH-responsive reversible fibrils for functional amyloids.

a) One example is the β-endorphin fibrils containing a buried glutamate, which is stable at low pH but disaggregate in neutral pH. The other example is suckerin A1H1 peptide with histidines buried in the core, which could self-assemble into fibrils under neutral pH but disaggregate in the acidic environment⁴⁰. b) The amino acid sequences of Pmel17 repeats, also forming pH-responsive functional amyloids. The ionizable glutamates are highlighted in red.

The pathological aggregations are irreversible because the corresponding fibrils are highly stable and stay intact even under highly denaturing conditions. Whether and how the disease-related amyloid peptides could be rendered pH-sensitive by strategically introducing one or more ionizable residues to the beta-sheet core are still need future computational and experimental exploration. Nevertheless, the pH-responsive and reversible properties of these functional amyloids have great application potentials, such as controlled drug delivery and release⁷⁸, regulating the drug release rate by the pH change^{39, 79}. Reversible fibrillization also enables promising applications in 3D printing and tissue regeneration. Therefore, introducing ionizable groups inside the cross-β core could allow the design of reversible amyloid fibrils as pH-responsive smart bio-nanomaterials.

Conclusion

In sum, we investigated the structural stability of β-endorphin fibril using atomistic MD simulations along with structure-based pKa predication. Our result showed that the experimentally determined amyloid fibril²⁷ was stable at acidic conditions with a pH of 5.5 or lower. After switching to the neutral pH, residues Glu8 from the outer layers along the fibrils would be deprotonated and cause sheet-to-coil conversion in the β1 region, which could make adjacent Glu8 residues in the inner layers partially exposed and subsequently deprotonated. Eventually, all Glu8 residues were progressively deprotonated and the fibril structures were significantly destabilized. Although a complete fibril disaggregation and monomer release were not observed in our simulation due to a high free energy barrier and long timescales, our computational study demonstrated the early events for the pH-triggered fibril disassembly and uncovered the key role of the buried Glu8 in the pH-responsive reversible fibrillization of β-endorphin. Introducing ionizable residues into the cross-β core might is, therefore, a potential approach to design reversible amyloid fibrils as pH-responsive smart bio-nanomaterials for drug delivery, 3D printing and tissue regeneration.