Surface force measurements and simulations of mussel-derived peptide adhesives on wet organic surfaces

Using the SFA (Fig. 2), we determined the interaction force (F) as a function of distance (D) between mica (with radius R) and either a CH₃-terminated (hydrophobic) SAM surface or an OH-terminated (hydrophilic) SAM surface in a solution of MFP-3s peptide. In agreement with our previous full-length MFP measurements on SAMs (17), MFP-3s peptides exhibit strong adhesion forces between hydrophobic surfaces and moderate-to-weak adhesion forces between hydrophilic surfaces. Firstly, SAMs were adsorbed onto smooth (rms roughness 0.2 nm) gold surfaces and then mounted in the SFA opposite from a mica surface. Control force–distance measurements between mica and the SAMs are shown in Fig. S1. In each experiment, increasing aliquots of MFP-3s peptides were injected between the two surfaces before force measurements were performed, until the surfaces were saturated with bound peptides. At saturation, the surfaces are fully covered with a single layer of peptide. Further injection of the peptide beyond the saturation point resulted in both a decrease in adhesion and an increase in the peptide film thickness (Fig. S2), attributable to stacking of multiple peptide layers. Using the Johnson–Kendall–Roberts theory (24) for adhering surfaces, the measured interaction force per unit radius is converted into an adhesion energy per unit area (E_ad = F_ad/1.5πR). The average adhesion energy required to separate a peptide film from a hydrophobic and hydrophilic SAM (E_ad) was −7.7 ± 1.9 and −0.4 ± 0.3 mJ/m², respectively. Energy minima for each system were observed at distances (r_min) of 2.5 ± 0.2 and 2.7 ± 0.3 nm from the gold SFA surface for hydrophobic and hydrophilic systems.

To gain further insight into the adhesion mechanisms of this peptide, we performed umbrella-sampling simulations to estimate free energy (or potential of mean force) as a function of distance (25). PMF simulations yielded 1D free-energy profiles of MFP-3s peptide adhesion as a function of the peptide’s center-of-mass distance from the SAM substrate, where attractive and repulsive regions can be seen in Fig. 3. These results show a large −34.7 kT energy minimum for the peptide at 1.9 nm above the hydrophobic SAM. In contrast, the peptide is only bound with −6.2 kT of energy on hydrophilic SAMs at an equilibrium distance (r_min) of 2.1 nm, almost six times weaker than on hydrophobic surfaces. These energies were then converted to an adhesion per unit area (Fig. 3, secondary axis) by dividing the adhesion energy by the average peptide contact area of about 8–9 nm². The simulated adhesion strengths for the MFP-3s peptide on hydrophobic and hydrophilic SAMs were 15.9 and 2.9 mJ/m², respectively.

If we compare the ratios of the energy minima (at r_min) between simulations (Fig. 3) and experiments (E_ad) (Fig. 2), we find that on hydrophobic SAMs, E_sim/E_exp = (−15.9 mJ/m²)/(−7.7 mJ/m²) = 2.1, and on hydrophilic SAMs, (−2.9 mJ/m²)/(−0.4 mJ/m²) = 7.2. These ratios are approximate and are subject to change if the SFA error bars are strictly applied; however, adhesion from the simulations was always stronger than what was measured in experiments. Simulations exhibit increased peptide adhesion because only one surface is present, where all adhering residues must detach for separation to occur. However, in SFA experiments there are two opposing surfaces, and peptides can remain bound to the upper mica surface as the mica separates from the opposing SAM surface. On average, if symmetric adhesion occurs on the SFA, where half the Dopa residues bind to opposite surfaces, we would expect to measure approximately half the adhesive force in SFA experiments compared with simulations. However, if nonsymmetrical binding of Dopa to lower SAM surfaces (compared with upper mica surfaces) occurs, this discrepancy can vary considerably, as observed on hydrophilic surfaces. Despite these factors, both methods agree that MFP-3s peptides bind more weakly to, and at slightly further distances from, OH-terminated SAMs.

Overall, the experimental and computational measurements are in good agreement with one another. In the simulations, the peptide’s distance is measured from the SAM substrate to the peptide center-of-mass. In the SFA experiments, distance is measured from the SAM substrate to the opposing mica surface. The length of the SAMs′ alkane chains in both experiment and simulation were also the same. Therefore, we obtain a similar r_min by offsetting the simulated value by the peptide’s radius of gyration (R_g), which was around 0.8 nm (Fig. 4). From simulations, we find that (r_min + R_g) = 2.7 nm on hydrophobic SAMs and 2.9 nm on hydrophilic SAMs, precisely within the uncertainty observed in SFA experiments (2.5 ± 0.2 and 2.7 ± 0.3 nm, respectively). Even the difference in r_min between hydrophobic and hydrophilic surfaces (0.2 nm) was accurately reproduced in the simulations.

At crystalline oxide or mineral surfaces—such as mica or TiO₂—MFPs adhere in solution via coordination or bidentate bonds with Dopa’s hydroxyls and electrostatic interactions with charged residues (26, 27). However, at organic surfaces with varying polarities and thermally mobile surface groups, the adhesive interactions are expected to differ. To probe the atomic binding behaviors and orientations of MFP-3s peptides to the two surfaces, we performed replica-exchange MD (REMD) simulations, as described in Methods. Fig. 4 highlights the most dominant peptide conformations found from REMD on each surface, where Dopa tends to be oriented parallel (≤45° from the plane of the surface) near hydrophobic SAMs and perpendicular (>45° from the plane of the surface) near hydrophilic SAMs. This observation confirms our earlier hypothesis (17) that Dopa will take on parallel and perpendicular orientations on hydrophobic and hydrophilic surfaces, respectively. In contrast to the bidentate interactions observed at crystalline surfaces, only monodentate hydrogen bonds were observed between Dopa and the thermally mobile OH-SAM surface. The bond lifetimes of monodentate interactions are significantly shorter than bidentate interactions, which dramatically reduces the peptide’s adhesion to the OH-SAM surface compared with mineral or oxide surfaces. These results also reveal that Dopa, which is amphiphilic, sometimes orients outward toward the bulk solution with its hydroxyls hydrated, whereas tryptophan (which is purely hydrophobic) orients toward the hydrophobic interior of the peptide’s globular core. Peptide secondary structures are also highlighted in Fig. 4, where hydrophobic SAMs promote extended, flatter conformations, whereas hydrophilic surfaces reinforce more globular bulk states, similar to other small proteins on organic surfaces (28). Detailed secondary structural information is presented in Fig. S3 as a function of the peptide’s amino acid sequence, where a small but noticeable increase in β-bridging occurs on the surface of hydrophobic SAMs. However, because the MFP-3s peptide is intrinsically disordered, no major secondary structures persist.

From simulations, we extract the top three MFP-3s peptide conformations found in bulk solution and on hydrophobic and hydrophilic SAMs (Fig. 4, Left). We find that whereas the peptide spends little time in any single conformation (where every cluster persists ≤6% of the total simulation time), the top three dominant structures (labeled clusters a, b, and c in Fig. 4) exhibit similar morphologies. These trends were not limited to the top three peptide conformations, as can be seen in Fig. S4. However, for clarity, only the top three peptide clusters are presented in detail, because the remaining structures were variants of the first three conformations. To better characterize the energetics of peptide folding, we plot complimentary 2D free-energy surfaces (FES) to the right of each set (Fig. 4, Right), as a function of the peptide’s end-to-end distance (R_ee) and radius of gyration (R_g). In bulk water, MFP-3s peptides can traverse across multiple FES minima by increasing their end-to-end distance from 0.5 to 2 nm; however, this movement always occurs at constant R_g (0.8 nm). Higher R_g are only accessible to MFP-3s peptides under energy penalties of 5–7 kT; however, surfaces significantly modify these barriers. For clarity, each cluster is explicitly identified on the FES, typically residing at free-energy minima.

On hydrophobic and hydrophilic SAMs, peptide adhesion is observed in all three of the most dominant peptide clusters, where Dopa facilitates peptide binding in parallel and perpendicular orientations on CH₃- and OH-terminated SAMs, respectively. Additional free-energy basins appear near hydrophobic SAMs, where MFP-3s peptides take on either comparable (0.8 nm) or larger (1.1 nm) radii of gyration compared with bulk. Similarly, larger end-to-end distances are observed near hydrophobic SAMs (up to 3.2 nm, as shown in clusters a and c), but these states are separated from other states (e.g., cluster b) by about 6 kT. However, peptides may change their radii of gyration (through extension or contraction) without leaving the SAM surface, as shown in Fig. 4, Far Upper Right. Conversely, hydrophilic surfaces stabilize bulk peptide conformations, although peptides remain bound to the surface by polar interactions between the interface and Dopa hydroxyls. The reduction in the number of free-energy basins near hydrophilic SAMs is also striking, with peptides primarily taking on compact (small R_ee, small R_g) conformations near the SAM surface. Multiple end-to-end peptide distances are also observed on hydrophilic SAMs, although clusters a and c are only separated from cluster b by about 2–3 kT, because of the persistence of the bulk state.

We also observed a decrease in the number of intramolecular hydrogen bonds formed between MFP-3s peptide backbones (C=O and N-H groups) near hydrophobic SAM surfaces. Fig. S5 shows the emergence of a number of new peptide structures on hydrophobic surfaces that contain very few hydrogen bonds, thereby reinforcing more extended peptide conformations with high-contact surface areas. Hydrophilic surfaces, however, do not strongly perturb the distribution of peptide hydrogen bonds from bulk, thereby encouraging peptide conformations with smaller surface contact areas.

A number of Arg-Asp salt bridges are also observed in MFP-3s peptides on the surface of hydrophobic SAMs. Fig. S6 shows multiple FES plotted as a function of the proximities between charged Asp, Arg, and Lys residues, in addition to the residues’ positions relative to the SAM surface. Energy minima at small distances (e.g., ≤0.3 nm) indicate strong electrostatic interactions between oppositely charged residues. Asp-Arg salt bridges persist on hydrophobic SAMs, although Lys-Asp salt brides are not often observed because Lys favors the SAM surface instead, despite the reduction of hydrating water molecules at the interface. On hydrophilic SAMs, both Arg-Asp and Lys-Asp salt bridges are observed, although Lys and Arg maintain contact with the OH SAM surface. Interestingly, Arg is bound so tightly to Asp that it sometimes pulls the negatively charged residue with it upon binding to the surface.

We also plot FES for individual Dopa and Trp residues in the MFP-3s peptide (Fig. S7), to extract the per-residue energetics of Dopa/Trp rotation at organic interfaces, as a function of the residues’ distances to the surface. We observed that almost all Dopa residues were oriented perpendicular (90°) to organic surfaces far above the interface (or in random orientations that averaged to 90°) but subsequently rotated parallel (0° or 180°) on hydrophobic surfaces during approach. The energy required for Dopa to transition from perpendicular to parallel orientations ranges from 4 to 4.5 kT; however, Trp is more freely able to rotate on hydrophobic organic surfaces with an energy penalty of only 1–2 kT. Results were quite different on hydrophilic surfaces, where multiple Dopa and Trp orientations were observed close to 45° from the surface.

Whereas Dopa residues clearly contribute to robust adhesion in MFPs, there are other amino acids (such as Lys) that can perturb the conformations of Dopa. To better understand the chemical context between Dopa and adjacent charged residues, and what roles they play in binding to organic surfaces, we performed umbrella-sampling simulations, where we used either a single-capped Dopa residue (Fig. 5), a single-capped Lys residue (Fig. S8, Upper), or a combined Dopa-Gly-Lys peptide (Fig. S8, Lower) for each SAM surface. This technique allowed for the decomposition and subsequent identification of adhesive contributions from individual amino acids, while also providing information about their bound conformations. Lys was selected as a complimentary amino acid because it was recently shown to enhance Dopa adhesion (29), but the molecular mechanisms behind this behavior are highly contested. Fig. 5 shows the free-energy profile for Dopa, where multiple interesting features stand out. When Dopa settles at hydrophilic r_min, it binds to OH-terminated SAMs at a somewhat tilted (∼45°) angle, where Dopa’s hydroxyl and peripheral hydrogens bind to the negatively charged SAM hydroxyl oxygens. This orientation is also observed during MFP-3s peptide binding, where multiple Dopa residues were tilted on hydrophilic surfaces (Fig. S7). Per unit area, Dopa was over 4.5 times more adhesive to hydrophobic surfaces compared with the entire MFP-3s peptide and 11.6 times more adhesive to hydrophilic surfaces than the MFP-3s peptide. Similarly, Lys was over 10 times more adhesive per unit area on hydrophilic surfaces compared with the MFP-3s peptide (attributable primarily to Lys′s positive electrostatic attraction to the surface, which extends far into bulk) and was also 1.7 times more adhesive on hydrophobic SAMs compared with the MFP-3s peptide. Near hydrophobic SAMs, the charged Lys terminus is hydrated in bulk water; however, the alkyl chain remains inside the hydrophobic region near the interface. When Dopa and Lys were combined in a model tripeptide (Fig. S8, Lower), adhesion to surfaces was cooperative, where peptide free energies on both hydrophobic and hydrophilic surfaces remained attractive out to 3.5 nm (further into bulk than any of the individual amino acids). Therefore, whereas Dopa is a better adhesive (per unit area) on hydrophobic surfaces, and Lys a better adhesive (per unit area) on hydrophilic surfaces, adjacent Dopa- and Lys-containing molecules (29) are better overall adhesives on a wide variety of organic surfaces because of the adjacent Dopa- and Lys-containing molecules’ concerted binding motifs.

By using theoretical models of peptide adhesion, which highlight the nanoscale interactions between Dopa and organic surfaces, and by validating these models with SFA measurements, we have taken important first steps in revealing the atomic mechanisms behind Dopa adhesion. These results successfully bridge atomistic theories of Dopa adsorption to macroscale material measurements of novel peptides on a wide-variety of wet organic surfaces, thereby enabling future studies to synthesize and optimize stronger, next-generation underwater adhesives.

These results (highlighted in Fig. 6) also validate earlier theoretical hypothesis (17) about the preferred orientations of Dopa on hydrophobic and hydrophilic organic surfaces and have far-reaching implications regarding the mechanism of MFP adhesion to generalized organic surfaces, including cellular membranes. We have demonstrated that interfacial solvation is inextricably tied to the peptide’s adhesive performance: at the fluctuating, vapor-like CH₃/water interface (30), hydrophobic association mediates strong peptide adhesion, whereas at the hydrophilic OH/water interface, the competition between water and the peptide for hydrogen bonds to the surface drastically reduces the overall adhesion. By better understanding how biological surfaces stabilize or destabilize intrinsically disordered proteins (31), and by using unique amphiphilic residues that modulate adhesion to these surfaces, we have provided a template for more sophisticated studies that seek to optimize next-generation bioadhesives under a number of unique biological conditions.

Self-assembled monolayers were constructed by creating a fixed 2D grid of sulfur atoms in a diamond geometry in the middle of a unit cell (30). Then, a 10-atom alkane chain (CH₂)₁₀ was attached to both the top and bottom of each sulfur anchor, terminated by either a CH₃ group (for hydrophobic interfaces) or an OH group (for hydrophilic interfaces). SAM chains were left unfrozen to observe if fluctuations in surface spacing would promote bidentate bonding between MFP dopa residues and surface groups. Parameterization of SAMs were performed by Garde et al. (30). The net charge of each SAM was kept neutral. Box dimensions were on the order of 8 × 8 × 10 nm³, and periodic boundary conditions were implemented to mitigate system size effects and reduce computation. The system was then hydrated at room temperature with a semiisotropic barostat that froze the XY box dimensions, whereas the Z dimension was allowed to fluctuate to establish 1 bar of pressure. An unfolded MFP-3s peptide (with +1e net charge) was then added to the bulk solution above the SAM, alongside a chloride counter ion. The Z box dimension was subsequently fixed after bulk water density was reestablished at 1 bar of pressure in Z, where the system was ready to be simulated under a canonical (constant volume) ensemble.

Simulations were carried out using GROMACS 4.6.1 (32) on the Stampede supercomputer at the Texas Advanced Computing Cluster. REMD (33) simulations consisted of 70 replicas at increasingly higher temperatures that (after a number of preliminary values were tested) yielded an average exchange rate of 25% between adjacent replicas during the initial 10 ns, ranging in temperature from 288 to 506 K. Exchanges between replicas were attempted every 3 ps to allow an adequate mixing of states to occur. Replicas were initially heated for 20 ns, followed by a 200-ns production run at constant temperature; however, only the final 100 ns was analyzed and summarized here. Peptide topologies were derived from the AMBER03* force field (30, 34, 35). Partial charges for dopa residues (Dopa) were parameterized by fitting quantum gas-phase potentials of each atom using the restrained electrostatic potential (RESP) method (36), whereas other parameters (bond, angle, torsion, and Lennard–Jones constants) were taken from AMBER03*. The use of idealized Lennard–Jones and Coulomb potentials in simulations allowed for the reproduction of a number of interactions between molecules such as van der Waals forces, dispersion, hydrogen-bonding, hydrophobicity, and dielectric responses. Although more complex Hamiltonians can be implemented in MD that incorporate additional terms, the Hamiltonian in the AMBER03* force field is sufficient for capturing the most dominant molecular interactions. TIP3P rigid water molecules (37) were used to hydrate each system, where up to 15000 explicit water molecules were used per simulation box. Simulations were initially equilibrated with a weakly coupled, semi-isotropic Berendsen barostat (38) at 1 bar, with an isothermal compressibility of 4.5 × 10⁻⁵ bar⁻¹, and a velocity-rescaled thermostat (39) that maintained temperature at 300 K. After the box-dimensions converged, the unit-cell volume was fixed, and switched over to a canonical NVT ensemble under a Nosé-Hoover thermostat (40) for the remainder of the study. A leapfrog algorithm was used to integrate Newton’s equations of motion with an integration time step of 2 fs. Peptide and SAM molecular bonds were constrained using the LINCS algorithm (41), while water bonds were constrained using the SETTLE algorithm (42). Short-range electrostatic and Lennard–Jones forces were truncated at 1 nm, where long-range interactions took over using a Particle Mesh Ewald (PME) algorithm (43).

Additional simulations were carried out to ascertain the potentials of mean force (PMF) on MFP-3s peptides, Dopa, Lys, and Dopa-Gly-Lys peptides on hydrophobic and hydrophilic SAMs. PMFs were extracted by umbrella sampling (25) molecules along a reaction coordinate normal to the SAM surface (in 1 Å bins) for a total of 10 ns per bin. A spring with a constant of 8,000 kJ mol⁻¹ nm⁻² was used to sample each reaction coordinate, and the weighted histogram analysis method (WHAM) (25) was used to extract free-energy profiles from normalized probability measurements. Bayesian bootstrapping was used (in conjunction with WHAM) to generate PMF error bars.

Molecular graphics were generated with Visual Molecular Dynamics (VMD) 1.9.1 (44). The GROMACS tools g_hbond, g_traj, g_gyration, and g_cluster were used to measure the probabilities of intramolecular hydrogen bonding, peptide end-to-end distance (R_ee), the radius of gyration (R_g), and clusters of dominant peptide morphologies at room temperature. Hydrogen bonds were defined with an O-H spacing of 0.25 nm or smaller and O-H-N angles of 30° or less. R_ee was measured from the N-Gly center of mass to the C-Dopa center of mass. MFP-3s peptide conformations were clustered according to the criteria of Daura et al. (45), which compares protein backbones (excluding terminal amino acids) and groups them together based on a root mean square cutoff of 0.14 nm or less. Secondary structures were extracted using the DSSP tool (46).

Peptides were commercially synthesized using solid-phase peptide synthesis (GenScript). Tyr residues in the peptides were then hydroxylated in-house to Dopa by using commercially available mushroom tyrosinase (Sigma Aldrich), where the yield of Dopa conversion was controlled by fine-tuning the enzyme/substrate ratios and reaction times. The resulting MFP-3s peptides contained, on average, seven Dopa units per molecule, which was confirmed by amino acid analysis and mass spectrometry.

Self-assembled CH₃ and OH monolayers were prepared on molecularly smooth gold surfaces using a previously described technique (17, 47). First, a 45-nm layer of gold was deposited onto a freshly cleaved mica surface using electron beam deposition. Individual pieces of gold-coated mica were glued, with the gold side down, onto cylindrical glass SFA discs [radius (R), ∼2 cm] using a UV-curable adhesive. After the glue had set, the discs were immersed in ethanol, and the mica back-sheets were carefully peeled away from the underlying gold layers. The templating procedure results in cylindrically shaped and molecularly smooth gold surfaces with 0.2-nm rms roughness. After cleaving, the pristine gold surfaces were immersed in 1 mM solutions of either 1-undecanethiol (CH₃-SAM) (Sigma Aldrich) or 11-mercapto-1-undecanol (OH-SAM) (Sigma Aldrich) in ethanol for 18 h to adsorb the respective monolayer. Following the adsorption, the surfaces were rinsed with ethanol for 30 s to remove excess SAM molecules, dried under N₂, and installed in the SFA for force measurements.

SFA measurements were performed with a SFA 2000 (SurForce), and the full details of the SFA technique may be found elsewhere (48). Briefly, a freshly cleaved, back-silvered mica surface was glued to a cylindrical glass disk and installed in the SFA with the mica surface facing the SAM surface. Droplets of pH 3 buffer solution (0.1 M acetic acid plus 0.25 M KNO₃; Sigma Aldrich) were injected into the gap between the mica and SAM surfaces (∼50 μL total volume), and force measurements were performed between the surfaces in buffer solution. Picomolar amounts of MFP-3s peptide—suspended in the same buffer solution—were then injected into the gap solution between the two surfaces and allowed to adsorb and equilibrate for 30 min. Force measurements were then performed between the surfaces at an approach/separation rate of ∼1 nm/s, with the interactions mediated by the adsorbed layers of MFP-3s peptide. The absolute separation distance between the two surfaces was measured with fringes of equal chromatic order (FECO) assuming a two-layer interferometer between the gold and silver layers (49, 50). After multiple force measurements were performed and found to be reproducible with a given amount of injected peptide, an additional picomolar aliquot of MFP-3s peptide was injected between surfaces. The peptides were again allowed to adsorb for 30 min, where additional force measurements were performed. Within a single experiment, this process was repeated until the adhesion between the surfaces either plateaued or began to decrease with greater amounts of injected MFP-3s peptide. The force measurements presented in this work are representative of at least four separate experimental setups for each SAM type.