Deciphering the Rules for Amino Acid Co-Assembly Based on Interlayer Distances

Abstract

Metabolite materials are extremely useful to obtain functional bioinspired assemblies with unique physical properties for various applications in the fields of material science, engineering, and medicine by selfassembly of the simplest biological building blocks. Supramolecular co-assembly has recently emerged as a promising extended approach to further expand the conformational space of metabolite assemblies in terms of structural and functional complexity. Yet, the design of synergistically co-assembled amino acids to produce tailor-made functional architectures is still challenging. Herein, we propose a design rule to predict the supramolecular co-assembly of naturally occurring amino acids based on their interlayer separation distances observed in single crystals. Using diverse experimental techniques, we demonstrate that amino acids with comparable interlayer separation strongly interact and co-assemble to produce structural composites distinctly different from their individual properties. However, such an interaction is hampered in a mixture of differentially layer-separated amino acids, which self-sort to generate individual characteristic structures. This study provides a different paradigm for the modular design of supramolecular assemblies based on amino acids with predictable properties.

Metabolites are essential components of life and perform multitudes of biological functionalities ranging from cellular fuels and signaling to cofactors and structural elements.^1,2 Functional metabolites assemblies, which originate from the higher-order supramolecular packing of simplest metabolite building blocks, are also abundant in nature.³ An important example is the photonic structures observed in biological systems which are comprised of a regular array of layered metabolite crystals.⁴ The shiny mirrorreflecting eyes of numerous animals result from the presence of assembled metabolites in the tapetum lucidum tissue which reflect light as well as divert it into a more useful range by fluorescence.^5–7 A cysteine-zinc complex was found to carry out this function in the ferret and canine tapetum lucidum.⁷ In addition, several amino acid-based metabolites, including Phe, Cys, Trp, and Tyr, were shown to form amyloid-like ordered supramolecular assemblies both in vitro and in vivo.^8–12

Metabolites self-assembly into distinct structural entities has emerged as a future avenue of research for the fabrication of diverse bioinspired materials for different applications, ranging from optics to energy harvesting devices. Similar to protein amyloids, self-assembly of amyloid fibers forming amino acids has been explored to design interesting biomaterials.^13–15 In deionized water, 4.0-5.5 mM Tyr self-assembled into fibers, which were altered to nanoribbon strip-like nanostructures upon decreasing the concentration to 2.8 mM.¹³ When dissolved in ethanol, Trp was found to form discrete nanotubes that exhibited fluorescence when excited at 385 (blue), 488 (yellow), and 561 (red) nm.¹⁴ In another report, self-assembly of the aromatic amino acids, Phe, Tyr, Trp, and His, was studied in water/methanol mixtures. In a 1:1 water:methanol ratio, Phe formed ordered fibers, whereas Tyr gave rise to rodlike microstructures. Trp was found to assemble into porous networks of twisted nanosheets, and His formed dendritic structures.¹⁵

Apart from morphological diversity, a recent study suggested a high piezoelectric voltage constant for β-glycine crystals, which was reported to be an order of magnitude larger than the voltage generated by any currently used ceramic or polymer.¹⁶ The ordered packing of β-glycine molecules along certain crystallographic planes and directions was found to be responsible for this high piezoelectric coefficient. Despite of progresses in the development of various functional materials, the use of unimolecular amino acid assemblies has been limited by a lack of chemical diversity and functional complexity.

In an interesting strategy, it has been recognized that the supramolecular co-assembly of two different short peptide building blocks into one ordered structural material can improve the stability,^17,18 biofunctionality,^19–21 and mechanical properties^22–25 of the composites, compared to the individual peptide assemblies. For amino acids, the pure centrosymmetric crystals of α-glycine were surprisingly reported to exhibit surface pyroelectricity by stereospecific doping with guest molecules.²⁶ Also the growth of centrosymmetric glycine crystals in the presence of specific additives was suggested to result in the generation and amplification of optical activity.²⁷ The hydrophobic faces of a series of racemic α-amino acid crystals were employed as a substrate onto which water vapor was cooled to freezing, thereby inducing the nucleation of ice.²⁸

Thus, co-assembly could be an efficient strategy to synergistically modulate the characteristics of supramolecular building blocks. However, tailor-made design of functional supramolecular ensembles is hampered by the lack of a structural correlation among different amino acids, thereby prohibiting a prediction of their resultant co-assemblies. Here, in an effort to establish a basic rule to predict the structure and supramolecular co-assembly relationship of the naturally occurring amino acids, we present the predictable co-assembly of amino acids with similar chirality depending on their interlayer separation distances observed in X-ray single crystal structures. This study provides different directions for the design of modular supramolecular assemblies of amino acid mixtures with desired physical, chemical, and mechanical properties.

In a recently published overview,²⁹ we introduced the concept of a supramolecular β-sheet structure for proteineous amino acids (L-amino acids) devoid of any covalent bond, in contrast to peptides/proteins. We have analyzed and demonstrated that in spite of the different side-chains, which significantly vary in their chemical properties, the crystal packing of all coded amino acids displays a layer-by-layer assembly stabilized by α-amine to α-carboxyl hydrogen bonds, resembling supramolecular β-sheet structures (Figure S1). Depending on their interlayer separation distances (Figure S2), amino acids can be classified into three different subgroups (Table S1) that correlate well with previously reported classifications of their β-sheet propensity based on various criteria, such as hydrophobicity, steric bulkiness, and folding. We postulate that due to their comparable interlayer separation, two amino acids from the same subgroup would interact with each other in mixture and co-assemble to produce a different structural architecture, while the members of different subgroups would not tend to co-assemble and retain their individual properties (Scheme 1). To experimentally test this hypothesis, we examined the admixing of amino acids from different subgroups using electron microscopy, mass spectrometry, X-ray powder diffraction, physical vapor deposition, isothermal titration calorimetry, and molecular dynamics (MD) simulation techniques. We chose Phe as a reference amino acid due to its well-documented supramolecular assembly and related biophysical and biochemical properties.^8–12 The co-assembly of Phe with either isoleucine (Ile) or methionine (Met), both from the same subgroup, as well as with either glycine (Gly) or alanine (Ala), representing a different subgroup, was thoroughly analyzed. According to our hypothesis, Phe would interact strongly with the Ile and Met in the supramolecular ensembles, whereas Gly and Ala would show minimal affinity for Phe.

Scheme 1. Schematic Structural Illustration of Amino Acid Assemblies in Pure and Mixed Conditions.

Results And Discussion

In order to investigate the underlying morphologies of selfassembled pure amino acids and mixed systems, we employed high-resolution scanning electron microscopic (HRSEM) image analysis. Figure 1a–e presents the architectures of the Phe, Ile, Met, Gly, and Ala individual amino acids in deionized water, respectively. Phe formed elongated micron-long fibers, similar to the previously reported amyloid-like fibrillar structures.^8,11,12 Ile and Met showed large hexagonal crystalline assemblies. HRSEM image of Gly indicated the presence of a one-dimensional (1D) long crystal and Ala produced flake-like ensembles. Formation of two-dimensional (2D) and threedimensional (3D) crystalline arrays and flake-like structures is very common for amino acid and peptide assembly.^30,31 The self-assembled morphology depends on several factors, such as concentration, solvent, temperature, presence of guest molecules, and many others.^13–15,30 It has been also observed that several peptides undergo self-assembly to form small structural ensembles in solution. When deposited over a 2D surface, the small structures underwent diffusion-limited aggregation (DLA) to fabricate self-similar fractals structures.^32–34 However, the self-assembled structures produced by the studied amino acids are quite large. They have a defined shape and are distinctly different from fractal structural organizations observed as an effect of drying. Under similar experimental conditions, co-assembly studies showed interesting results. The equimolar mixture of Phe and Ile revealed the formation of a spherical architecture 5-7 μm in diameter (Figure 1f), which was completely different from both the elongated fibers of Phe and the large crystal observed for Ile. Thus, the self-assembly of Phe was altered by the presence of Ile, indicating a two-component co-assembly resulting in different structures. Similarly, the presence of flake-like nanostructures in the equimolar mixture of Phe and Met (Figure 1g) indicated their interaction and co-assembly. The interactions were quite strong, as upon mixing of two separately prepared amino acid solutions, the fabrication of different structural ensembles was observed (Figure S3). Intriguingly, both elongated fibers (for Phe) and long crystals (for Gly) or flake-like structures (for Ala) were observed in the equimolar mixtures Phe with Gly (Figure 1h) or Ala (Figure 1i), respectively. Hence, these two amino acid pairs showed reduced affinity to interact with each other and independently assembled, giving rise to their individual characteristic morphologies. As predicted, the microscopy studies showed that Phe synergistically co-assembled with Ile and Met resulting in distinct morphological features not achievable through single component assembly, whereas the Phe:Gly and Phe:Ala mixtures maintained the native supramolecular structures characteristic of the individual components, indicating a self-sorting process.

Figure 1.

HRSEM images of (a) Phe, (b) Ile, (c) Met, (d) Gly, (e) Ala alone, (f) Phe:Ile (1:1) mixture, (g) Phe:Met (1:1) mixture, (h) mixed system of Phe and Gly (1:1), inset showing edge of the image specifies the growth of Gly crystal, and (i) mixed system of Phe and Ala (1:1) contained both Phe fiber and Ala flake structures. Scale bars are indicated in each image.

Mass spectrometry has recently emerged as an important tool to understand the interaction of small molecules with amyloid forming building blocks and metaclusters.^35–38 To detect the co-assembled and self-sorted amino acid systems, we used electrospray ionization mass spectrometry. For the Phe:Ile mixture, in addition to m/z peaks at 130 and 164 in negative ion mode corresponding to Ile (I) and Phe (F), respectively, the signals for the Ile-Phe (IF) complex at 295 and 317 were also observed (Figure 2a). This suggested the formation of a co-assembled structure by Phe and Ile. Similarly, along with peaks for individual Met (M) and Phe, the signals of the Met-Phe (MF) composite were also obtained for the Phe:Met mixture (Figure 2b). However, mixing of Gly (G) with Phe resulted in the characteristic signals of the single components (Gly at 76 and Phe at 166 in positive ion mode) only, but no signal corresponding to their complex mass was observed, signifying self-sorted assembly (Figure 2c). Likewise, only monomeric peaks at 90 and 166, corresponding to Ala (A) and Phe were detected for the Phe:Ala system (Figure 2d), specifying a self-sorted assembly mechanism.

Figure 2.

Mass spectra of (a) Phe:Ile (1:1), (b) Phe:Met (1:1), (c) Phe:Gly (1:1), and (d) Phe:Ala (1:1) mixtures. “IF” and “MF” represent the notation for Ile-Phe and Met-Phe complexes, respectively.

Wide-angle powder X-ray scattering (WAXS) techniques were employed to probe the molecular arrangement of pristine assemblies and co-assemblies. All the examined amino acids and their mixtures retained an ordered molecular packing in supramolecular aggregates, as indicated by the presence of relatively sharp peaks (Figure 3). A self-sorted system would be expected to provide a diffraction pattern corresponding to the linear combination of the individual amino acid peaks, while a co-assembly system would afford a different diffraction pattern.²³ The powder X-ray diffraction pattern of Phe implied the presence of a monohydrate phase under the chosen experimental conditions.³⁹ The diffraction pattern of the Phe:Ile mixture differed from the diffraction pattern obtained for either Phe or Ile alone (Figure 3a). Moreover, the additive diffraction peaks of Phe and Ile did not overlay with the peaks obtained for the mixed system. Most notably, the correlation peak corresponding to the interlayer separation at ~13.5 Å shifted to ~14.5 Å, indicating the formation of a different composite. This phenomenon was more prominently observed in the powder diffraction pattern of Phe:Met system (Figure 3b), suggested by a large change in the diffraction pattern at wide angles. This strongly demonstrated supramolecular coassembly for this mixed system at the molecular level. However, the diffraction peaks of the Phe:Gly system overlaid on the additive diffraction peaks of individual Phe and Gly, without any shifts (Figure 3c). This again underlined the fact that Phe and Gly self-sorted under the mixed condition and self-assembled separately. Analogous additive diffraction peaks were also observed in the Phe:Ala system (Figure 3d), further supporting their self-sorting.

Figure 3. WAXS signals obtained from dried powder of single amino acids and their mixtures.

(a) Phe (green), Ile (red), Phe:Ile-additive (black), Phe:Ile-experimental (blue); (b) Phe (green), Met (red), Phe:Met-additive (black), Phe:Met-experimental (blue); (c) Phe (green), Gly (red), Phe:gly additive (black), Phe:gly experimental (blue); and (d) Phe (green), Ala (red), Phe:Ala-additive (black), Phe:Alaexperimental (blue). For (a) and (b), the colored section has been magnified in the inset.

Physical vapor deposition (PVD) of short peptides has been employed to construct large, homogeneous arrays of bionanostructures on macroscopic surfaces.⁴⁰ The use of selected amino acid mixtures in the experimental process will allow to develop either uniform or layer-by-layer deposition of mixed amino acids onto a surface with desired properties. To demonstrate this, we studied the PVD of co-grinded all four amino acid combinations in a vacuum chamber at a temperature of 230 °C and pressure 1 × 10^-6 mbar. Under this low pressure, the solid powder of amino acids directly converted into vapor through a sublimation process upon temperature increase. The HRSEM image of the resulting surface deposition of Phe:Ile showed a horizontally oriented, homogeneously distributed array of flake structures (Figure 4a). The evaporation also produced a uniform layer over a large surface area. The chemical composition of the surface layer produced by vapor deposition was examined by time-of-flight secondary ion mass spectrometry (ToF-SIMS).^41,42 Previous reports showed that ToF-SIMS can provide insights regarding the structural orientation of larger protein over a surface based on the intensity differences of secondary ions stemming from amino acids located asymmetrically within the protein.⁴³ It can also allow imaging within a phase-separated lipid membrane with a small lateral resolution of ~100 nm and quantification of the lipid composition within small regions of the bilayer.⁴⁴ The chemical imaging of a membrane monolayer portion using SIMS provided insight regarding the abundance ratio of particular atoms, confirming particular biological processes.⁴⁵ At a smaller length scale, this method was employed to identify the presence of two different dipeptide building blocks in a preformed nanostructure prepared in solution by the co-assembly of two monomers.²² The result confirmed that the nanotube was indeed a co-assembly of the two monomers. Herein, we have employed ToF-SIMS to understand the chemical composition of a surface prepared by the vapor deposition technique of a solid amino acid mixture to identify their co-evaporation or separate evaporation. High-resolution mass spectra in positive ion mode of the same deposited surface one characterized by HRSEM showed peaks at 132, corresponding to Ile, and at 166, corresponding to Phe, pointing out the presence of both amino acids in the uniform layer (Figure S4). From the intensity of mass peaks or measuring the total ion counts, it is possible to shed light on the relative amounts of two amino acids present over the deposited surface.^22,45 The presence of peak at 297 corresponding to Phe:Ile conjugate indicated the co-assembly formation in solid state during co-grinding and vapor deposition (Figure S4). Chemical ion mapping was applied by selected specific ions from mass spectra (132 or 166) to observe their location over a precise area (Figure 4b). By separately fixing the red color for mass 166 (Phe) and green color for ions corresponding to mass 132 (Ile), a flake-like structure was observed for both the amino acids in the surface layer. Furthermore, co-assembly of these two amino acids was validated by overlapping the red and green channels. As observed in the far right image of Figure 4b, the merge region demonstrating distinctive yellow color flakes arises due to the co-assembly of Phe and Ile at the nanoscale. The co-assembly nature of these flakes was further confirmed via ion mapping by fixing the mass of 297 (Phe:Ile conjugate) (Figure S5). This result confirmed that Phe and Ile interacted at the atomic level and co-assembled during the grinding. Thus, the two amino acids evaporated simultaneously at high temperature, and the deposited surface embodied the side chain characteristics of both amino acids separated from each other at the molecular level. A similar result was also obtained by vapor deposition of Phe:Met, indicating their co-assembly in solid state (Figures S6 and S7). In contrast, the HRSEM image of the Phe:Gly deposited surface clearly showed the presence of two distinct layers (Figure 4c). Phe evaporated earlier to generate a uniform layer of flakes, which was later covered by the deposition of the Gly layer composed of small particles. Mass spectra obtained from ToF-SIMS analysis of the deposited surface indicated the occurrence of both amino acids (Figure S8). Chemical ion imaging of the ion corresponding to mass 166 (Phe) by labeling in red color showed a flakes structure (Figure 4d). While the Gly (76) showed small particle structures labeled in green color, although the signal was a very weak characteristic of Gly. Combination of both the ions mass for a single imaging clearly represented the presence of small green Gly particles over the red flake structures of the Phe layer. The formation of two different layers was also observed for the composition of the surface produced by the vapor deposition of Phe:Ala (Figure S9). The Ala layer composed of small particles fabricated initially was completely covered by the large flake-like structures of Phe, which evaporated later (Figure S9). The mass spectra obtained from ToF-SIMS analysis of the deposited surface showed only a peak corresponding to Phe, as Ala particles remained hidden under the large Phe flakes (Figure S10). The two layers produced by separate evaporation of Ala and Phe fabricated a very thick surface, which caused difficulties in capturing an image on ion mapping. The formation of two different layers further confirmed that these two amino acid pairs could not form a co-assembly during co-grinding, but rather self-sorted into distinct assemblies with different interlayer separation distances.

Figure 4. Coated surfaces after vapor deposition.

(a, b) Phe:Ile deposition: (a) HRSEM image, (b) ToF-SIMS ion image of (left to right) Phe labeled in red color, Ile labeled in green color, and overlapping of Phe and Ile with corresponding color. (c, d) Phe:Gly deposition: (c) HRSEM image, (d) ToF-SIMS ion image of (left to right) Phe labeled in red color, Gly labeled in green color, and overlapping of Phe and Gly with the corresponding color.

To evaluate the thermodynamic properties of Phe interaction with the other amino acids, isothermal titration calorimetry (ITC) measurements were performed. A freshly prepared solution of Phe (1 mM) was titrated into a cell containing each of the other amino acids (2 μM) to measure the corrected heat and the enthalpy values. Results of the titration profile and the thermodynamic values were calculated and are displayed in Figure S11. The calculated constants are presented in Table S2. Titration of Phe to Ile and Met resulted in endothermal peaks. In contrast, titration of Phe into buffer solution resulted in low exothermal peaks. However, both Gly and Ala titrations resulted in low endothermal interaction that was complete during the titration, as indicated by the exothermal peaks observed near the end of the experiment. The calculated dissociation constants (K_d) confirmed that Phe interacted strongly with Ile and Met (1.02 × 10^-4 and 1.85 × 10^-5, respectively), while both Gly and Ala showed much smaller K_d values (6.77 × 10^-6 and 7.08 × 10^-6, respectively). Interestingly, all the interactions of Phe with the examined amino acids showed an endothermic pattern, suggesting that the interaction is driven by entropy. This is indicated by the calculated values of Gibbs free energy, where ΔG < 0 in all cases due to the relation of TΔS > ΔH. It was reported that entropy-driven interactions occurred through hydrophobic contacts, whereas enthalpy-driven binding occurred through hydrogen-bonding and electrostatic interactions.^46,47 Thus, the ITC results suggested that the interaction of Phe with Ile and Met was dominated by hydrophobic interactions, as depicted in Scheme 1.

We further examined the supramolecular interaction of Phe with Ile or Gly by performing MD simulations on Phe:Ile (1:1) and Phe:Gly (1:1) mixture containing 1000 amino acids for each system. All amino acids were modeled using coarsegrained (CG) models based on Martini 2.2refP force field.⁴⁸ We refined the main-chain representation by adding two virtual charges to the main-chain bead to mimic the electric dipole moment of single amino acids (Figure 5a and details in Supporting Information). An 1 μs-long simulation was performed on each system using Gromacs-2018.3 package⁴⁹ at 300 K and 1 atm. The two systems showed distinct assembly pathways (Figure 5b,c). It can be seen from Figure 5b, starting from disordered states, molecules in Phe:Ile system first aggregated quickly into small clusters at 60 ns, which then started to fuse together into larger aggregates. At 200 ns, a single spherical assembly was formed, while other molecules scattered around and distributed almost evenly in the solution. This assembly gradually grew bigger in the remaining simulation time, and finally a large spherical assembly was formed. This aggregation-fusion assembly pathway can be further shown by tracking the number of clusters in the system as a function of time (Figure 5d, red line). The cluster number increased rapidly and reached a peak value of 7 at 60 ns and then decreased quickly to 1, corresponding to the quick formation and fusion of small clusters. In sharp contrast, the aggregation ability of Phe:Gly system was much weaker (Figure 5c). A small aggregate was formed at 60 ns and then gradually grew into a larger assembly by recruiting dispersed molecules. The number of clusters quickly reached 1 and fluctuated around this value in the remaining simulation time (Figure 5d, green line). Furthermore, the largest cluster in Phe:Ile system was much larger than that in the Phe:Gly system throughout the simulation time (Figure 5e). These results reveal that the Phe:Ile system has higher aggregation propensity than the Phe:Gly system.

Figure 5. Results of MD simulations on Phe:Ile (1:1) mixture and Phe:Gly (1:1) mixture.

(a) All-atom and CG model of single amino acids (Phe, Ile, and Gly). In the CG model, main-chain beads, N and C terminal virtual beads, and side chain beads are colored red, green, purple, and yellow, respectively. (b, c) Snapshots of Phe:Ile (b) and Phe:Gly (c) systems at 0, 60, 200, and 1000 ns. (d) The number of clusters as a function of simulation time. (e) The number of amino acid molecules in the largest cluster as a function of simulation time. (f-h) Number of contacts between Phe and Phe (f), between Ile and Ile (or between Gly and Gly) (g), and between Phe and Ile/Gly (h). (i) The snapshot of the largest cluster in Phe:Ile and in Phe:Gly systems at 1.0 μs.

To understand the distinct aggregation propensity of Phe:Ile and Phe:Gly systems, we analyzed the molecule-molecule interactions by calculating the average number of contacts between amino acids in each system (larger contact number corresponds to stronger interaction). A contact is considered if the minimum distance between two amino acids is within 0.65 nm. The average contact number is defined as the total number of contacts divided by the total number of molecules. The contact numbers between homo-molecules in Figure 5f,g show that the order of interaction strength from strongest to weakest is as follows: Phe-Phe > Ile-Ile > Gly-Gly. The contact number between Phe and Ile is extremely larger than that between Phe and Gly (Figure 5h), indicating stronger interactions of Phe with Ile than with Gly and the importance of hydrophobic interaction between Phe and Ile. Interestingly, from the snapshot of the largest cluster in the two systems at 1 μs (Figure 5i), we found that the largest cluster in Phe:Ile system is a mixture of Phe and Ile amino acids, while the cluster in Phe:Gly system consists of almost purely Phe (with three Gly molecules attached on the surface of the cluster). All of these results indicate that two amino acids from the same subgroup (Phe and Ile) have a higher propensity to interact with each other in mixture and co-assemble, while the members of different subgroups (Phe and Gly) have a lower tendency to co-assemble.

To understand whether the proposed rule based on interlayer separation to govern the co-assembly or self-sorting behavior of amino acids is valid as a general law, we have extended our study to other amino acids. Further investigation showed that Phe co-assembled also to produce different structures with all other amino acids, which have comparable interlayer spacing, namely Trp, Leu, and Val (Figure S13). However, with all other amino acids displaying extremely different layer separation, including Asp, Thr, and Pro, Phe showed self-sorting assembly (Figure S13). These studies firmly establish the proposed rule of amino acids interaction based their interlayer distances.

The results reported here demonstrate that the interaction of amino acids is strongly driven by matching of their interlayer separation distances. Having an interlayer spacing of ~15 Å, Phe strongly interacted with similar layer-spacing amino acids, Ile (~14 Å) and Met (~15 Å) (Scheme 1). However, such an interaction was considerably suppressed with amino acids of much lower interlayer separation, Gly (~5 Å) and Ala (~6 Å) (Scheme 1). This result is also consistent with the attempted co-crystallization of several hydrophobic amino acids and their derivatives with mixed chirality.^50,51 Although no complex of two amino acids with the same chirality has been described, the observation of poor quality co-crystals for L-Phe:D-Ala and L-Ile:D-Ala is highly consistent with our theory. Also, the crystal quality was found to significantly improve when D-Ala was replaced by amino acids with larger side-chains strongly amenable to layer matching guided interaction. Thus, the hypothesis presented in this study may be extended beyond the natural L-amino acids.

Conclusion

In conclusion, amino acids, peptides, and proteins constitute the fundamental functional units of life. Although prediction of the assembly of protein subunits or peptide molecules is highly explored, there are currently no insights to envision the compatibility of different amino acids to form synergistic supramolecular aggregates. The experimental evidence presented in this study demonstrates that the individual amino acid preferentially co-assembled with the group of amino acids showing a similar interlayer separation in a single crystal X-ray structure. Furthermore, the synergistic self-assembly was maintained during PVD confirming the robustness of our hypotheses. This study lays the basis for a different paradigm to understand the interactions among individual amino acids in proteins as well as opens a direction for the generation of different amino acid-based multifunctional supramolecular assemblies both in solution and on solid surfaces.

Materials and methods

materials. All the amino acids were purchased from Sigma (purity >98%).

Scanning Electron Microscopy (SEM)

The amino acids were dissolved in deionized water by heating at 90 °C at a concentration of 2 mg/mL, followed by gradual cooling of the solutions. For coassembly, 2 mg of each amino acids were dissolved in 1 mL of deionized water by a similar procedure. A 5 μL aliquot was allowed to dry on a microscope glass coverslip at ambient conditions overnight and coated with Au. SEM images were recorded using a JSM-6700F FE-SEM (JEOL, Tokyo, Japan) operating at 10 kV.

Mass Spectrometry

The samples of amino acids were prepared for mass spectrometry by dissolving them at a concentration of 1 mM, in deionized water by heating to 90 °C. Mass spectrometry was recorded using an Acquity UPLC system coupled to a TQD XEVO triple quadrupole ESI source mass spectrometer system from Waters (Milford, MA, USA).

Wide-Angle X-ray Scattering (WAXS)

The amino acids were dissolved in deionized water by heating at 90 °C at a concentration of 2 mg/mL for individual amino acids or (2 + 2) mg/mL for coassembly. The assembled structures were lyophilized and poured inside a quartz capillary of diameter 1.5 mm. WAXS measurements were performed using an in-house X-ray scattering system, with a GeniX (Xenocs) low divergence Cu Kα radiation source (wavelength of 1.54 Å) and a scatterless slits setup.⁵² Two-dimensional scattering data, with a momentum transfer wave vector (q) range of 0.07-2.5 Å^-1 at a sample-to-detector distance of about 160 mm, were collected on a Pilatus 300 K detector (Dectris, Baden-Daettwil, Switzerland) and radially integrated using Matlab (MathWorks, Natick, MA, USA) based procedures (SAXSi). Calibration was performed using silver behenate. The scattering data of the empty capillary were collected as background and used to subtract the quartz capillary and spurious scattering from the WAXS system itself, for example, Kapton vacuum windows and air gaps. For the measurement in solution, amino acids of above-mentioned concentration in water were sealed in quartz capillaries with a 1.5 mm diameter and used for the experiment. The scattering data of the solution without the amino acids were collected as background and used for subtraction.

Physical Vapor Deposition (PVD)

The amino acids were deposited on different substrates, such as glass and silicon wafer, using a biomolecule vapor deposition method in PVD system custom-built by El-Tan Technologies (Hod Hasharon, Israel). In a typical method, ~20 mg of lyophilized powder of either pure amino acid or mixture was ground using a mortar and pestle and placed in a small copper boat to serve as the source material. The substrate was placed above the source at a vertical distance of ~2 cm. The chamber was then set to 230 °C with a heating rate of 10 °C/min at a constant pressure of 1 × 10^-6 mbar and kept for 30 min at the final temperature. The resulting products were collected on the downwardfacing side of the substrate.

Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS)

The vapor deposited layer of amino acids on silicon wafer was analyzed by PHI Model 2100 TRIFT II ToF-SIMS instrument. The system used a pulsed primary ion beam to desorb and ionize species from the amino acids surface. The resulting secondary ions were accelerated into a mass spectrometer, where they were mass analyzed by measuring their time-of-flight from the sample surface to the detector. In addition, an image was generated by rostering a finely focused beam across the sample surface. Due to the parallel detection nature of TOF-SIMS, the entire mass spectrum was acquired from every pixel in the image. The ions related to m/z 132, m/z 166, and m/z 76 were used to identify and evaluate the ionic image of Ile⁺, Phe⁺, and Gly⁺, respectively. The mass spectrum and the secondary ion images were then used to determine the composition and distribution of sample surface constituents.

Isothermal Titration Calorimetry (ITC)

Fresh monomeric Phe (1000 μM) was dissolved in 40 mM PBS, and Ile, Met, Gly and Ala were diluted in PBS to a working concentration of 2 μM. The solutions were kept at 25°C and incubated at that temperature before measurement for 10 min. A sample of 300 μL of the amino acid solution was inserted into the Nano ITC low volume cell (TA Instruments, Newcastle, DE, USA), and the titrating syringe was filled with 50 μL Phe solution. The system was allowed to reach a stable temperature of 25°C along 2000 s and collected baseline for 500 s. Subsequently, Phe was titrated to the amino acid solution or PBS as a control. Titration was carried out in 5 μL aliquots and allowed to equilibrate for 400 s before the next drop, along 10 drops, of total 47.5 μL (first drop was half volume). The resulted isotherm was analyzed using Nanoanalyze software using an independent interaction model. Baseline correction was performed by titrating Phe to the PBS blank.

Coarse-Grained MD Simulations

The martini force field^48,53 has been widely used in the study of peptide self-assembly.^54,55 But as the main-chain of each residue is described using a single CG bead, it fails to describe single amino acids because the electrical dipole moment of main-chain cannot be modeled using a single bead. Inspired by the off-center charge model introduced in Martini 2.2P force field for charged residues,⁴⁸ we added two virtual charge points around the main-chain bead, while keeping the real main-chain bead uncharged. To determine the partial charge of these two virtual charge points, we built the all-atom structure of Phe, Ile, and Gly, calculated their electrostatic potential using Hartree-Fock methods under tight-binding approximation, and obtained the restrained electrostatic potential (ESP) charges by fitting the electrostatic potential using the AmberTools package.⁵⁶ The charges of the N terminal amino group (-NH₃⁺) and of the C terminal carboxyl group (-COO^-) are listed in Table S3. Based on the dipole moment calculated using these charges, we set the virtual charges of the N and C terminal as +0.3 and -0.3 e, respectively.

Supplementary Material

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsnano.8b07775.

Layer arrangement of the amino acids crystal structures, HRSEM images of simple mixing, mass spectrum, and ion images obtained from ToF-SIMS of Phe:Ile and Phe:Met decorated surfaces, mass spectrum and HRSEM images of Phe:Gly and Phe:Ala decorated surfaces, ITC curve and calculated constants, CG model for amino acids, HRSEM images of amino acids and their combinations (Phe, Trp, Val, Leu, Pro, Thr, Asp, Phe:Trp, Phe:Val, Phe:Leu, Phe:Pro, Phe:Thr, Phe:Asp), classification of amino acids based on interlayer distances, calculated charges on N-terminal and C-terminal of Phe, Ile, and Gly (PDF)