Cite this Article
Deng Li, Zhao Yurong, Zhou Peng, Xu Hai, Wang Yanting. Modulation of intra- and inter-sheet interactions in short peptide self-assembly by acetonitrile in aqueous solution. Chinese Physics B, 2016, 25(12): 128704  
Permissions
Modulation of intra- and inter-sheet interactions in short peptide self-assembly by acetonitrile in aqueous solution
Deng Li1, Zhao Yurong1, Zhou Peng1, Xu Hai1, †, , Wang Yanting2, 3, ‡,
Center for Bioengineering and Biotechnology, China University of Petroleum (East China), Qingdao 266580, China
CAS Key Laboratory of Theoretical Physics, Institute of Theoretical Physics, Chinese Academy of Sciences (CAS), Beijing 100190, China
School of Physical Sciences, University of Chinese Academy of Sciences, Beijing 100049, China

 

† Corresponding author. E-mail: xuh@upc.edu.cn

‡ Corresponding author. E-mail: wangyt@itp.ac.cn

Project supported by the National Basic Research Program of China (Grant No. 2013CB932804), the National Natural Science Foundation of China (Grant Nos. 91227115, 11421063, 11504431, and 21503275), the Fundamental Research Funds for Central Universities of China (Grant No. 15CX02025A), and the Application Research Foundation for Post-doctoral Scientists of Qingdao City, China (Grant No. T1404096).

Abstract
Abstract

Besides our previous experimental discovery (Zhao Y R, et al. 2015 Langmuir, 31, 12975) that acetonitrile (ACN) can tune the morphological features of nanostructures self-assembled by short peptides KIIIIK (KI4K) in aqueous solution, further experiments reported in this work demonstrate that ACN can also tune the mass of the self-assembled nanostructures. To understand the microscopic mechanism how ACN molecules interfere peptide self-assembly process, we conducted a series of molecular dynamics simulations on a monomer, a cross-β sheet structure, and a proto-fibril of KI4K in pure water, pure ACN, and ACN-water mixtures, respectively. The simulation results indicate that ACN enhances the intra-sheet interaction dominated by the hydrogen bonding (H-bonding) interactions between peptide backbones, but weakens the inter-sheet interaction dominated by the interactions between hydrophobic side chains. Through analyzing the correlations between different groups of solvent and peptides and the solvent behaviors around the proto-fibril, we have found that both the polar and nonpolar groups of ACN play significant roles in causing the opposite effects on intermolecular interactions among peptides. The weaker correlation of the polar group of ACN than water molecule with the peptide backbone enhances H-bonding interactions between peptides in the proto-fibril. The stronger correlation of the nonpolar group of ACN than water molecule with the peptide side chain leads to the accumulation of ACN molecules around the proto-fibril with their hydrophilic groups exposed to water, which in turn allows more water molecules close to the proto-fibril surface and weakens the inter-sheet interactions. The two opposite effects caused by ACN form a microscopic mechanism clearly explaining our experimental observations.

PACS: 87.10.Tf;87.14.ef;87.15.bk
Keyword:solvent effect;peptide self-assembly;molecular dynamics simulation
1. Introduction

In view of the promising applications of peptide self-assembly in biomaterials and biomedicines,[1,2] the underlying microscopic mechanisms have been extensively studied, aiming to develop novel methods to control the self-assembly process and tune the self-assembled morphology. Furthermore, the fibrillar aggregates formed by peptide self-assembly are known to be closely related to many neurodegenerative diseases, such as Alzheimer’s, Huntington’s, and Parkinson’s.[3,4] Although peptides in different diseases have different amino acid sequences, their fibrillar aggregates share a common feature, i.e., a highly ordered cross-β sheet structure, in which peptide backbones are aligned along the direction perpendicular to the long axis of fibrils.[5–7] In addition to the ordered fibrils, other ordered supramolecular nanostructures such as twisted ribbons, helical ribbons, and nanotubes have also been observed.[8–11] Different self-assembled nanostructures possess different physicochemical properties and technological potentials. As a result, unraveling the microscopic mechanisms of different ordered structures formed by peptide self-assembly is not only essential for understanding the origin of neurodegenerative diseases but also favoring the manipulation of self-assembled peptide nanostructures for different applications.

Just like protein folding whose process is hierarchical,[12,13] the peptide self-assembly process to form supramolecular nanostructures also spans multiple spatial and temporal scales. Aggeli et al.[5,14,15] proposed a hierarchical peptide self-assembly model, in which the peptide fibril corresponds to a lamellar structure composed of cross-β sheets; the growth along the long axis of the fibril is governed by the intra-sheet H-bonding interactions between neighboring backbones; and its width growth is determined by the inter-sheet hydrophobic and/or electrostatic interactions between neighboring cross-β sheets. The width growth stops when the system reaches an energetic balance between the attraction of neighboring cross-β sheets and the twisting of the cross-β sheets. Fiber-XRD and SANS experiments on many peptide self-assembled nanostructures confirmed the existence of the lamellar structure proposed in the hierarchical model.[16,17] In addition to fibrils, lamellar structures were also found in the formation of nanotubes, where other ordered supramolecular nanostructures such as thin fibers, twisted ribbons, and helical ribbons have been observed by acting as intermediates.[11,18,19] Furthermore, by analyzing the correlation between the width and the morphology of supramolecular nanostructures in peptide self-assembly, Childers et al.[20] demonstrated that the width has a direct relation with the final self-assembling morphology: when the width of the self-assembly is larger than 40 nm, the morphology is a helical ribbon or nanotube; but when it is smaller than 40 nm, the morphology is a twisted ribbon. The above findings imply that the competition between the intra- and inter-sheet interactions in lamellar structures determines the final morphology of supramolecular nanostructures.

Peptide intra- and inter-sheet interactions consist of non-covalent interactions including hydrophobic, electrostatic, van der Waals (VDW), π–π stacking, and H-bonding interactions,[21,22] which could possibly be tuned by varying peptide sequence, such as the number and types of hydrophobic residues as well as the pattern of charged residues along the backbone.[23–25] However, the change of peptide sequence may have comprehensive influences on various non-covalent interactions, so it is usually not a good way to fine-tune individual interactions. Instead, a better way is changing solution conditions, such as pH, ion strength, and content of small organic molecules. Because the side chains with – COOH or −NH2 groups can be in either a protonated or a deprotonated state, the electrostatic interactions between peptides can be tuned by changing solution pH values. For instance, the morphologies of self-assembled nanostructures of short Aβ(16–22) depend on solution pH values: nanofibers at pH = 6 and nanotubes/nanoribbons at pH = 2.[26] The strength of the electrostatic interactions can also be tuned by changing ion strength. One example is that Adamcik et al.[27] observed untwisting of fibrils formed by β-lactoglobulin in aqueous solution with increasing NaCl concentration. Recently, many works have been devoted to tune non-covalent interactions by adding small organic molecules into aqueous solution. Small organic molecules modulate not only electrostatic interactions by changing the solution polarity at the macroscopic level, but also hydrophobic and H-bonding interactions by directly varying the interactions among water, small organic molecules, and peptides at the microscopic level. For instance, after β-lactoglobulin fibrils were incubated in 40 w% ethanol for 5 weeks, Jorden et al.[28] observed a disassembly of fibrils, which was attributed to the decrease of the dielectric constant of the solution caused by adding ethanol. By increasing the methanol content in solution, Castelletto et al.[29] demonstrated a morphological transformation of peptide AAKLVFF from nanofibrils to nanotubes due to the variations of the H-bonding pattern inside β-sheets and the stacking pattern between side chains by methanol.

Unraveling the roles of microscopic intra- and inter-sheet interactions involved in peptide self-assembly will be highly valuable for the realization of tuning peptide self-assembling morphology by changing physicochemical conditions. This goal is unlikely to be fulfilled solely by experiment because dynamics and intermolecular forces at the microscopic level are usually difficult to be determined in experiments. With the development of modern computer technology for modeling and simulation of biological molecules, molecular simulation techniques have been adopted to study the microscopic dynamics and interactions involved in peptide self-assembly.[30–34] In order to understand the influence of pH on the self-assembled morphology of peptide EAK16, Emamyari et al.[35] systematically studied conformation and dimerization of EAK16 monomers with differently charged states under different pH values by atomistic molecular dynamics (MD) simulations, and found that both the charge pattern and size of side chains play an important role in the early stages of self-assembly. Detailed simulations for Aβ peptide fragments and model peptide Ac-GGAGG-NH2 in different solutions suggested that monomer conformation is mainly determined by the competition between the H-bonding interactions between backbones and solvents and those between backbones and backbones.[36–38] By conducting computer simulations on Aβ16–22 monomer and oligomer in the solutions with different urea contents, Klimov et al.[39] found that the electrostatic interactions between solvent and peptide backbones not only determine the monomer conformation, but also have a strong influence on the interactions between the strands which stabilize oligomers in water. By performing simulations on the aggregation of FF and AA in water and methanol, respectively, Rissanou et al.[40] observed that both peptides have a weaker propensity of self-assembling in methanol than in water, which can be attributed to the weaker H-bonding interactions between peptide strands in methanol.

We had previously observed by experiment that nanotubes self-assembled by KI4K in water can gradually transform to nanofibrils with increasing ACN content in aqueous solution.[41] Further experiments with more cases, as reported in Supplementary material, confirmed that observation. As shown in Fig. S1 in Supplementary material, AFM images demonstrated that KI4K molecules form nanotubes with diameters of 80 nm ∼160 nm in pure water, helical nanoribbons with widths of 30 nm ∼90 nm in 20-vol% ACN, nanofibrils with decreased widths of about 18 nm in 40-vol% ACN, and twisted nanotapes with a width of about 10 nm in 80-vol% ACN. Furthermore, as shown in Fig. S2 in Supplementary material, the SANS intensity of self-assemblies in pure water, 20-vol% ACN, and 80-vol% ACN is significantly stronger than that in 40-vol% ACN, indicating that the mass of self-assemblies does not monotonically change with increasing ACN in solution. These results mean that the ACN in solution can tune not only the morphology but also the mass of the KI4K self-assembled structure.

The experimental time scale for peptides to self-assemble into ordered states with cross-β sheets starting from disordered states is about a few hours,[11,42] but the time scale of solvent relaxation around peptide is about hundred picoseconds, as demonstrated by simulations.[43,44] It is impractical to simulate the whole self-assembly process by all-atom MD simulation, so we have conducted a series of all-atom MD simulations for a KI4K monomer, a KI4K cross-β sheet with six strands, and a KI4K proto-fibril with four cross-β sheets in different solutions to elucidate the microscopic mechanisms of ACN’s effects on the intra- and inter-sheet interactions within the framework of the hierarchical peptide self-assembly theory.[15,45] The simulation results for the monomer in different solutions demonstrate that the interactions of ACN with different groups of KI4K are different from those of water due to the amphiphilic feature of ACN. The results of the cross-β sheet demonstrate that the intra-sheet interactions between neighboring strands are strengthened by ACN in aqueous solutions. The results of the proto-fibril confirm that the intra-sheet interactions are strengthened by ACN but the inter-sheet interactions are weakened by ACN in aqueous solutions. By analyzing the correlation between different groups and the behavior of solvent around the proto-fibril surface, we have found that polar and nonpolar atomic groups of ACN molecules play different roles in affecting the self-assembled nanostructure: the weaker correlation between the backbone and the polar group of ACN enhances the intra-sheet interactions mainly composed of H-bonding interactions between backbones, but the stronger correlation between the hydrophobic side chain and the nonpolar group of ACN weakens the inter-sheet interactions mainly composed of hydrophobic interactions between β-sheets in pure water. The opposite effects of adding ACN on intra- and inter-sheet interactions revealed by simulation provide a clear microscopic mechanism for understanding the experimentally observation decreasing of widths and turnover of masses of KI4K self-assemblies with increasing ACN content.

2. Simulation methods

In this work, all-atom MD simulations of a KI4K monomer, a cross-β sheet, and a proto-fibril in different solutions were conducted by using the GROMACS software.[46] Both peptides and ACN molecules were modeled by the OPLS-AA force field,[47–49] and water molecules were modeled by the TIP4P force field.[50] The cross-β sheet structure with six strands was constructed according to the optimized trimer structure composed of an anti-parallel and a parallel β-sheet substructure, determined by replica exchange MD simulation.[51] The proto-fibril contains four such cross-β sheets with neighboring ones parallel to each other, since there are repulsive electrostatic interactions between charged lysines and attractive hydrophobic interactions between nonpolar isoleucines. The molecular models of ACN, water, KI4K, the cross-β sheet and the proto-fibril are shown in Fig. 1. Since the unit for the ACN content is represented by volume ratio in experiment but molar ratio in simulation, the relations between the two kinds of ratios as well as the numbers of components in different simulated systems are listed in Table S1 in Supplementary material. The lysine residues were protonated in the solutions with water and Cl− ions were added to neutralize the simulated system, while non-protonated lysines were simulated in pure ACN. In all the simulations, the system was firstly energetically minimized by the steepest decent method, and then the solvent was relaxed for 0.5 ns in the NPT [P = 1 atm (1 atm = 1.01325×105 Pa) and T = 300 K] ensemble with the peptide positions fixed. The MD simulations of the monomer in different solutions were conducted for 100 ns to ensure adequate sampling for analyzing the correlation between solvent and peptide molecules. The cross-β sheet and the proto-fibril were also simulated for 100 ns in each case to exam their stability in pure water. The cross-β sheet and the proto-fibril in the solutions with ACN were simulated for 10 ns in each case to study the effect of ACN on intra- and inter-sheet interactions. All simulations were conducted in the NPT ensemble with the temperature T = 300 K and the pressure P = 1 atm. The electrostatic interactions were treated with the particle-mesh-Ewald method.[52,53] The cutoffs for the VDW interactions and the real part of the electrostatic interactions were both set to be 1.2 nm. The time step was 1 fs and 1000 instantaneous configurations were evenly sampled.

Fig. 1. Molecular structures of ACN, water, KI4K, and proto-fibril.
3. Results and discussions
3.1. ACN strengthens the β-strand propensity of monomer

In peptide self-assembly, the propensity of a monomer to form a β-strand is the key to form a cross-β structure. We characterized the β-strand formation propensity of KI4K by calculating the end-to-end distance, defined as the distance between the Cα atoms of the two lysine residues at both ends. The probability distributions of the end-to-end distance in different solutions (pure water, ACN/Water mixture with 50% molar ratio of ACN, and pure ACN) are shown in Fig. 2. There exist two peaks at 1.3 nm and 1.55 nm for the monomer in the solution of pure water, but only one peak at 1.5 nm in the two solutions with ACN. The peak at 1.5 nm in pure ACN is much higher than that in the mixture. The average value of the end-to-end distance in pure water is 1.38 nm, smaller than 1.40 nm in ACN/Water and 1.48 nm in pure ACN. These results demonstrate that the conformation of KI4K in solution with ACN is more extended than that in pure water, i.e., the addition of ACN into aqueous solution strengthens the β-strand formation propensity of KI4K.

Fig. 2. Probability distribution of the end-to-end distance of KI4K at different solutions (Water, ACN, and ACN/Water mixture).

H-bonding interactions between peptides and solvent molecules were known to have a strong influence on the conformation of peptide monomer.[36–38] In our data analysis, we regard an H-bond as formed when the distance between the hydrogen donor and acceptor is less than 3.5 Å, and simultaneously, the angle between the vector from the donor to its hydrogen atom and the vector from the donor to the acceptor is less than 30°. The average numbers of the H-bonds formed between the peptide and solvent molecules in different solutions are given in Table 1. With increasing ACN content, the average total number of H-bonds decreases: 19.2 for pure water, 13.6 for ACN/Water, and 6.1 for pure ACN. In the mixture solution, the number of the H-bonds of backbone with ACN is 2.5, smaller than 6.4 with water; and the number of the H-bonds of side chain with ACN is 1.3, also smaller than 3.3 with water. The H-bond number between peptide and water is 19.2 in pure water and 9.7 in the ACN/water mixture, demonstrating that the H-bonding interactions between peptide and water is reduced by ∼49% when ACN is added in pure water. The H-bond number between peptide and ACN is 6.1 in pure ACN and 3.9 in the mixture, demonstrating that the H-bonding interactions between peptide and ACN is reduced by ∼ 36% when water is added into pure ACN. These results demonstrate that, although the number of H-bonds between ACN and peptide is smaller than that between water and peptide possibly because a water molecule can act as both an H-bond acceptor and a donor but an ACN molecule can only be an acceptor, the strength of the H-bonding interaction between ACN and peptide is stronger than that between water and peptide, implying that there exist additional interactions between peptide and ACN besides H-bonding.

Table 1.

H-bond numbers formed between peptide and solvent molecules.

.
3.2. ACN strengthens the intra-sheet interactions of the β-sheet

A 100-ns all-atom MD simulation was performed for the cross-β sheet structure in pure water to ensure that the system was well equilibrated without the memory of the manually constructed initial configuration. The root mean square deviation (RMSD) of the sampled structures with respect to the final structure in the trajectory and the H-bond number between peptides have been calculated and are shown in Fig. 3. It can be seen that, after 5 ns, the initial cross-β sheet structure with a RMSD of about 0.5 nm relaxes to a more stable structure with a RMSD of about 0.2 nm (Fig. 3(a)), and the H-bond number in the cross-β sheet structure fluctuates around 26 (Fig. 3(b)). The stable RMSD value and the H-bond number demonstrate that the relaxed cross-β sheet structure can be stabilized in pure water for a long time.

Fig. 3. Time evolutions of the RMSD relative to the final structure (a) and the H-bond number (b) of the cross-β sheet in pure water.

To understand the influence of ACN on the cross-β sheet structure, we conducted 10-ns simulations for the cross-β sheet structure in pure water, pure ACN, and the mixture with 50% molar ratio of ACN. The distributions of the H-bond numbers in the cross-β sheets in different solutions have been calculated and are shown in Fig. 4(a). It can be seen that the peak value shifts from 26 for pure water to 27 for the mixture, and finally to 29 for pure ACN, indicating that the addition of ACN in solution promotes H-bond formation between peptides, i.e., ACN strengthens intra-sheet interactions. To investigate the effect of strengthened intra-sheet interactions on the cross-β sheet structure, the average twisting angles between neighboring strands were calculated. The distribution of the average twisting angles in Fig. 4(b) has a peak shift from 0.28 rad for pure water and the mixture to 0.26 rad for pure ACN, indicating that ACN can slightly weaken the twisting degree of the cross-β sheet.

Fig. 4. Probability distributions of H-bond number (a) and twisting angle between neighboring strands (b) in β-sheets in different solutions.
3.3. Opposite effects of ACN on intra- and inter-sheet interactions of the proto-fibril

To study the effects of ACN on inter-sheet interactions, a proto-fibril with four cross-β sheets was manually constructed and a 100-ns all-atom MD simulation was performed to assess its stability in pure water. The evolution of the RMSD with respect to the final configuration, radius of gyration (Rg), inter-sheet distance (ISD), and hydrophobic area of the proto-fibril were calculated to quantify the dynamics of the proto-fibril in pure water. In our calculations, the ISD is defined as the average distances between the mass centers of neighboring sheets in the proto-fibril.

The figure 5(a) shows that, after relaxing for 20 ns, the RMSD of the proto-fibril decreases from 0.72 nm for the initial structure to smaller than 0.3 nm. By contrast, it only takes about 2 ns for the Rg and ISD decrease from 1.78 nm to 1.67 nm (Fig. 5(b)) and from 1.33 nm to 1.24 nm (Fig. 5(c)), although there still exist small fluctuations for Rg (between 1.67 nm to 1.62 nm) and ISD (from 1.24 nm to 1.17 nm), respectively. Consequently, the hydrophobic area of the proto-fibril decrease from 85 nm2 to 70 nm2 and keep unchanged after about 2 ns (Fig. 5(d)). These results demonstrate that the proto-fibril becomes more compact and stable in pure water after relaxation and the interactions between hydrophobic side chains drive the association of β-sheets, but there still exist structural relaxation for the associated inter-sheet structure for a longer timescale.

Fig. 5. Time evolutions of the RMSD with respect to the final configuration (a), radius of gyration (b), average inter-sheet distance (c), and hydrophobic area of the proto-fibril (d) in pure water.

Considering the limitation of computational resource and the timescale for association for strand and cross-β sheet structure, we further performed all-atom MD simulations for the proto-fibril in solutions with different ACN contents and studied the effects of ACN on intra- and inter-sheet interactions. Figure 6 shows the evolutions of the proto-fibril structure in different solutions. As time goes by, the proto-fibril in pure water has its structure relatively unchanged, but the cross-β sheets of the proto-fibril in pure ACN and in the mixture more or less separated from their neighbors, indicating that ACN weakens inter-sheet interactions between cross-β sheets, and consequently weakens the width-growth propensity of the supramolecular nanostructure, consistent with our experimental observations provided in Supplementary material.

Fig. 6. Snapshots of the simulated proto-fibrils in different solutions. Significant separations of neighboring β-sheets are marked by red rectangles.

To quantify the influence of ACN on intra- and inter-sheet interactions in the proto-fibril, the hydrophobic contact numbers and the inter-molecular H-bond numbers between β-strands in the four cross-β sheets were calculated for the solutions with different ACN contents. Two hydrophobic side chains in different residues are considered as in contact if the distance between two carbon atoms (CG2) is close to 6 Å. The average H-bond numbers and hydrophobic contact numbers in different solutions are shown in Fig. 7. Figure 7(a) demonstrates that the H-bond number increases with increasing ACN content, indicating that ACN strengthens the intra-sheet H-bonding interactions between neighboring backbones, consistent with the results in Fig. 4(a). By contrast, the number of hydrophobic contacts in the proto-fibril decreases with increasing ACN content (Fig. 7(b)), indicating that ACN can reduce the inter-sheet interactions between hydrophobic side chains in neighboring cross-β sheets.

Fig. 7. Average H-bond numbers (a) and hydrophobic contacts (b) in solutions with different ACN contents.
3.4. Opposite effects of polar and nonpolar groups of ACN on KI4K monomer

To understand why the β-strand propensity of KI4K monomer is enhanced in ACN and the ACN/water mixture, the radial distribution functions (RDFs) have been calculated to characterize the density correlations between different groups of solvent and peptide molecules, as shown in Fig. 8. The oxygen atom (OW) is used to represent the water molecule, the nitrogen atom (NZ) of an ACN molecule is used to represent its polar group, and the carbon atom (CT) to represent its nonpolar group. For peptide KI4K, the carbon atom (CG2) on the isoleucine side chain is used to represent its nonpolar group, and the oxygen (O) and nitrogen (N) atoms on the peptide backbone are used to represent its polar group. The molecular positions of these atoms are shown in Fig. 1.

Fig. 8. RDFs between different groups of solvent and peptide monomer. (a) RDFs between water and peptide molecules in pure water. (b) RDFs between water and peptide molecules in the ACN/water mixture. (c) RDFs between ACN and peptide molecules in pure ACN. (d) RDFs between ACN and peptide molecules in the ACN/water mixture. OW represents the oxygen atom of water, CT the methyl carbon of ACN, NZ the nitrogen of ACN, CG2 the side chain carbon of isoleucine, N the backbone nitrogen of isoleucine, and O the backbone oxygen of isoleucine. All of these atoms have been labeled in Fig. 1.

Figures 8(a) and 8(b) demonstrate that the first peaks of the RDFs between peptide and water are at some 0.3 nm for both gOW−N(r) and gOW−O(r), but the peak height of gOW−N(r) is lower than that of gOW−O(r). This indicates that water molecules form H-bonds with both the backbone nitrogen and oxygen atoms, but the former are weaker than the latter. From Figs. 8(c) and 8(d), we can see that the first peak of gNZ−N(r) is at some 0.3 nm and that of gNZ−O(r) is at some 0.4 nm. Furthermore, the first peak height of gNZ−N(r) is larger than that of gNZ−O(r). This indicates that acetonitrile molecules only form H-bonds with backbone nitrogens because they can only be H-bond acceptors. The first peaks of gOW−N(r) and gOW−O(r) at some 0.3 nm in Fig. 8(b) are higher than those in Fig. 8(a), but the first peak of gOW−CG2(r) at some 0.4 nm in the mixture solution is lower than that in pure water, indicating that the addition of ACN strengthens the correlation between water and the backbone but weakens the correlation between water and the hydrophobic side chain. At the same time, the first peaks of gNZ−N(r) and gNZ−O(r) in Fig. 8(d) are lower than those in Fig. 8(c), and the first peak of gCT−CG2(r) at some 0.4 nm in Fig. 8(d) is higher than that in Fig. 8(c), indicating that the introduction of water weakens the correlation between ACN and the backbone but strengthens the correlation between ACN and the hydrophobic side chains. In the mixture solution, the fact that the first peak of gOW−O(r) in Fig. 8(b) is higher than that of gNZ−N(r) in Fig. 8(d), together with the fact that there lacks a first peak of gNZ−O(r) at some 0.3 nm in Fig. 8(d), indicates that the correlation between water and backbone is stronger than that between ACN and backbone, consistent with the above analysis on the H-bonding interactions between peptide and solvent molecules (Table 1). The first peak of gOW−CG2(r) in Fig. 8(b) is lower than that of gCT−CG2(r) in Fig. 8(d), indicating that the correlation of the hydrophobic side chain with water is weaker than with ACN. Overall, these results manifest that ACN has two opposite effects on peptide monomer: the weaker interaction between ACN and the backbone strengthens the intra-backbone interaction (intramolecular H-bonding) and thus weakens the β-strand propensity of KI4K, but the favorable interaction between ACN and the hydrophobic side chain prevents the collapse of the side chains and thus strengthen the β-strand propensity of KI4K. As demonstrated by the end-to-end distribution of the KI4K monomer (Fig. 2), ACN strengthens the β-strand propensity of KI4K, indicating that the interactions between the methyl group of ACN and the hydrophobic side chain dominate the conformation of KI4K monomer in aqueous solutions.

To understand the interactions between solvent molecules and peptides in the proto-fibril simulations, the RDFs between different groups in different solutions have been calculated, as shown in Fig. 9. Comparing the first peak heights of gOW−CG2(r) in Fig. 9(a) and those of gCT−CG2(r) in Fig. 9(b), the correlation of the hydrophobic side chains of KI4K with the nonpolar groups of ACN is stronger than that with water; while comparing the first peak heights and positions of gOW−N(r) and gOW−O(r) in Figs. 9(c) and 9(e) with those of gNZ−N(r) and gNZ−O(r) in Figs. 9(d) and 9(f), the correlation of the peptide backbone with the polar group of ACN is weaker than that with water. These results are consistent with the RDFs for the KI4K monomer. As a result, when the ACN content increases in the mixture solution, the correlations between the peptide backbone and solvent are weakened, leading to stronger interactions between peptide backbones in the same cross-β sheet, which is advantageous for self-assembling along the H-bonding direction. At the same time, the interactions between cross-β sheets are weakened due to stronger correlations between the hydrophobic side chains and solvent, which is disadvantageous for the lateral stacking of cross-β sheets.

Fig. 9. RDFs between different groups of solvents and peptides in the proto-fibril simulations in different solutions: panels (a), (c), and (e) are the results between water and nonpolar side chain (gOW−CG2(r)), between water and nitrogen on backbone (gOW−N(r)), and between water and oxygen on backbone (gOW−O(r)), respectively; panels (b), (d), and (f) are the results between nonpolar group of ACN and nonpolar side chain (gCT−CG2(r)), between polar group of ACN and nitrogen on backbone (gNZ−N(r)), and between polar group of ACN and oxygen on backbone (gNZ−O(r)), respectively.

From the RDFs shown in Figs. 9(a), 9(c), and 9(e), we can see that, from 0.5 nm to 2.0 nm, the RDF values increase with increasing ACN content, indicating that the existence of ACN molecules allow water molecules to be close to the proto-fibril surface. From Figs. 9(b), 9(d), and 9(f), we can see that, from 0.5 nm to 1.5 nm, the RDF values decrease with increasing ACN content, indicating that the existence of water molecules also allow ACN molecules to be close to the proto-fibril. In other words, the distribution of ACN and water around the proto-fibril can be enhanced by each other.

3.5. Accumulation and orientation of solvent around the proto-fibril

Since the main driving force for lateral stacking of cross-β sheets is inter-sheet hydrophobic interaction, which is a collective effect caused by water,[54] we have calculated the surface RDFs of the polar and nonpolar groups of ACN and water to quantify the accumulation of solvent molecules around the proto-fibril. In our calculations, the water molecule and the polar and nonpolar groups of ACN are represented by OW, NZ, and CT, respectively, as labeled in Fig. 1. The surface RDF SX(r) (X = OW, NZ, CT) between solvent and the proto-fibril is defined as the ratio of the local number density of solvent around the proto-fibril to the average density of solvent, where the local number density is the number of solvent molecules in different shells around the proto-fibril divided by the shell volume, i.e., number/volume. The width of each shell was set to 0.1 nm. The area of each shell around the proto-fibril was approximated by the total accessible surface of the proto-fibril, which can be calculated with the help of a probe sphere whose radius is the distance between the shell and the proto-fibril. A typical shell around the proto-fibril is shown in Fig. 10(b), and shell areas in different solutions are listed in Table S2 in Supplementary material.

Fig. 10. (a) Surface RDFs of solvent molecules around the proto-fibril in different solvents. (b) A typical shell and illustration of the molecular orientation angle around the proto-fibril.

The surface RDFs in different solutions are shown in Fig. 10(a). The first peak position of the surface RDF of the polar group for pure ACN solution is at 0.3 nm, the same as that of water for pure water solution, and their peak heights are similar. For pure solutions, the first peak position of SNZ(r) for pure ACN solution is at 0.3 nm, the same as that of SOW(r) for pure water solution, and the peak heights of SNZ(r) and SOW(r) are similar. The first peak position of SCT(r) for pure ACN solution is at 0.4 nm, larger than that of SNZ(r), but the peak height of SCT(r) is larger than that of SNZ(r). The larger peak position of SCT(r) relative to SNZ(r) demonstrates that the polar group can be closer to the surface of the proto-fibril due to the H-bond formation between the proto-fibril and the polar group of ACN as well as the larger size of methyl in the nonpolar group of ACN. However, the higher peak of SCT(r) than SNZ(r) indicates that the nonpolar group of ACN has a stronger tendency to accumulate around the proto-fibril than the polar group of ACN. In the ACN/Water mixture, the first peak positions of the surface RDF for all groups retain the same as in pure solutions. However, the peak height of SOW(r) is significantly higher than for SNZ(r) and SCT(r), demonstrating that the accumulation of water around the proto-fibril surface is stronger than that of ACN in the mixture solution. Comparing the surface RDFs for the mixture solution with those for the pure solutions, we can see that the accumulation of the polar group of ACN has little change and the accumulation of the nonpolar group of ACN at 0.4 nm is weakened, but the accumulations of water molecules in the shells from 0.2 nm to 0.5 nm are significantly strengthened, indicating that ACN can assist water to wet the surface of the proto-fibril. The theoretical work of Patel et al.[55,56] suggested that a larger wetting degree corresponds to a weaker hydrophobic interaction between two surfaces. In our case, because the wetting of the proto-fibril is strengthened by adding ACN into the solution, the inter-sheet interaction (hydrophobic attraction) in pure water is reduced.

The above results demonstrate that the accumulation of different solvents around the proto-fibril surface is different from that in the bulk, consistent with the results for solvents around biomolecules reported by many groups,[31,57] but the associated microscopic solvent environment is unclear. We compare the orientation of solvent molecules on the proto-fibril surface with that in the bulk of different solutions. A solvent molecule is considered as on the surface if the minimum distance between the solvent molecule and the proto-fibril is smaller than 0.3 nm, and is in the bulk if the distance is larger than 1 nm. In our calculations, water and ACN molecules are still represented by OW and CT, as labeled in Fig. 1. The orientation of a solvent molecule is characterized by the angle θ between the solvent molecule direction (from the polar atom to the nonpolar atom) and the normal vector of the proto-fibril surface approximated by the direction from a designated atom of a solvent molecule to the mass center of the proto-fibril. For ACN, the solvent direction is from NZ to CT, and the normal vector of the proto-fibril surface is from CT to the mass center of the proto-fibril. For water, the solvent direction is from OW to the center of the two hydrogen atoms, and the normal vector of the proto-fibril surface is from OW to the mass center of the proto-fibril. The molecular orientation angles are schematically shown in Fig. 10(b). The orientations of ACN and water molecules at the surface and in the bulk in different solutions are shown in Fig. 11. From Figs. 11(a) and 11(c), we can see that in pure water and in pure ACN, the orientation distribution for solvent molecules on the surface of the proto-fibril has a left shift relative to the one in the bulk, which indicates that the hydrogen atoms of water and the methyl group of ACN are more apt to orient towards the proto-fibril surface. Figure 11(b) shows that the left shift of water orientation still exists in the mixture solution. By contrast, Figure 11(d) shows no shift of ACN orientation in the mixture solution. This indicates that, in the mixture solution, ACN has almost no influences on the orientation of water on the surface of the proto-fibril, but water has a remarkable influence on the orientation of ACN.

Fig. 11. Orientation distributions of solvent molecules on the proto-fibril surface and in the bulk. (a) Water molecules in pure water. (b) Water molecules in ACN/water mixture. (c) ACN molecules in pure ACN. (d) ACN molecules in ACN/water mixture.

The orientations of water and ACN in the mixture solution provide a clear microscopic physical picture for understanding the surface RDF results in Fig. 10(a). In the mixture solution, because ACN molecules accumulate on the proto-fibril surface with its nonpolar group orienting inward to the hydrophobic surface and with its polar group orienting outward, water molecules are attracted by the polar group of ACN, which allows water easier to approach the proto-fibril surface. In this case, the orientation of water molecules changes little in comparison with that in pure water, due to the repulsion between the polar group of ACN and the oxygen atom of water. However, the aggregation of water around the polar group of ACN might change the microscopic environment of ACN and thus lead to the orientation change of ACN on the proto-fibril surface.

3.6. Microscopic mechanisms

Based on all our simulation results, we provide a clear microscopic physical picture to explain the ACN effects on peptide self-assembly. First, we have to keep in mind that intra- and inter-sheet interactions play different roles in peptide self-assembly. The polar intra-sheet interaction comes from the H-bonding interaction between peptide backbones and determines the growth along the H-bond direction, as shown in Fig. 12(b). Peptide backbones interact stronger with water molecules than with the polar groups of ACN, resulting in a stronger interaction between backbone and backbone in pure ACN than in pure water. In simulations of both the cross-β sheet and the proto-fibril, the H-bonds between peptides in the solutions with ACN are more than that in pure water (Fig. 4(a) and Fig. 7), so the growth along the H-bond direction can be strengthened by adding ACN into aqueous solution.

Fig. 12. Schematic diagram for the opposite effects of ACN on peptide self-assembly. (a) Symbols for peptide, water, and ACN molecules (polar groups in red and nonpolar groups in green) as well as polar (red) and nonpolar (green) interactions. (b) Since water has a stronger hydrophilic interaction with peptide than ACN, hydrophilic interactions between peptide molecules in solutions with ACN are stronger than in pure water. (c) Since ACN has a stronger nonpolar interaction with peptide than water, cross-β sheets in solutions with ACN effectively have a less hydrophobic surface than in pure water, so the lateral stacking in former is weaker than in latter.

The inter-sheet attraction comes from the interactions of hydrophobic side chains between neighboring sheets, which leads to the lateral stacking along the side chain direction, as shown in Fig. 12(c). Hydrophobic peptide side chains have a weaker interaction with water than with the nonpolar group of ACN, so water molecules are more difficult to accumulate than ACN molecules on the hydrophobic surface of the cross-β sheet. Due to its strong interaction with hydrophobic side chains, the nonpolar group of ACN is apt to stay closer to the hydrophobic surface of cross-β sheets. The ordered aggregation of ACN around the cross-β surface weakens the hydrophobicity of the cross-β sheets, resulting in a weaker attraction among cross-β sheets. Consequently, the width growth of lateral stacking along the side-chain direction can be weakened by adding ACN in aqueous solution, which explains our experimental observation that the width of the self-assembled nanostructure can be monotonically tuned by adding ACN. Attributed to the opposite effects on intra- and inter-sheet interactions, with increasing ACN content in solutions, the aggregation is strengthened along the H-bonding direction but weakened along the side chain direction, so the mass of aggregation does not monotonically change with increasing ACN content, as observed in experiments.

4. Conclusions

Our previous experiments demonstrated that adding ACN in aqueous solution strongly affects the self-assembled morphology: with increasing ACN content, the width of the self-assemblies decreases and intermediate morphologies towards forming a nanotube, such as nanofibrils, twist ribbons, and helical ribbons, appear.[41] In this work, further experiments with more cases studied have shown that the self-assembled mass is also affected: the mass first decreases and then increases with increasing ACN content. A series of all-atom MD simulations reveal that the above phenomena originate from two opposite effects of ACN molecules on the peptide self-assembly process. Because an ACN molecule is composed of a polar atomic group and a nonpolar atomic group, the correlation between the polar group of ACN and backbone is weaker than that between water and backbone, but the correlation between the nonpolar group of ACN and hydrophobic peptide side chain is stronger than that between water and hydrophobic peptide side chain. The MD simulations for the proto-fibril show that ACN molecules have two opposite effects on the proto-fibril: ACN strengthens the intra-sheet interaction between peptides in the same cross-β sheet due to the weaker correlation between ACN and backbone, but weakens the inter-sheet interaction between two cross-β sheets due to the stronger correlation between ACN and hydrophobic side chain, which in turn strengthens the wetting degree of cross-β sheets. The wetting attenuates the effective inter-sheet attractions, so the width of self-assembled structure decreases with increasing ACN content. On the other hand, ACN also strengthens the intra-sheet interaction, which enhances peptide aggregation along the H-bond direction. Because of the two opposite effects, the aggregation propensity of peptides will be weakest at a critical ACN content. The microscopic mechanism we have revealed for the solvent effect on proto-fibril is anticipated to promote peptide self-assembly experiments which program the self-assembled structure through tuning the contents of small amphiphilic organic molecules in aqueous solution.

Reference
1Ulijn R VSmith A M 2008 Chem. Soc. Rev. 37 664
2Chiti FDobson C M 2009 Nat. Chem. Biol. 5 15
3Caughey BLansbury P T 2003 Ann. Rev. Neurosci. 26 267
4Dobson C M 2003 Nature 426 884
5Nyrkova ISemenov A NAggeli ABell MBoden NMcLeish T C 2000 Eur. Phys. 17 499
6Nelson RSawaya M RBalbirnie MMadsen A ORiekel CGrothe REisenberg D 2005 Nature 435 773
7Sawaya M RSambashivan SNelson RIvanova M ISievers S AApostol M IThompson M JBalbirnie MWiltzius J J WMcFarlane H TMadsen A ORiekel CEisenberg D 2007 Nature 447 453
8Ghadiri M RGranja J RMilligan R AMcRee D EKhazanovich N 1993 Nature 366 324
9Deechongkit SPowers E TYou S LKelly J W 2005 J. Am. Chem. Soc. 127 8562
10Scanlon SAggeli A 2008 Nano Today 3 22
11Adamcik JCastelletto VBolisetty SHamley I WMezzenga R 2011 Angew. Chem. Int. Ed. 50 5495
12Baldwin R LRose G D 1999 Trends Biochem. Sci. 24 26
13Baldwin R LRose G D 1999 Trends Biochem. Sci. 24 77
14Nyrkova ISemenov A NAggeli ABoden N 2000 Eur. Phys. 17 481
15Aggeli ANyrkova I ABell MHarding RCarrick LMcLeish T C BSemenov A NBoden N 2001 Proc. Natl. Acad. Sci. USA 98 11857
16Lu KJacob JThiyagarajan PConticello V PLynn D G 2003 J. Am. Chem. Soc. 125 6391
17Lamm M SRajagopal KSchneider J PPochan D J 2005 J. Am. Chem. Soc. 127 16692
18Pashuck E TStupp S I 2010 J. Am. Chem. Soc. 132 8819
19Ziserman LLee H YRaghavan S RMor ADanino D 2011 J. Am. Chem. Soc. 133 2511
20Childers W SAnthony N RMehta A KBerland K MLynn D G 2012 Langmuir 28 6386
21Whitesides G MGrzybowski B 2002 Science 295 2418
22Knowles T PFitzpatrick A WMeehan SMott H RVendruscolo MDobson C MWelland M E 2007 Science 318 1900
23Bowerman C JRyan D MNissan D ANilsson B L 2009 Mol. Biosyst. 5 1058
24Xu HWang JHan S YWang J QYu D YZhang H YXia D HZhao X BWaigh T ALu J R 2009 Langmuir 25 4115
25Lee N RBowerman C JNilsson B L 2013 Biomacromolecules 14 3267
26Mehta A KLu KChilders W SLiang YDublin S NDong J JSnyder J PPingali S VThiyagarajan PLynn D G 2008 J. Am. Chem. Soc. 130 9829
27Adamcik JMezzenga R 2011 Soft Matter 7 5437
28Jordens SAdamcik JAmar-Yuli IMezzenga R 2011 Biomacromolecules 12 187
29Castelletto VHamley I WHarris P J FOlsson USpencer N 2009 J. Phys. Chem. 113 9978
30Hua LZhou R HThirumalai DBerne B J 2008 Proc. Natl. Acad. Sci. USA 105 16928
31Hwang SShao QWilliams HHilty CGao Y Q 2011 J. Phys. Chem. 115 6653
32Shao QFan Y BYang L JGao Y Q 2012 J. Chem. Phys. 136 115101
33Candotti MEsteban-Martin SSalvatella XOrozco M 2013 Proc. Natl. Acad. Sci. USA 110 5933
34Xu W XPing JLi W FMu Y G 2009 J. Chem. Phys. 130
35Emamyari SFazli H 2014 Eur. Biophys. J. Biophy 43 143
36Wei G HShea J E 2006 Biophys. 91 1638
37Yang CLi J YLi YZhu X L 2009 J. Mol. Structure-theochem 895 1
38Li W FQin MTie Z XWang W 2011 Phys. Rev. 84 041933
39Klimov D KStraub J EThirumalai D 2004 Proc. Natl. Acad. Sci. U.S.A. 101 14760
40Rissanou A NGeorgilis EKasotaids EMitraki AHarmandaris V 2013 J. Phys. Chem. 117 3962
41Zhao Y RDeng LWang J QXu HLu J R 2015 Langmuir 31 12975
42Buchanan L EDunkelberger E BTran H QCheng P NChiu C CCao PRaleigh D Pde Pablo J JNowick J SZanni M T 2013 Proc. Natl. Acad. Sci. USA 110 19285
43Krone M GHua LSoto PZhou R HBerne B JShea J E 2008 J. Am. Chem. Soc. 130 11066
44Jose J CKhatua PBansal NSengupta NBandyopadhyay S 2014 J. Phys. Chem. 118 11591
45Fitzpatrick A W PDebelouchina G TBayro M JClare D KCaporini M ABajaj V SJaroniec C PWang L CLadizhansky VMuller S AMacPhee C EWaudby C AMott H RDe Simone AKnowles T P JSaibil H RVendruscolo MOrlova E VGriffin R GDobson C M 2013 Proc. Natl. Acad. Sci. USA 110 5468
46Van der Spoel DLindahl EHess BGroenhof GMark A EBerendsen H J C 2005 J. Comput. Chem. 26 1701
47Jorgensen W LMaxwell D STitado-Rives J 1996 J. Am. Chem. Soc. 118 11225
48Jorgensen W LTirado-Rives J 2005 Proc. Natl. Acad. Sci. USA 102 6665
49Caleman Cvan Maaren P JHong M YHub J SCosta L Tvan der Spoel D 2012 J. Chem. Theory Comput. 8 61
50Jorgensen W LWilliam LChandrasekhar JMadura J DImpey R WKlein M L 1983 J. Chem. Phys. 79 926
51Deng LZhou PZhao Y RWang Y TXu H 2014 J. Phys. Chem. 118 12501
52Darden TYork DPedersen L 1993 J. Chem. Phys. 98 10089
53Essmann UPerera LBerkowitz M LDarden TLee HPedersen L G 1995 J. Chem. Phys. 103 8577
54Ravikumar K MHwang W 2011 J. Am. Chem. Soc. 133 11766
55Patel A JVarilly PJamadagni S NAcharya HGarde SChandler D 2011 Proc. Natl. Acad. Sci. USA 108 17678
56Patel A JVarilly PJamadagni S NHagan M FChandler DGarde S 2012 J. Phys. Chem. 116 2498
57Shao Q 2014 J. Phys. Chem. 118 6175