Molecular Engineering of Rigid Hydrogels Co-assembled from Collagenous Helical Peptides Based on a Single Triplet Motif

The potential of ultra-short peptides to self-assemble into well-ordered functional nanostructures makes them promising minimal components for mimicking the basic ingredient of nature and diverse biomaterials. However, selection and modular design of perfect de novo sequences are extremely tricky due to their vast possible combinatorial space. Moreover, a single amino acid substitution can drastically alter the supramolecular packing structure of short peptide assemblies. Here, we report the design of rigid hybrid hydrogels produced by sequence engineering of a new series of ultra-short collagen-mimicking tripeptides. Connecting glycine with different combinations of proline and its post-translational product 4-hydroxyproline, the single triplet motif, displays the natural collagen-helix-like structure. Improved mechanical rigidity is obtained via co-assembly with the non-collagenous hydrogelator, fluorenylmethoxycarbonyl (Fmoc) diphenylalanine. Characterizations of the supramolecular interactions that promote the self-supporting and self-healing properties of the co-assemblies are performed by physicochemical experiments and atomistic models. Our results clearly demonstrate the significance of sequence engineering to design functional peptide motifs with desired physicochemical and electromechanical properties and reveal co-assembly as a promising strategy for the utilization of small, readily accessible biomimetic building blocks to generate hybrid biomolecular assemblies with structural heterogeneity and functionality of natural materials.


Scanning Electron Microscopy (SEM).
A 5 L aliquot was allowed to dry on a microscope glass cover slip under ambient conditions overnight, and then coated with Au. SEM images were recorded using a JSM-6700F FE-SEM (JEOL, Tokyo, Japan) operating at 20 kV.
Absorbance Kinetics Measurement of the Gels. Samples of 150 μL of each hydrogel were placed in wells of a 96-well plate. Absorbance at 350 nm was measured every 3 min for 2 h using a TECAN Infinite M200PRO plate reader.

Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS).
A bundle of thin fibrils of hybrid hydrogels was prepared between two thin capillaries by drying the gels. The dried fibril was deposited on a silicon wafer and analyzed by a PHI model 2100 TRIFT II ToF-SIMS instrument. The system useda 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 120 and 70 m/z were used to identify and evaluate the ionic image of the co-assembled hybrid hydrogels. The mass spectrum and the secondary ion images were then used to determine the composition and distribution of sample surface constituents.
Atomic Force Microscopy. AFM images were obtained by depositing 5 μL of the hydrogel solution onto freshly cleaved V1 grade mica (Ted Pella) immediately after preparation. The samples were allowed to dry under ambient conditions for 24 h. The samples were imaged using AFM (JPK Instruments AG) performed using Nano Wizard 3 with 5 N/m spring constant tips and a resonance frequency of ∼150 kHz in soft tapping mode. The images were processed and analyzed by the JPK Data Processing software. The AFM nanoindentation experiments were performed using the same AFM. The force curves were obtained using the commercial software from JPK and analyzed by a custom-written procedure based on Igor Pro 6.12 (Wavemetrics Inc.). Silicon nitride cantilevers (Budget Sensors Company with the triangle shape, length 100 µm, width 16 µm, thickness 520 nm, frequency in air: ∼30 kHz) were used in all experiments. In a typical experiment, the QI mode was used to detectthe Young's modulus (conditions: pixels: 126 × 126; Z length: 0.3 μm; extend and retract speed: 2μm s −1 , Z resolution: 80 000 Hz). The spring constant of the cantilevers was 0.27 N m −1 , and the maximum loading force was set at 1.34 nN. All AFM experiments were carried out in air at room temperature. In a typical measurement, the cantilever was brought above the crystals with the help of an optical microscope. Then, the cantilever was extended to the surface of the crystal and retracted at a constant speed of 2μm s −1 . The extending and retracting force curves were recorded. The Young's modulus values of the crystals were obtained by fitting the extend curve to the Hertz model.
Density Functional Theory Calculations. Electromechanical properties were predicted from periodic DFT calculations 1 on the single crystal using the VASP 2 code. Electronic structures were calculated using the PBE functional 3 with Grimme-D3 dispersion corrections 4 and projector augmented wave (PAW) pseudopotentials 5 . The crystal structure was optimised using a plane wave cut-off of 600 eV with a 4x4x4 k-point grid. A finite differences method was used to calculate the stiffness tensor, with each atom being displaced in each direction by ± 0.01 Å, and piezoelectric strain constants and dielectric tensors calculated using Density Functional Perturbation Theory 6 (DFPT), with a plane wave cut-off of 600 eV and k-point sampling of 2x2x2. Young's Moduli were derived from the stiffness and its inverse compliance matrix components. Values are presented as Voigt-Reuss-Hill averages 7,8 . Crystal structures were visualised using VESTA 9 .

Details of MD Simulation Setup
400 peptide molecules were placed in a cubic 17 nm X 17 nm X 17 nm box containing 2000 DMSO molecules. The Fmocresidue was parametrized with ParamChem 10,11 , which provides the CHARMM General Force Field (CGenFF) 12 parameters. Additionally, the partial charges for Fmoc atoms were derived using the Restrained Electrostatic Potential (RESP) 13 scheme based on quantum mechanical calculations with Gaussian 14 , followed by charge fitting with Antechamber 15 . The peptide molecules and DMSO were represented by the CHARMM 36m 16 force field, and solvated with 72,000 molecules of CHARMM-modified TIP3P 16 explicit water.
MD simulations were carried out using the Gromacs 2018.4 17 package with a time step of 2 fs using the Leap frog integrator. 18 Bond lengths to hydrogen were constrained using the LINCS 19 (protein) and the SETTLE 20 (water) algorithms. Background ions were added to neutralise full protein formal charges. Long-range electrostatics were treated by the Particle mesh Ewald (PME) method. 21 Protein and non-protein molecules (water, DMSO and ions) were coupled separately to an external heat bath (300 K) with acoupling time constant of 1 ps using the velocity rescaling method. 22 All systems were minimised for 100 ps, and equilibrated for 500 ps in constant volume NVT ensemble followed by another 500ps of NPT equilibration with the reference pressure set at 1 bar and a time constant of 5ps using the Parrinello-Rahman barostat. 23 The production runs were carried out for 600 ns in constant pressure NPT ensemble. Structures were saved every 20 ps. MD trajectories of the Fmoc-Phe-Phe assembly and Fmoc-Phe-Phe/Fmoc-Gly-Pro-Hypco-assembly are provided in movies S1 and S2. SUPPLEMENTARY NOTES S1. Computed Properties from MD Simulations S1.1 Supramolecular Clustering Analyses. To compute the formation of self-assembling and co-assembling molecular clusters as a function of time, we used the gmxclustsize routine in Gromacsat a molecule-molecule cut-off distance of 0.5 nm. Figure S12 shows the number of peptide clusters formed as a function of simulation time. The pure self-assembling Fmoc-Phe-Phe and the four co-assembling Fmoc-Phe-Phe/Fmoc-tripeptides revealed the formation of peptide clusters evident from a rapid drop in the number of clusters (#clusters) within the first 10 ns of the simulation (inset, Figure S12 and main text Figure 6). As expected, pure Fmoc-Phe-Phe formed the largest clustersas the Phe-Phe units formed tight supramolecular networks via π−π extensive packing. The rank order of clusters formation amongst the co-assemblies was Fmoc-Phe-Phe/Fmoc-Gly-Pro-Hyp>Fmoc-Phe-Phe/Fmoc-Gly-Pro-Pro>Fmoc-Phe-Phe/Pro-Pro-Gly≈Fmoc-Phe-Phe/Hyp-Pro-Gly. The superstructuring of pure Fmoc-Phe-Phe and of the fastest co-assembly system Fmoc-Phe-Phe/Fmoc-Gly-Pro-Hyp is shown in the main text Figure 6c-f. The superstructures formed for other co-assembled systems are shown in Figure S12b-g.

S1.2 Contribution from Inter-Peptide Hydrogen Bonds (HBs). The number of HBs formed
during assembly was calculated with a characteristic donor-acceptor distance of 0.3 nm and an angle of 20º using VMD software. 24 Figure S13 shows the number of HBs (#HBs) as a function

S1.3 Role of Encapsulated Water in Directing Hydrophobic and Hydrophilic Inter-Peptide
Contacts. In order to identify the different roles of the individual groups in the Fmoc-Phe-Phe self-assembly and the four co-assemblies in the formation of the hydrogel scaffolds, we computed the fraction of their solvent accessible surface area (SASA) relative to their initial randomly dispersed states (Figure 6i-k and S15a, b). The SASA fractions of Fmoc-Phe1-Phe2 assembly (where Phe1 is the first phenyl group and Phe2 is the second phenyl group in the sequence; a similar nomenclature is followed throughout for the residues in the tripeptidesaccording to their position in the sequence for direct comparison) rapidly droped during the first 25 ns for each group, in the order of Phe1 >Fmoc> Phe2. The gels remained in these super-structured states for the remainder of 600 ns MD (see Figure 6i). The mid-peptide Phe1 underwent maximum SASA loss indicating the potential to be buried inside the hydrophobic core of the aggregates with a propensity to engage in - T-stacks. The terminal Phe2 remained significantly more exposed to water, with the Fmoc unit intermediate relative to the two Phe units. After 600 ns, the SASA fraction values were 0.48 for Fmoc, 0.32 for Phe1 and 0.57 for Phe2 (see Figure 6i).
The SASA fractions of individual groups belonging to the Fmoc-Phe-Phe/Fmoc-Gly-Pro-Pro (Fmoc1-Phe1-Phe2/Fmoc2-Gly-Pro1-Pro2) co-assembly droped rapidly except for Pro2 ( Figure   S15a), which mostly remained exposed to the solvent (0.93 after600 ns). This is due to the charged C-terminus and "tail" position of this unit. The degree of burial was Phe1 > Fmoc1 > Phe2 ≈ Fmoc2 > Pro1 ≈ Gly> Pro2. The final SASA fraction values of 0.49 for Fmoc1, 0.37 for Phe1 and 0.59 for Phe2 suggest that Fmoc1-Phe1-Phe2 remained as buried during co-assembly as it did during the single self-assembly of Fmoc-Phe1-Phe2. In the co-assembling Fmoc2-Gly-Pro1-Pro2 peptide, Fmoc2 remained buried to a similar degree as Phe2 with a final SASA fraction value of 0.65, followed by Pro1 and Gly (~0.75 each). Pro1 and Gly share similar SASA values as the latter is very small and does not exhibit any site for H-bonding or - stacking.
Therefore, its exposure to the solvent is solely driven by Pro1. Similar rapid loss of SASA fractions was observed for theFmoc1-Phe1-Phe2 groups (Phe1 > Fmoc1 > Phe2) as opposed to the Fmoc2-Hyp-Pro-Gly groups in the co-assembly of Fmoc1-Phe1-Phe2/Fmoc2-Hyp-Pro-Gly ( Figure S15b). Gly, the terminal group of the tripeptide, remained mostly exposed to the solvent (as opposed to Gly in Fmoc2-Gly-Pro1-Pro2), sometimes displaying a larger fraction (>1) than its initial randomly dispersed state. Pro as the penultimate group in Fmoc2-Hyp-Pro-Gly remained buried to a similar extent (0.8) as Pro1 in Fmoc2-Gly-Pro1-Pro2. Hyp following Fmoc remained more exposed (0.85) than Pro following Hyp. The order of the degree of SASA loss for the second tripeptide only was Fmoc2 > Pro >Hyp>Gly.
Rapid, significant, and sustained loss of SASA fractions was also observed for individual groups belonging to Fmoc1-Phe1-Phe2 (converged values of 0.35 for Phe1, 0.48 for Fmoc1 and 0.57 for Phe2) in the Fmoc1-Phe1-Phe2/Fmoc2-Pro1-Pro2-Gly (Fmoc-Phe-Phe/Fmoc-Pro-Pro-Gly) co-assembly (see Figure 6j) compared to groups in Fmoc2-Pro1-Pro2-Gly (0.6 for Pro2, 0.65 for Fmoc2, 0.85 for Pro1 and 0.95 for Gly). The terminal solvent-exposed Gly group transiently showed fractions larger than 1, like Fmoc2-Hyp-Pro-Gly ( Figure S15b). The placement of Pro2 in the second position of the tripeptide as opposed to the terminal position in Fmoc2-Gly-Pro1-Pro2 significantly improved its participation in the assembly by burying in the core. As expected, SASA fractions of individual groups belonging to buried Phe1 were the lowest, as seen in all other cases in the co-assembly of Fmoc1-Phe1-Phe2/Fmoc2-Gly-Pro-Hyp.
However, interestingly, the fraction SASA losses of the Fmoc-Gly portion of Fmoc2-Gly-Pro-Hyp were almost at par with Fmoc1-Phe1-Phe2 (order Phe2 >Gly ≈ Fmoc1 > Fmoc2 ≈ Phe1), in contrast to the other co-assemblies (see Figure 6k). The MD data indicates that the rate of formation of Fmoc-Phe-Phe/Fmoc-Gly-Pro-Hyp superstructures may be the fastest amongst the co-assemblies, consistent with the measured kinetic data (Figure 2). The simulations also predict that replacing Pro with Hyp at the third position of the tripeptide sequences improves their ability to formwater-mediated hydrogen bonds in the hydrogel scaffolds (see Figure S13a and S13b for comparison). and Fmoc-Phe-Phe/Fmoc-Gly-Pro-Hyp within a threshold of 8 Å. These plots were generated by sampling the FEL during the last 100 ns of the simulation (500-600 ns). The FEL of Fmoc-Phe-Phe self-assembly was majorly T-shaped - stacking (planes of rings were perpendicular to each other) at a distance of ~5-6 Å (see Figure 6l and Figure S15c). There was also evidence of -parallelstacking or parallel displaced stackingat shorter distances of ~4 Å.

S1.4 Free Energy Mappingof Buried
The FEL for the co-assemblies of the Fmoc-Gly-Pro-Pro, Fmoc-Hyp-Pro-Gly and Fmoc-Pro-Pro-Gly tripeptides with Fmoc-Phe-Phe (Figure S15e, f and 6m) again revealed primarily Tshaped - stacking configurations (at or around 90 degrees) at a centroid-centroid distance of ~5-6 Å (red basin) with less prominent (cyan basin) - sandwich or parallel displaced stacking (180 degrees) at ~4 Å. The minimal basin for T-shaped stacking ( Figure S15d) of the Fmoc-Phe-Phe/Fmoc-Gly-Pro-Hyp system occured around 5-6 Å with an additional small minimum at longer distances of~8 Å (see Figure 6n). The maps show how both the local and long-range structure of the stacking network depend critically on the sequence of the peptides. These results show the resemblance between the π−π network in the co-assemblies between Fmoc-Phe-Phe and the two tripeptides with adjacent proline moieties and the very rigid pure Fmoc-Phe-Phe self-assembled structure, which produces a tightly packed peptide supramolecular organization, as reflected in the measured mechanical strengths (Figure 2).

Figure S1
Crystal structure of Fmoc-Gly-Pro-Hyp (1).Side-by-side H-bond connection of a single helical chain with two nearby helical chains. For visualization, the three helices are colored differently. Intermolecular H-bonds are represented by blue dotted lines.