The self-assembly mechanism of tetra-peptides from the motif of β-amyloid peptides: a combined coarse-grained and all-atom molecular dynamics simulation

Abstract

Understanding the self-assembly mechanisms of peptides into nanostructures is essential for the rational design of bio-nanomaterials. Moreover, the natural fiber formation of Alzheimer’s β-amyloid peptides is crucially involved in Alzheimer’s disease but the mechanism still remains obscure. Herein, the assembly of the tetra-peptide motif VFFA from Aβ peptides and its derivations KFFA and FFFA into different nanostructures was investigated with combined coarse-grained (CG) and all-atom (AA) models. The primary structures of the tetra-peptides were found to be the most important factor to form special nanostructures rather than the concentration of the tetra-peptides. FFFA tends to form nanosheets, while VFFA tends to form nanospheres and KFFA tends to form nanorods from the CG simulation. The stabilities of the aggregated structures from the CG simulation were investigated and confirmed by AA simulations. In addition, FFFA and VFFA have a greater tendency to assemble into ordered nanostructures than KFFA, and VFFA prefers to form a large beta-sheet like structure from cluster analysis.

---

1. Introduction

Understanding of the formation mechanisms of high-order nanostructures is essential for their potential applications in nanotechnology.^1–3 Owing to the biocompatibility and inherent molecular recognition properties of biomolecules, nanostructures formed by peptide self-assembly have attracted a lot of attention in recent years.^4–7 Peptides from two to seven amino acid residues have been confirmed to form ordered nanostructures from many experiments,^8–11 and the ordered nanostructures could be nanotubes, nanovesicles and bilayers etc.^11–13 Thus, the understanding of the nature of self-assembly of nanostructures from short peptides is significant for the rational design of controlled nano-architectures.

The diphenylalanine (FF) motif of Alzheimer’s β-amyloid peptides was thought to play a key role in the amyloid fibrillation process, and it has been investigated as a diagnostic target for Alzheimer’s disease.¹⁴ FF is the smallest self-assembled peptide sequence in the Alzheimer’s β-amyloid motif¹⁰ and its derivatives can assemble into highly ordered aromatic dipeptide nanostructures including nanotubes and nanowires, as revealed by many experiments.^15–17 From X-ray diffraction studies, Görbitz pointed out that the phenyl rings in the FF peptides promote nanostructure formation.¹⁸ Kumaraswamy et al. found that the pH and substrates could affect the assembly process of peptides.¹⁹ Self-assembly into peptide nanotubes from AAKLVFF was studied in methanol, and the important role of aromatic interactions between phenylalanine residues in driving β-sheet self-assembly was confirmed by circular dichroism spectroscopy.²⁰ These findings revealed many details about the nanostructure’s formation from short aromatic peptides, however, the process and the mechanism of the nanostructure’s formation are still unclear at the molecular level. Besides the experiments, molecular simulations including coarse-grained (CG) and all-atom (AA) models have been used successfully to investigate the mechanism of assembly of short peptides.^21–23 By using both coarse-grained²⁴ and all-atom models,²⁵ Schatz et al. pointed out that peptide amphiphiles could self-assemble into nanofibers. Zhou et al. found that the self-assembly mechanism of nanostructures from FF and triphenylalanine (FFF) is different by using the coarse-grained model.²⁶ Tuttle et al. successfully predicted the aggregation propensity of different dipeptides by using the MARTINI V2.1 CG model, and confirmed it experimentally.²⁷ Using CG methods, Frederix et al. developed a tool to enable the peptide sequence space to be searched for supramolecular properties, and predicted the self-assembly nanostructures of all 8000 possible tripeptides.²³ All these simulations have greatly improved the understanding of the process and dynamics of self-assembly of short peptides. Although significant advances have been made in the study of the aggregation process of short peptides, the assembly pathways of the ordered nanostructures still remain obscure. Particularly, the effects of the concentration of the peptides and the residue type of the peptide motif on the aggregation mechanism of the amyloid have not been well studied, and aggregation mechanisms from the motif of amyloid beta-peptides are still unclear.

Based on this background, the CG model with the MARITNI V2.1 force field was used to investigate the self-assembly process of motifs of beta-peptides on a microsecond scale in this study. VFFA and its analogues KFFA and FFFA were selected to study the effect of residue type on the aggregation mechanism. To investigate the stability of the aggregated structures from the CG simulation, all-atom model based simulations were also carried out in our study.

By performing a series CG molecular dynamics (MD) simulations, we found that FFFA could assemble into nanosheets and VFFA could assemble into nanospheres in both high and low concentrations, while KFFA could not assemble into an ordered nanostructure. All the results of the CG MD simulations were confirmed by AA MD simulations. The mechanism of the self-assembly of peptides into nanostructures was investigated in this study.

2. Methods

Coarse-grained simulations

Coarse-grained simulations have been widely used in the study of biomolecules.^28–31 The all-atom tetra-peptide models with different sequences (VFFA, KFFA, and FFFA) were constructed using Hyperchem software.³² During the CG simulation, one atomistic tetra-peptide molecule was first immersed into TIP3P³³ water molecules. Then the tetra-peptide with atomistic details was energy minimized. After that, the tetra-peptide was mapped into a CG structure, and then copied into a 5 × 5 × 12 array in x, y and z directions, respectively. Therefore, the whole system contains 300 CG tetra-peptides. Then a box of 20.0 nm × 20.0 nm × 12.5 nm was constructed into which these CG tetra-peptides were placed and 32 000 water beads were added, making the concentration of tetra-peptides around 40 mg mL⁻¹. The structure of all the tetra-peptides after this process was used as the initial structure for the simulation. When simulating the assembly process of the tetra-peptides, a high peptide concentration, as used in the simulations of other groups, was chosen in this study to accelerate the assembly process.^34–37 The parameters of the water beads and tetra-peptides in our CG model are taken from the MARTINI V2.1 force field.³⁸ In this force field, four main types of interaction sites were constructed and used to map the main chain and side chain of the residues: polar (P), nonpolar (N), apolar (C) and charged (Q). One water bead represents four atomistic water molecules. As shown in Fig. 1, one bead represents the main chain of one residue in the tetra-peptide. Val is composed of one bead for the main chain and the other bead for the side chain. Ala is only composed of one bead, and the benzene ring of phenylalanine is represented by three beads. Compared with the all-atom model, the MARTINI force field allows a 4-fold reduction in the number of simulated particles. After a 100 ns equilibration, 5 μs simulations were performed with a time step of 20 fs in all systems. To investigate the effect of concentration on the self-assembly of tetra-peptides into nanostructures, systems with different concentrations (10 mg mL⁻¹ and 40 mg mL⁻¹) of tetra-peptides (VFFA, KFFA, and FFFA) were constructed and simulated. All simulations were carried out in the NPT ensemble using the GROMACS 4.6.5 package.³⁹ Temperature and pressure baths are weakly coupled by Berendsen coupling methods,⁴⁰ and a constant temperature of 310 K and a constant pressure of 1 atm were used in the simulation. The cutoff of electrostatic and van der Waals (vdW) interactions is 1.2 nm, and the vdW interaction is shifted from 0.9 nm to 1.2 nm. Periodic boundary conditions (PBC) are imposed in all directions. Visual molecular dynamics (VMD) was used to visualize the results and generate the snapshots.⁴¹

Fig. 1 Mapping and back-mapping between the all-atom models and coarse-grained models (FFFA, KFFA, VFFA). The all-atom structures were displayed using Licorice drawing style and coarse-grained structures were displayed by transparent VDW drawing style in VMD.⁴¹

All-atom molecular dynamics in an explicit solvent

After a 5 μs CG simulation, tetra-peptides and water beads in systems with concentrations of 40 mg mL⁻¹ were back-mapped into an AA model with a CHARMM 36 force field using a back-mapping tool.⁴² All three systems included around 420 000 atoms with 300 tetra-peptide molecules and 128 000 water molecules (one bead for water molecules represents four atomistic water molecules). All atoms including hydrogen atoms were represented explicitly. The cutoff for non-bonded van der Waals interactions was set by a switching function starting at 1.0 nm and reaching zero at 1.2 nm. A time step of 2 fs was used in all simulations. The temperature was maintained at 310 K using a Berendsen thermostat with a constant of 0.1 ps, and the pressure was maintained at 1 atm using a Berendsen barostat. Particle mesh Ewald (PME) summation was used to calculate the long range electrostatic interactions, with a cutoff of 1.2 nm for the separation of the direct and reciprocal space summation. The cutoff distance for the vdW interaction was 1.2 nm. All systems were performed for 100 ns in the GROMACS 4.6.5 package.³⁹

Analysis

The data was analyzed by a home scripts and GROMACS package. The number of tetra-peptides in the beta-sheet like structure was calculated based on two conditions: (1) the distance of two chains is less than 0.6 nm; (2) the angle between two chains is less than 30° or larger than 150°. Herein, any pair of two chains is considered as parallel or anti-parallel-aligned if the angle between them is less than 30° or larger than 150°. After that, the possibility of tetra-peptides assembling into beta-sheet like structures with a different size was calculated. The possibility of forming a beta-sheet like structure is defined as follows:

P = N_i/N_tot (1)

where N_i is the total number of tetra-peptides forming beta-sheets with i tetra-peptides, and N_tot is the total number of tetra-peptides in the system.

3. Results and discussion

Coarse-grained simulations

The initial structures of the tetra-peptide FFFA at a concentration of 40 mg mL⁻¹ are shown in Fig. 2. After a 5 μs CG simulation, a spontaneous aggregation process of all tetra-peptides was observed. All systems were repeated three times, and the results from the same system are similar and the ordered nanostructures are the same from the same system. Thus, we will focus on the simulation to study the aggregation process. As displayed in Fig. 2, different nanostructures varying from nanospheres and nanorods to nanosheets formed by the tetra-peptides were observed in the simulation. FFFA could self-assemble into nanorod-like structures within 0.2 μs, and turn into nanosheet-like structures within 1.0 μs. After that, the FFFA molecules that were out of the membrane fused into the membrane and the membrane became more flat at 2.0 μs. The bilayer-membrane formed by FFFA remained stable during the 5 μs simulation after 2.0 μs. The tetra-peptide KFFA tended to form a long nanorod at 0.5 μs, and then the length of the KFFA nanorod started to decrease. During the last 3 μs of the simulation, the KFFA peptides formed more ordered nanorods. This indicates that KFFA peptides could form stable and ordered nanorod structures. For the tetra-peptides of VFFA, they tended to aggregate into nanosphere-like structures within 5 μs. During the aggregation process of all three different tetra-peptides, nanorod-like structures were found at the initial time (0.5 μs for FFFA, 0.5 μs for KFFA and 0.2 μs for VFFA). However, the stable and ordered nano-structure at the end of the simulation depended on the primary structure of the tetra-peptide.

Fig. 2 Self-assembly process of different tetra-peptides (FFFA, KFFA, and VFFA) into different nanostructures. The left figure is the initial structure of 300 FFFA in the coarse-grained model from the top view. The beads of the backbone are shown in red and the side chain beads in white.

To understand the mechanism of nanostructure formation from the different types of tetra-peptides, the solvent accessible surface area (SASA) and radius of gyration (R_g) of all tetra-peptides during the first 2 μs of the simulation were measured. As shown in Fig. 3, the SASA greatly decreased in the three different systems. In the system of FFFA, the value of the SASA decreased from 1152 nm³ to 490 nm³. The value of the SASA in the KFFA system decreased from 935 nm³ to 680 nm³ within 0.4 μs, and kept stable for the rest of the 2 μs simulation. The value of the SASA in the VFFA system decreased from 1085 nm³ to 650 nm³ and kept stable after 1.2 μs. The SASA of the tetra-peptides decreased a lot during the aggregation process of the three different tetra-peptides. This implies the importance of the SASA in the aggregation process of the tetra-peptides. The SASA of FFFA peptides decreased more than half. This is most probably due to the fact that strong hydrophobic interactions between phenylalanines could facilitate the aggregation of FFFA peptides. It implies that the change in the SASA during the aggregation process also depended on the composition of the tetra-peptides. Moreover, the evolution of R_g in the three different systems is rather distinct, as shown in Fig. 3. Fig. 2 visually showed that the nanostructures formed by the different tetra-peptides are significantly different, and the change of R_g could reflect the structure change of these nanostructures. The value of R_g in the FFFA system decreased after 0.5 μs when the nanostructure changed from a nanorod-like structure into a nanosphere-like structure. However, this structure is less ordered than the bilayer membrane and is still not stable, and so the position of the peptides experienced a certain change and aggregated into the ordered bilayer membrane within 1.9 μs. The nano-membrane structure composed of 300 FFFA tetra-peptides was very stable after 1.9 μs. This is the reason that R_g increased even after the FFFA bilayer formation at 1.0 μs. After 1.9 μs, the less ordered bilayer membrane structure changed to an ordered bilayer membrane structure with the decrease of R_g. The nano-membrane structure composed of 300 FFFA tetra-peptides was very stable after 2.0 μs with a low R_g and SASA. The value of R_g in the KFFA system is always larger than 6.0 due to its long nanorod structures. The value of R_g in the VFFA system is very small since it formed a nanosphere structure during the simulation. Although the values of R_g for the KFFA and VFFA systems are almost the same after 1.0 μs, the x, y and z components of R_g in these two systems are different, as shown in Fig. S1 (see the ESI†). The x, y, and z components of R_g in the VFFA system are almost the same near 1.0 μs, and the x component of R_g is much smaller than the value of the y and z components of R_g in the KFFA system. This shows that the nanostructures formed by KFFA and VFFA are different.

Fig. 3 The evolution of the solvent accessible surface area (SASA) and radius of gyration (R_g) in the self-assembly process of three different types of tetra-peptide: (A) FFFA, (B) KFFA and (C) VFFA. The change in SASA and R_g are represented by black and blue lines, respectively.

To better explore the tetra-peptide self-assembly pathway, the projection of free energy in two reaction coordinates (R_g and SASA) was calculated. More specifically, the possibility distribution P(R_g, SASA) in two reaction coordinates R_g and SASA was calculated from a whole 5 μs simulation. The free energy (S) calculated from eqn (2) could reflect the possibility of non-equilibrium events as referenced in Zhou’s paper.⁴³

S = −RT log P(R_g, SASA) (2)

The nanostructures of self-assembled FFFA, KFFA and VFFA were nanosheets, nanorods and nanospheres, as shown in Fig. 2. In the aggregation process, the SASA decreases a lot in all three different tetra-peptides, as shown in Fig. 4. This implied the importance of SASA in the aggregation process of the tetra-peptides. However, the nanostructures formed by the tetra-peptides with different primary sequences were different. As shown in Fig. 4, the SASA could not be used to determine the nano-structures in all three different systems. The free energy landscape for self-assembly from the different tetra-peptides is different. The formation pathway of the FFFA nanostructures aligns with a decrease in R_g from nano-rod like structures to nano-membrane like structures, as shown by the red dotted line in Fig. 4A. The formation pathway of the KFFA nanostructures aligns with an increase in R_g and the formation pathway for the VFFA nanostructures aligns with a decrease in R_g, which has almost no barrier in the formation pathway. These results indicate that it is very easy for KFFA and VFFA to self-assemble into nanostructures but not for FFFA. The formed nanostructures of FFFA and VFFA peptides are different although both the SASA and R_g decreased in the two different systems. It implies that the ordered nanostructures formed by the different tetra-peptides are not determined by the SASA and R_g but are dependent on the different primary sequences of the tetra-peptides. The structure of FFFA has one more F than the other two types of tetra-peptide, and it could strengthen the pi–pi interaction between the different FFFA peptides and increase the parallel configuration of this system. Thus, the formed nano-structure from FFFA was a nano-membrane in the simulation, which is also found in the aggregation of FFF tripeptides.²⁶ The structure of VFFA is different from FFFA, and the side chain of the Val residue is a hydrophobic isopropyl. VFFA tended to form nanospheres with a hydrophobic interior. The nanostructures formed from VFFA and KFFA were different from that of the dipeptide FF.²⁶ It indicates that the assembly process of FF could be changed with the addition of other residues. To understand the nanostructures from different tetra-peptides better, the distribution of side chains of the tetra-peptides in the different nanostructures was analyzed. As shown in Fig. S3,† the first peak of the radial distribution functions (RDFs) of the side chain (SC) to peptide main chain (BB) in the FFFA system is a little higher than that of KFFA and VFFA. Due to the bilayer structure formed by FFFA, the SCs of the FFFA molecules were located in the middle of the bilayer structure. Therefore, the first peak of the RDFs in Fig. S3A† is a little larger than that of VFFA and KFFA. In Fig. S3B,† the first peak of the RDFs between the SCs of peptides in the VFFA system is a little higher than that of FFFA and KFFA. The structure of the nanosphere formed by VFFA made more SCs aggregate as a hydrophobic core. In addition, the nano-structure formed by KFFA is a nanorod, and the first peak of the non-ordered structure from KFFA in Fig. S3A and B† is much lower than that from VFFA and FFFA. This confirmed that the structure from KFFA is much more disordered than the other structures in the VFFA and FFFA systems.

Fig. 4 The free energy landscape of the self-assembly processes of three different tetra-peptides in two reaction coordinates (R_g and SASA): (A) FFFA; (B) KFFA; (C) VFFA.

The tetra-peptide VFFA is the motif of Aβ peptides for Alzheimer’s disease, thus the propensity of all tetra-peptides forming beta-sheet like structures was analyzed. This was done using a cluster method (details in the Methods part). By calculating the possible sizes of the beta-sheet like structures in the nanostructures formed by the different sequences of tetra-peptides, the propensity of the different tetra-peptides for forming beta-sheet like structures was compared. As shown in Fig. 5, a monomer with only one tetra-peptide has a probability of around 41.1%, 48.2% and 37.6% in the VFFA, KFFA and FFFA nanostructures, respectively. This implies that the monomer tetra-peptide is the largest part even using the core motif of beta amyloid proteins, and more monomer exists in the KFFA system compared with the other two sequences of tetra-peptides. The possibility of two-chain beta-sheets and three-chain beta-sheets from KFFA nanostructures is larger than that from VFFA and FFFA nanostructures. With the increase in size, the possibility of five- to twelve-chain beta-sheets from FFFA nanostructures is larger than VFFA, and both of them are much larger than that from KFFA nanostructures. To form beta-sheets with more than twelve chains, the propensity of VFFA is larger than FFFA, although the possibility of forming that large beta-sheet is very small (<3%) in VFFA nanostructures. This indicates that FFFA and VFFA have a larger propensity for forming beta-sheet like structures than KFFA, and VFFA has a larger propensity for forming large beta-sheets (more than twelve chains). From the results of Laio’s simulation, the free energy landscape for Aβ aggregation decreased after the amyloid fibril nucleation process.⁴⁴ Thus, the size of a beta-sheet like structure could be expected to increase as the simulation time extended. The propensity of VFFA for forming beta-sheet like structures proves the importance of the tetra-peptide in Aβ aggregation, and it may be a target for the inhibition of the amyloid fibrillation process.

Fig. 5 The distribution of size of beta-sheet like structures formed from tetra-peptides with different sequences. (A) Containing less than 11 tetra-peptides; (B) containing more than 10 tetra-peptides.

To understand the effect of the concentration of tetra-peptides on the formed nanostructure, simulations of FFFA at a low concentration (10 mg mL⁻¹) were carried out. As shown in Fig. 6, both of the nanostructures of FFFA at different concentrations (10 mg mL⁻¹ and 40 mg mL⁻¹) are bilayer membranes. The main chain with four beads is at the bilayer membrane surface and it is accessible by water molecules, as shown in red. The hydrophobic benzene rings are in the inner part of the membrane, as shown in white. The nanostructure aggregated from the low FFFA concentration (10 mg mL⁻¹) is quite similar to that from the high FFFA concentration (40 mg mL⁻¹). The formed nanostructure is not affected by the FFFA concentration, which implies that the concentration is not as important as the sequence of tetra-peptides to the self-assembled nanostructure. A similar phenomenon was also observed in low concentration systems of KFFA and VFFA. As shown in Fig. S2 (see ESI†), in the system of KFFA (10 mg mL⁻¹) at low concentrations, the self-assembled nanostructure is still a nanorod, and it is still a nanosphere for VFFA at low concentrations (10 mg mL⁻¹). All the simulation results from the systems of FFFA, KFFA and VFFA tend to form ordered nanostructures at different concentrations.

Fig. 6 The last configuration of FFFA at the concentrations of (A) 10 mg mL⁻¹ and (B) 40 mg mL⁻¹. This implies that the concentration has little influence on the aggregated configuration.

All-atom simulations

To understand the details of the nanostructures better, AA simulations for the FFFA and KFFA nanostructures were carried out and the structures of these AA structures were back-mapped from the last frame of the 5 μs CG simulation. Fig. 7A displays the FFFA bilayer nanostructure after a 100 ns AA simulation. After equilibration of the CG structures and the back-mapping procedure, the AA structure is relatively easy to equilibrate and 50 ns was used to facilitate the equilibration of the AA structures. In a CG simulation, the calculation of the structure and thermodynamics of a benzene or residues containing phenyl rings is one of the most challenging to perform, mainly due to their aromaticity and planar structures.⁴⁵ Therefore, in this study, we chose the angle distributions between the neighboring aromatic planes in each FFFA tetra-peptide molecule as the criterion of our CG model. The angle distributions between the neighboring aromatic planes in each FFFA tetra-peptide molecule were calculated from the last 50 ns of the AA simulation and the last 1 μs from the CG simulation. As displayed in Fig. 7C, the angle distribution is almost the same from both the CG and AA simulations. It proves that the results from the CG simulation are reasonable and trustable. From the data in Fig. 7C, the angle between the neighboring benzene rings tends to be 90°, and the two planes tend to form a T-shape to keep the structure stable.

Fig. 7 The angle distribution between neighboring benzene rings in FFFA nanostructures. (A) is the nano-structure represented by an all-atom FFFA model. N atoms are represented by red spheres, and other atoms are represented by blue lines. (B) The definition of angles between neighboring benzene rings. (C) The angle distribution between neighboring benzene rings in the coarse-grained (CG) simulation and the all-atom (AA) simulation.

The dihedral angle distribution of four N atoms in one tetra-peptide was calculated, as is shown in Fig. 8. Most of the dihedral angles between four N atoms are smaller than 30° in the FFFA nanostructures. This is due to the fact that the nanostructure of FFFA is a bilayer membrane. The main chains are located on the outside of the membrane and are accessible by water molecules, while the hydrophobic benzene rings are in the inner part of the membrane. To keep the membrane flat and stable, the dihedral angles between four N atoms in the main chain are generally very small, and most of them are less than 30°. However, the KFFA nanostructures are nanorod like, and some KFFA units need to touch each other to form a circular structure from the top view. Thus, most of the dihedral angles between four N atoms are in the range of 0° to 60°.

Fig. 8 The dihedral distribution of the main four beads in the backbone in FFFA and KFFA nanostructures.

4. Conclusions

In summary, combined coarse-grained and all-atom MD simulations were used to investigate the self-assembly process of the beta-amyloid tetra-peptide core VFFA and its variants FFFA and KFFA. The nanostructures formed by FFFA, KFFA and VFFA are bilayer membranes, nanorods and nanospheres, respectively. The formation of different ordered nanostructures was mainly affected by the primary sequence of the tetra-peptides and not by the concentration. The results were confirmed by all-atom simulations. From the cluster analysis, FFFA and VFFA have a larger propensity for forming beta-sheet like structures than KFFA, and VFFA has a large propensity for forming large beta-sheets (more than twelve chains). These findings provide a deep insight into the self-assembly of tetra-peptides and the properties of tetra-peptide nanostructures. This work may promote future research into high-order peptide nanostructure design and its potential applications in bio-nanotechnology.

Acknowledgements

We acknowledge financial support from the National Natural Science Foundation of China (Grant No. 21674032, 21503186, 21403049), and Zhejiang Provincial Natural Science Foundation of China (Grant No. LY14B030008).

References

1. M.-C. Daniel and D. Astruc, Chem. Rev., 2004, 104, 293–346.
2. M. R. Buck, J. F. Bondi and R. E. Schaak, Nat. Chem., 2012, 4, 37–44.
3. F. Meng, S. Zheng, H. Li, Q. Liang and T. Liu, Macromolecules, 2006, 39, 5072–5080.
4. E. Gazit, Chem. Soc. Rev., 2007, 36, 1263–1269.
5. S. Zhang, D. M. Marini, W. Hwang and S. Santoso, Curr. Opin. Chem. Biol., 2002, 6, 865–871.
6. K.-B. Lee, J.-H. Lim and C. A. Mirkin, J. Am. Chem. Soc., 2003, 125, 5588–5589.
7. L. Adler-Abramovich and E. Gazit, Chem. Soc. Rev., 2014, 43, 6881–6893.
8. G. Fichman and E. Gazit, Acta Biomater., 2014, 10, 1671–1682.
9. S. Zhang, Nat. Biotechnol., 2003, 21, 1171–1178.
10. M. Reches and E. Gazit, Nano Lett., 2004, 4, 581–585.
11. C. Guo, Y. Luo, R. Zhou and G. Wei, ACS Nano, 2012, 6, 3907–3918.
12. X. Zhao, F. Pan, H. Xu, M. Yaseen, H. Shan, C. A. E. Hauser, S. Zhang and J. R. Lu, Chem. Soc. Rev., 2010, 39, 3480–3498.
13. S. Vauthey, S. Santoso, H. Gong, N. Watson and S. Zhang, Proc. Natl. Acad. Sci. U. S. A., 2002, 99, 5355–5360.
14. C. H. Görbitz, Chem. Commun., 2006, 2332–2334.
15. M. Reches and E. Gazit, Science, 2003, 300, 625–627.
16. O. Carny, D. E. Shalev and E. Gazit, Nano Lett., 2006, 6, 1594–1597.
17. X. Yan, P. Zhu and J. Li, Chem. Soc. Rev., 2010, 39, 1877–1890.
18. C. H. Görbitz, Chem.–Eur. J., 2001, 7, 5153–5159.
19. P. Kumaraswamy, R. Lakshmanan, S. Sethuraman and U. M. Krishnan, Soft Matter, 2011, 7, 2744–2754.
20. M. J. Krysmann, V. Castelletto, J. E. McKendrick, L. A. Clifton, P. J. F. Harris and S. M. King, Langmuir, 2008, 24, 8158–8162.
21. C.-A. Palma, M. Cecchini and P. Samorì, Chem. Soc. Rev., 2012, 41, 3713–3730.
22. J. B. Matson and S. I. Stupp, Chem. Commun., 2012, 48, 26–33.
23. P. W. J. M. Frederix, G. G. Scott, Y. M. Abul-Haija, D. Kalafatovic, C. G. Pappas, N. Javid, N. T. Hunt, R. V. Ulijn and T. Tuttle, Nat. Chem., 2014, 7, 30–37.
24. O.-S. Lee, V. Cho and G. C. Schatz, Nano Lett., 2012, 12, 4907–4913.
25. O.-S. Lee, S. I. Stupp and G. C. Schatz, J. Am. Chem. Soc., 2011, 133, 3677–3683.
26. C. Guo, Y. Luo, R. Zhou and G. Wei, Nanoscale, 2014, 6, 2800–2811.
27. P. W. J. M. Frederix, R. V. Ulijn, N. T. Hunt and T. Tuttle, J. Phys. Chem. Lett., 2011, 2, 2380–2384.
28. S. O. Nielsen, C. F. Lopez, G. Srinivas and M. L. Klein, J. Phys.: Condens. Matter, 2004, 16, R481.
29. C. F. Lopez, P. B. Moore, J. C. Shelley, M. Y. Shelley and M. L. Klein, Comput. Phys. Commun., 2002, 147, 1–6.
30. H. I. Ingólfsson, C. A. Lopez, J. J. Uusitalo, D. H. de Jong, S. M. Gopal, X. Periole and S. J. Marrink, Wiley Interdiscip. Rev.: Comput. Mol. Sci., 2014, 4, 225–248.
31. L. A. Patel and J. T. Kindt, Soft Matter, 2016, 12, 6.
32. I. Gutowska, Z. Machoy and B. Machaliński, J. Biomed. Mater. Res., Part A, 2005, 75, 788–793.
33. W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey and M. L. Klein, J. Chem. Phys., 1983, 79, 926–935.
34. P. J. Bond, J. Holyoake and A. Ivetac, et al., J. Struct. Biol., 2007, 157, 593–605.
35. G. Srinivas, D. E. Discher and M. L. Klein, Nat. Mater., 2004, 3, 638–644.
36. Y. Sun, Z. Qian, C. Guo and G. Wei, Biomacromolecules, 2015, 16, 2940–2949.
37. N. Thota, Z. Luo and Z. Hu, et al., J. Phys. Chem. B, 2013, 117, 9690–9698.
38. L. Monticelli, S. K. Kandasamy, X. Periole, R. G. Larson, D. P. Tieleman and S.-J. Marrink, J. Chem. Theo. Comput., 2008, 4, 819–834.
39. B. Hess, C. Kutzner, D. Van Der Spoel and E. Lindahl, J. Chem. Theory Comput., 2008, 4, 435–447.
40. H. J. Berendsen, J. V. Postma, W. F. van Gunsteren, A. R. H. J. DiNola and J. R. Haak, J. Chem. Phys., 1984, 1984(81), 3684–3690.
41. W. Humphrey, A. Dalke and K. V. M. D. Schulten, Visual molecular dynamics, J. Mol. Graphics Modell., 1996, 14, 33–38.
42. T. A. Wassenaar, K. Pluhackova, R. A. Böckmann, S. J. Marrink and D. P. Tieleman, J. Chem. Theory Comput., 2014, 10, 676–690.
43. R. Zhou, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 13280–13285.
44. F. Baftizadeh, X. Biarnes, F. Pietrucci, F. Affinito and A. Laio, J. Am. Chem. Soc., 2012, 134, 3886–3894.
45. A. Villa, C. Peter and A. N. F. van der Vegt, J. Chem. Theory Comput., 2010, 6, 2434–2444.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra18204f

---