Effect of calcium ions on peptide adsorption at the aqueous rutile titania (110) interface

The use of titanium-based materials in medical implants is widespread, mainly owing to their biocompatibility, low allergenicity, and excellent corrosion and mechanical properties. Dental implants extensively utilize titanium (Ti) because it rapidly osseointegrates to the surrounding tissues, hence building a shield against bacterial infection.¹ The surfaces of orthopedic and dental titanium implants are functionalized or coated not only to improve implant osseointegration but also to reduce implant loosening and undesired reactions.^2,3 Over the years, a variety of surface modification techniques^4,5 have been proposed to improve the bone–implant adhesion properties of titanium. One approach that has delivered promising results involves the biomimetic modification of implant surfaces, where biologically active molecules are used to coat the surface of implants to enhance cell adhesion and tissue regeneration.^6–9 The specific adsorption of peptides or other biomolecules is a viable alternative to the more expensive covalent immobilization techniques to confer materials with biological functions.^10,11

Design strategies have focused on exploiting implant surfaces presented with integrin ligands such as proteins and short motifs of the extracellular matrix (ECM) to improve implant–cell adhesion, thereby accelerating cell colonization and biointegration.¹² Cells recognize and attach to components of the ECM via integrin receptors that contain an exposed ubiquitous adhesive motif of the sequence Arg–Gly–Asp (RGD). Several experimental studies have demonstrated the effectiveness of RGD-coated titanium implants in enhancing bone growth and healing;^2,13–16 however, the mechanism by which this and other peptides adsorb to titanium surfaces is not yet fully understood. This process is believed to be greatly influenced by inorganic elements of the ECM such as calcium which is heavily involved in the bone repair and remodeling process.^17,18 Calcium ions (⁠ $Ca 2 +$⁠) play critical roles in cell signaling and the activation and adhesion of platelets due to their ability to mediate biomolecule– and cell–material interactions.^19–22 

In addition, calcium (as well as other divalent cations) alters the isoelectric point and $ζ$-potential of titania;^23,24 an inert, thin oxide layer quickly forms on the Ti surface when exposed to air or biological fluids. Although the surface of a titanium implant is negatively charged at physiological environments, the presence of $Ca 2 +$ and other divalent cations at material interfaces has long been suggested²⁵ and later shown^26–29 to facilitate the adsorption of biomolecules such as proteins and DNA onto like-charge surfaces (including titania) via electrostatic shielding. Although possible with monovalent ions, divalent cations, particularly $Ca 2 +$⁠, are more effective in shielding the surface charge. The adsorption of (negatively charged) DNA onto sand particles (i.e., silica) was found to be 100-fold more effective in solutions containing divalent than monovalent cations.³⁰ However, electrostatic screening is but one example, and other, more complex alternative mechanisms have been suggested to explain monovalent versus divalent cation-based differences for silica adsorption.³¹ 

By altering the surface properties of titania,^23,24 however, the coadsorption of charged species such as $Ca 2 +$ adds to the complexity of an already complex^32,33 aqueous $TiO 2$ interface. This poses significant implications for the adsorption of biomolecules onto $TiO 2$⁠, thereby influencing the biocompatibility of Ti implants as demonstrated in several studies.^34–39 The coadsorption of $Ca 2 +$ ions on the $TiO 2$ surface will most certainly influence the adsorption of peptides by altering their conformation at the interface. Given the difficulty with which this can be investigated experimentally in atomistic details, molecular dynamics (MD) simulations provide a viable alternative tool to examine, at the molecular level, the impact of $Ca 2 +$ ions on peptides adsorption at the aqueous $TiO 2$ interface. Toward this end, Wu et al.⁴⁰ evaluated the role of various cations including $Ca 2 +$ in determining the binding propensity of the Asp residue of RGD to the aqueous rutile $TiO 2$ (110) interface using MD simulations. The authors reported that the binding propensity of Asp at the $TiO 2$ interface was strongly influenced by the ionic radii and charge of the mediating cation. Furthermore, divalent cations appear to be more effective than monovalent cations in mediating the interaction between the negatively charged Asp residue and the negatively charged $TiO 2$ surface.⁴⁰ 

In previous work, we used an integrated approach which combined the outcomes from Replica Exchange with Solute Tempering (REST) MD simulations in partnership with metadynamics simulations, along with experimental quartz crystal microbalance measurements of the titania-binding affinity of two experimentally identified 12-mer peptide sequences: Ti-1 (QPYLFATDSLIK) and Ti-2 (GHTHYHAVRTQT). We found good consistency in terms of trends with the experimental binding data and our predicted binding free energies. Moreover, our simulation data enabled a molecular-level elucidation of the structural traits that give rise to these binding affinities. In both the experiments and simulations, we considered a solvent comprising 0.15 M NaCl solution, and therefore, effects from the presence of calcium ions were not explored.

Here, we investigate the impact of $Ca 2 +$ ions on the adsorption of titania-binding peptides at the negatively charged aqueous rutile $TiO 2$ (110) interface using atomistic MD simulations. To enable constructive comparisons with our previous work, we have focused on the Ti-1 and Ti-2 peptide sequences. We have examined the possible effect that $Ca 2 +$ ions have on the binding propensity and peptide conformations of Ti-1 and Ti-2 in the surface-adsorbed state. To our knowledge, these Ca $2 +$ effects on the adsorption of large peptides at the titania/water interface have been relatively under-explored by MD simulation approaches. Our data provide insights into this influence and may help with the goal of designing peptides with tunable affinity to titania, paving the way for the development of novel and reliable surface treatments for titanium implants.

Owing to the typical complexity of biointerface in liquid water, the use of targeted, advanced conformational sampling approaches is essential for capturing the Boltzmann-weighted ensemble of adsorbed conformations.⁴¹ To this end, we used a Hamiltonian-based replica exchange approach, namely, Replica Exchange with Solute Tempering (REST-MD) simulations. Our previous tests indicate that standard MD simulations provide an inferior description in comparison with REST-MD simulations.⁴² We performed two REST-MD simulations,^43,44 one for each peptide sequence, Ti-1 and Ti-2, in the presence of CaCl $2$ solution, adsorbed at the negatively charged rutile TiO $2$ (110) interface. We used gromacs 4.5.5⁴⁵ simulation package throughout. All simulations were carried out in the Canonical (NVT) ensemble at 300 K and were maintained at this temperature via the Nosé–Hoover thermostat,^46,47 with a $τ T$ value of 0.2 ps. The leapfrog algorithm was used to integrate Newton’s equations of motion, with an integration timestep of 1 fs. Coordinates and velocities were saved every 1000 steps. Particle Mesh Ewald⁴⁸ summation was used for long-range electrostatic interactions, with a real-space cutoff of 12 Å. The cutoff for Lennard–Jones interactions was also set at 12 Å.

The orthorhombic simulation cell comprised one peptide chain, a five-layer $TiO 2$ slab (containing 6100 atoms), solvated with 6952 water molecules, and calcium and chloride ions. The lateral dimensions of the cell (and therefore the titania slab) were 60.7  $×$ 57.5 Å , whereas the dimension of the simulation cell in the direction normal to the slab (⁠ $z$-direction) was adjusted to $∼$57.5 Å so as to recover bulk water density (at 300 K and ambient pressure) in the center of the inter-slab space (i.e., between the slab and its periodic image in the $z$-direction).

The negatively charged surface was represented using the “charged nonhydroxylated 12.5% surface” model reported by Prédota et al.⁴⁹ Note that this surface model carries surface hydroxyls, despite the name given for this model. Twenty-five hydroxyl groups were attached at randomly selected Ti surface sites to each side of the $TiO 2$ slabs, giving a 12.5% surface hydroxyl coverage. To elaborate, the 25 OH $−$ groups were positioned randomly on the upper slab surface and their positions were mirrored on the opposite face of the slab with an additional 25 hydroxyls. This hydroxyl coverage corresponds to a surface charge density of $σ$ = $−0.104$ C m $− 2$⁠, which is close to the experimentally measured charge density for this $TiO 2$ surface at room temperature and at near physiological pH.⁴⁹ This random distribution of hydroxyl groups gives the $TiO 2$ interface a pseudo-random structural character. Here, the $TiO 2$ slab atoms were restrained during the simulations, except for the surface hydroxyls which were allowed to move within the constraints of the bonded terms of the force-field developed by Prédota et al.⁴⁹ Incorporation of surface flexibility was previously found to have very little impact on the calculated binding free energy compared to a rigid surface.⁵⁰ To balance the overall $−50e$ charge of the titania slab, 27/26 Ca $+ 2$ and 4/2 Cl $−$ ions were added to liquid water as counterions, for Ti-1/Ti-2, respectively. The peptides and liquid water were described using the CHARMM22* force-field^51,52 and TIPS3P model,^53,54 respectively. We recognize that the model of Prédota et al.⁴⁹ was devised in partnership with the SPC/E water model. However, in earlier work, we adapted and tested this force-field to work with the TIPS3P model,⁵⁵ to enable our use of the CHARMM family of force-fields for this biointerface.

For the REST-MD simulation of each peptide sequence, Ti-1 and Ti-2, 16 replicas were prepared, with each replica containing a unique initial peptide structure constructed manually, covering a range of common folded-backbone secondary structure motifs. Where applicable, following previously published work,^42,56 the solute group was defined as the peptide chain and the subset of counterions required to neutralize the charge of the peptide. In all REST simulations, the thermal temperature of all replicas was set to 300 K. However, each replica featured a unique “effective temperature” associated with scaling of terms in the Hamiltonian, with all 16 replicas spanning the “effective temperature” range of 300–433 K. Following previously published successful protocols,^42,57 the “effective temperature” set for all the replicas was 300.00, 305.35, 310.89, 317.25, 323.88, 331.57, 339.88, 349.11, 358.07, 367.37, 380.98, 389.80, 398.22, 406.85, 419.97, and 433.00 K. The “effective temperature” of 300.00 K corresponded to the unscaled Hamiltonian, which provided the reference trajectory which was subjected to subsequent analysis. Prior to commencing the REST-MD simulations, all replicas were energy-minimized and equilibrated for 0.5 ns at their target Hamiltonian. During the REST-MD simulation runs, an exchange attempt between two adjacent replicas was made every 1 ps. Each REST-MD simulation was run for 15 ns, yielding an aggregate of 0.24  $μ$s simulation time (⁠ $16×15$ ns = 0.24  $μ$s). Consistent with our previously published simulations of peptide-TiO $2$ adsorption in NaCl solution, both peptides were found to be adsorbed to the surface throughout most of the simulation, specifically in more than 90% of the REST reference replica (i.e., unscaled Hamiltonian) trajectory.

We also performed and analyzed REST-MD simulations for both Ti-1 and Ti-2 in the absence of the titania surface, in equivalent concentrations of Ca $2 +$ and Cl $−$ ions, using simulation settings as specified above.

The Boltzmann-weighted ensemble of conformations produced by our REST-MD simulations can be characterized by identifying the key types of conformations that comprise this ensemble. To do this, we performed a clustering analysis of the reference replica trajectory. The 15,000 conformations (one frame generated every 1 ps) obtained from each of these 15 ns REST-MD simulations were grouped into “clusters” of like-structures following the method reported by Daura et al.⁵⁸ This approach compares peptide backbone conformations based on the root mean-squared deviation between their backbone atom positions to identify the number of thermally accessible peptide conformation types. A cutoff distance of 2 Å was found to be appropriate in previous studies of oligopeptide adsorption to aqueous inorganic interfaces^42,56,57,59 and thus was chosen in this work for the clustering analysis. This analysis yields both the total number of distinct backbone conformations in the ensemble (the total number of clusters) and the fractional population of each distinct conformation in the ensemble. Further, we used these data to calculate the conformational entropic contribution for this ensemble, as defined in the study reported by Palafox-Hernandez et al.,⁵⁶ such that

where n is the total number of clusters and $p i$ is the population of the $i$th conformation. We also characterized the relative proportion of secondary structural motifs for both Ti-1 and Ti-2, both in solution and when adsorbed at the aqueous $TiO 2$ interface. These were determined using Ramachandran analyses of the conformations generated from the entire trajectories of the REST-MD simulations.

We characterized the degree of structuring of both the water and ions at the interface in the absence of the peptide. We used in-house codes to calculate the vertical mass density profile of waters and ions as a function of distance from the surface. Furthermore, following our earlier work,⁵⁹ we calculated the degree of residue–surface contact, both for “direct” and for “solvent-mediated” contact modes using trajectories of the REST-MD simulations. Contact between a given residue and the $TiO 2$ surface was defined as “direct” if there were no intervening water molecules between the residue side chain and the surface and “solvent-mediated” if the residue was adsorbed to the first interfacial water layer rather than directly adsorbed to the surface itself.

For the purpose of calculating residue–surface distances, the $TiO 2$ surface was defined as the uppermost plane of the slab’s basal Ti atoms. As explained and justified in previous work,⁵⁹ a residue was considered in direct contact if a reference site assigned to the residue was found within a given cutoff separation (⁠ $d$⁠) from the surface (i.e., $z≤d$⁠) and was defined to be in solvent-mediated contact for $d<z<$(⁠ $d+2$ Å). Residue reference sites and cutoff distances are listed in Table I. For each of the 15,000 frames of each reference replica REST-MD simulation trajectory, we therefore counted if each residue was in “direct” or “solvent-mediated” contact. The fraction of total frames that satisfied either of these criteria was then converted into a percentage.

There is compelling experimental evidence that $Ca 2 +$ ions are deeply involved in many aspects of biomolecular adsorption at the aqueous $TiO 2$ interface, including the mediation of biomolecule–material interactions^{9,19–22,26–28} and manipulation of $TiO 2$ surface properties.^23,24 Despite the critical roles Ca $2 +$ ions play in the realm of bio- $TiO 2$ interactions, and the concomitant impact of $Ca 2 +$ ions on the biocompatibility of titanium implants, only a handful of computational studies,^40,60,61 employing nonadvanced MD simulations, have investigated the effect of $Ca 2 +$ ions on bio- $TiO 2$ interactions at the molecular level. Moreover, these studies investigated the impact of $Ca 2 +$ ions on the adsorption of tripeptide motifs only, and thus knowledge of such impact on the adsorption of larger peptides remains to be explored. Accordingly, elucidating how $Ca 2 +$ ions could impact the adsorption mechanism of dodecamers Ti-1 and Ti-2 to the aqueous $TiO 2$ interface, using advanced simulation techniques, will enhance our currently limited understanding of the mediation of bio- $TiO 2$ interactions via $Ca 2 +$ ions. Such understanding, at the molecular level, bodes well for the development of strategies to manipulate biomolecular interactions at the $TiO 2$ interface and thus control the biological and cellular response to Ti implants.

Based on earlier studies,^32,59,62 before evaluating the surface adsorption of Ti-1 and Ti-2 in $CaCl 2$ solution, it is relevant to investigate any possible impact of $Ca 2 +$ on the structuring of the interfacial solvent. As shown in Fig. 1, the presence of $Ca 2 +$ ions at the interface has negligible influence on interfacial water layering compared with the case of $NaCl$ (Fig. 1). Again, it is noted that the surface is defined here as the plane of uppermost Ti atoms on the slab surface. For the $CaCl 2$ solution, the density of the three interfacial water layers was slightly higher, with small differences in the vertical position (along the z-axis) of the corresponding peaks; the first two water layers were positioned further away from the surface (by 0.02 and 0.08 Å, respectively), whereas the third layer was closer (by 0.04 Å). These marginal differences are expected to be insignificant in terms of the influence on peptide–surface binding. The position and the height of the peaks in the water density profile in NaCl (data taken from Sultan et al.⁵⁹) are in good agreement with previous work on the same surface.^49,55

In addition, while the height of the peaks in our water density profile in $CaCl 2$ is in fair agreement with previous work based on a charged, partially hydroxylated $TiO 2$ aqueous interface,⁶⁰ the position of the peaks shown in Fig. 1 is a little further from the surface (2.8, 3.6, and 5.7 Å) in comparison with previous work (2.2, 3.6, and 4.8 Å).⁶⁰ Furthermore, the position of the peaks of the first two interfacial water layers is in fair agreement with that reported by Wu et al.⁶¹ (⁠ $2.34±0.20$ and $3.76±0.28$ Å) for their study of the adsorption of the RGD tripeptide adsorbed at the charged rutile $TiO 2$ (110) interface. In addition to the difference in peptide size of RGD compared to Ti-1 and Ti-2 dodecamers, Wu et al.⁶¹ used a $TiO 2$ surface model featuring a surface charge density twice as great (⁠ $σ$ = $−0.208$ C m $− 2$⁠) compared with our $TiO 2$ surface model.

As the results presented herein will show, the position of $Ca 2 +$ ions at the interface can have a significant impact on the binding of residue side chains to the surface. We emphasize here that the ions were free to move in our simulations and showed acceptable mobility during our REST-MD simulations. The location and density of $Ca 2 +$ at the interface showed similarities compared with the case of $Na +$ (Fig. 2). Both $Na +$ and $Ca 2 +$ featured four well-defined layers at the interface, with the first two layers being significantly more dense than the third and fourth layers. While the position of the first two peaks from the surface (3.1 and 3.4 Å) were similar for $Na +$ and $Ca 2 +$⁠, the latter showed a more dense second peak compared with $Na +$ (Fig. 2). Furthermore, both ions featured a small peak at 3.9 Å from the surface. However, the fourth peak for $Ca 2 +$ was positioned further the surface (5.3 Å) compared with that for $Na +$ (4.5 Å). This distinction in ion-surface separation distances for Na $+$ and Ca $2 +$ can be rationalized in terms of the inner-sphere and outer-sphere adsorption models.⁶¹ Inner-sphere adsorption corresponds to a case where (in this particular instance) a surface hydroxyl or a first-layer water molecule can substitute for one of the water molecules in the solvation shell of the ion. In contrast, the outer-sphere adsorption mode corresponds to the surface adsorption of an ion via the ion adsorption shell. In particular, the close positioning of the Na $+$ ions corresponds with inner-sphere adsorption, while the more distant location of the Ca $2 +$ in the fourth peak corresponds with outer-sphere adsorption.

The position of the peaks corresponding to the first two $Ca 2 +$ layers in Fig. 2 (3.1 and 3.4 Å) is in very good agreement with data reported by Prédota et al.⁶³ using MD simulation. These authors reported at least two $Ca 2 +$ layers at the negatively charged nonhydroxylated rutile $TiO 2$ (110) aqueous interface (the same $TiO 2$ surface modeled here), with the peaks of these two layers at 3.0 and 3.47 Å from the surface.⁶³ Although the position of these peaks corresponds to the $TiO 2$ surface with a charge density twice as strong (⁠ $σ$ =  $−0.208$ C m $− 2$⁠) compared with our $TiO 2$ surface model, the authors also performed similar simulations at $σ$ =  $−0.104$ C m $− 2$ and found that the position of the peaks changed by only less than 0.05 Å, with no appreciable changes in the long-range structure.⁶³ The position of the first two $Ca 2 +$ peaks in Fig. 2 is also in good agreement with Monti et al. regarding their simulation of the KEK tripeptide at the charged partially hydroxylated rutile $TiO 2$ (110) aqueous interface.⁶⁰ In contrast, in their study of RGD adsorption at the $TiO 2$ interface, Wu et al.⁶¹ reported three $Ca 2 +$ peaks at the interface, with the positions of these peaks (3.5, 4.1, and 5.4 Å) in fair agreement with those observed here, but for the second (3.4 Å), third (3.9 Å), and fourth (5.3 Å) peaks, respectively (Fig. 2). However, unlike the results shown in Fig. 2, the third $Ca 2 +$ peak (5.4 Å) reported by Wu et al.⁶¹ was found to be significantly more dense than the first two peaks located closer to the surface (3.5 and 4.1 Å), which also contradicts the results of previous simulations^60,63 that reported one or two dense peaks within 3.5 Å of the surface. The above results notwithstanding show that the position of $Ca 2 +$ at the negatively charged $TiO 2$ interface modeled here (Fig. 2) is in fair agreement with previous simulations at the charged $TiO 2$ interface.^60,61,63 In closing, we remark here that differences in the solvation model (SPE/E versus TIPS3P), ion potentials, degree of hydroxylation, and surface charge density can all contribute to differing vertical density profiles for adsorbed ions.

REST-MD simulations of Ti-1 and Ti-2 in $CaCl 2$⁠, in the presence and absence of the aqueous $TiO 2$ interface, were carried out to study how $Ca 2 +$ ions affect peptide structure at the interface. In each case, the conformational ensemble of the peptide was evaluated via a clustering analysis, as detailed in Sec. II. The total number of clusters resulting from this analysis indicates the number of thermally accessible structures for each peptide. A high number of clusters indicates a greater degree of structural disorder relative to a low number of clusters. The presence of $Ca 2 +$ ions appeared to have a far greater impact on the structure of Ti-1 than Ti-2, particularly when the Ti-1 peptide was adsorbed at the $TiO 2$ interface. The total number of clusters for Ti-1 and Ti-2 in $CaCl 2$⁠, when adsorbed at the aqueous $TiO 2$ interface (130 and 81, respectively) and when free in solution (175 and 157, respectively), was reduced compared to those corresponding to the same simulations in NaCl⁵⁹ (182 and 113 at the interface and 270 and 200 in solution). This suggests that these two peptides were less disordered in $CaCl 2$ than in NaCl. This fact notwithstanding, the total number of clusters for both peptides in $CaCl 2$ (Table II) was comparable to other materials-binding peptides.^42,56 Furthermore, similar to the case of NaCl,⁵⁹ Ti-1 featured more clusters than Ti-2 (i.e., a greater number of thermally accessible conformations) in $CaCl 2$⁠, both at the interface and in solution. This suggests that similar to the results in NaCl,⁵⁹ Ti-1 may be more disordered than Ti-2 in $CaCl 2$ solution. In addition, in $CaCl 2$ solution both peptides unsurprisingly featured a greater number of clusters in solution than at the interface, demonstrating adsorption-induced limitations on the range of conformations available to both these peptides. Furthermore, the difference between the total number of clusters at the interface and that in solution was greater for Ti-2 than for Ti-1, suggesting that, at the interface, the latter peptide experienced less conformational restriction upon adsorption.

The population of the top ten most populated clusters of both peptides in $CaCl 2$ and $NaCl$⁠, both when adsorbed at the interface and when free in solution, is shown in Table II. Similar to the results reported for NaCl,⁵⁹ the population of the top ten clusters in $CaCl 2$ did not indicate a dominant conformation for either peptide when at the interface or when in solution. However, both Ti-1 and Ti-2 showed similar populations for the top two clusters in $CaCl 2$⁠, particularly at the interface where they accounted for $∼$45% of the total ensemble (Table II). This differed from what was observed for both peptides in NaCl, where the sum of the population of the top two clusters for Ti-1 and Ti-2 at the interface was $∼$22% and $∼$36%, respectively (Table II). This difference in cluster populations for both peptides between the $CaCl 2$ and $NaCl$ cases indicates that, at the interface, $Ca 2 +$ ions had a greater impact on the ensemble of Ti-1 than on that of Ti-2. However, the population of the top two clusters in $CaCl 2$⁠, together with the smaller number of total clusters compared with that in NaCl (Table II), suggests that $Ca 2 +$ ions have restricted the structural diversity of the conformational ensemble for both peptides at the interface. The population of the top clusters in $CaCl 2$ at the interface is comparable to those reported for gold-binding peptide AuBP2 and silver-binding peptide AgBP2 at the aqueous Au interface in water.^42,56

We also performed a cross-cluster comparison to check if the most populated types of peptide backbone conformation were shared for both Ti-1 and Ti-2 in the presence of $Ca 2 +$ ions (and compared to their conformations in NaCl). These data are summarized in Table III. Similar to what was noted for the two peptides in NaCl,⁵⁹ Ti-1 showed fewer matches in $CaCl 2$ between surface-adsorbed and in-solution clusters compared with Ti-2. Indeed, only one match was found (using the 2 Å RMSD cutoff) for Ti-1 which, in this case, involved the least populated of the top ten clusters identified in the surface-adsorbed state. The same comparison for Ti-2 revealed three matches including the second and third most populated clusters of the surface-adsorbed conformations (Table III). This analysis suggested that in $CaCl 2$⁠, the most populated cluster for either peptide adsorbed at the interface was distinct from any clusters of the same peptide when free in solution; note that matches with low-populated (i.e., beyond top ten) clusters in solution are regarded as insignificant due to their relatively minor populations. Overall, however, the fact that Ti-1 and Ti-2 showed fewer matched clusters between the “in solution” and surface-adsorbed states in $CaCl 2$ (Table III) compared with NaCl,⁵⁹ together with a smaller number of total clusters in $CaCl 2$ than was predicted in NaCl,⁵⁹ indicates that both peptides featured more unique conformations in the presence of $Ca 2 +$ than $Na +$ ions. Representative structures of Ti-1 and Ti-2 “in solution” and adsorbed at the interface (both in $CaCl 2$⁠) are shown in Figs. 3 and 4, respectively.

One possible way the presence of $Ca 2 +$ could impact the adsorption of both peptides to the surface is by influencing peptide structure. To assess this possible impact, Ramachandran analysis of secondary structure motifs of both peptides at the interface and in solution was performed. The results revealed only insignificant differences in the population of peptide secondary structure motifs for both peptides as a result of adsorption, as illustrated in Table IV. In addition, the variation in the population of peptide secondary structure motifs in $CaCl 2$ compared with NaCl was greater for Ti-1 than Ti-2 (Table IV). This suggests that the influence of $Ca 2 +$ ions on peptide structure was more significant for Ti-1, which featured a slightly larger population for random coil in NaCl than $CaCl 2$ in the adsorbed state.

This Ramachandran analysis, however, did not reveal the similarities and differences between peptide structures identified in $NaCl$ and those found in $CaCl 2$⁠. To gain such information, a cross-cluster comparison between the ten most populated adsorbed and in-solution clusters for Ti-1 and Ti-2 identified in $NaCl$⁵⁹ and $CaCl 2$ was performed (Table V). Compared with Ti-1, the Ti-2 peptide showed more matches between clusters obtained for the NaCl and $CaCl 2$ solution conditions for both the surface-adsorbed and in-solution states. In addition, the most populated “in-solution” cluster for both peptides in $CaCl 2$ solution (Fig. 3) was found to be unique, matching none of the corresponding clusters in NaCl solution. At the interface, however, the top three Ti-2 clusters in $CaCl 2$ were found to match with top ten Ti-2 clusters in NaCl (Table V), indicating that $Ca 2 +$ ions did not alter the structure of the Ti-2 significantly, which is consistent with the Ramachandran analysis results (representative structures of the absorbed peptides are shown in Fig. 4). Ti-1, on the other hand, showed no matches between top ten clusters of the adsorbed peptide in $CaCl 2$ and NaCl. This further suggests that $Ca 2 +$ ions exert a stronger influence on the structures of Ti-1, particularly when adsorbed at the $TiO 2$ interface.

The above analysis notwithstanding, a large number of distinct peptide clusters is not the sole indicator to a high degree of peptide conformational disorder; a materials-binding peptide may, for example, feature over 100 clusters in total but might have a population of, e.g., 70% for the most populated cluster, indicating the presence of a well-defined conformation. To probe this, the conformational entropic contribution, $S conf$⁠, of both Ti-1 and Ti-2 was calculated (described in Sec. II). Indeed, the presence of $Ca 2 +$ ions was found to reduce the conformational entropic contribution for both peptides at the interface and in solution compared with the case of $Na +$ ions (Table VI). Not surprisingly, both Ti-1 and Ti-2 were found to lose conformational entropic contribution upon adsorption to the surface. However, this loss was greater for Ti-1 in $CaCl 2$ (0.9) compared with that in NaCl (0.75). Also, upon adsorption in $CaCl 2$⁠, the reduction in the conformational entropic contribution for Ti-2 in $CaCl 2$ was less than that in NaCl (1.04 versus 1.2). Overall, the conformational entropic contributions for Ti-1 and Ti-2 in $CaCl 2$ were similar in magnitude to those reported for gold-, silver-, and graphene-binding peptides in water.^56,64 In summary, the clustering and secondary structure analyses indicate that $Ca 2 +$ ions exerted a greater impact on the structure of Ti-1 than Ti-2, particularly when the peptide was adsorbed at the $TiO 2$ interface.

We next investigated the degree of peptide–surface contact for Ti-1 and Ti-2. We found that the presence of $Ca 2 +$ ions at the $TiO 2$ interface exerted a significant influence in the direct contact and solvent/ion-mediated contact modes, particularly for Ti-1 residues. We evaluated this by calculating the degree of surface contact for each residue (and each terminus) of both Ti-1 and Ti-2, illustrated in Fig. 5. These $CaCl 2$ data show substantial differences with the corresponding NaCl contact data⁵⁹ in terms of direct contact. The direct contact of the C-terminus of both peptides in $CaCl 2$ was found to be small (less than 10%), in contrast to what was observed in $NaCl$ (27% for Ti-1 and 44% for Ti-2). The N-terminus of both peptides, however, maintained considerable direct contact with the surface in $CaCl 2$ (28% for Ti-1 and 59% for Ti-2), although this degree of contact was lower for Ti-2 in $CaCl 2$ than in $NaCl$ (82%).

Interestingly, the contact between K12 in Ti-1 and the surface increased significantly in the presence of $Ca 2 +$ ions (80%). However, this interaction was not mediated by $Ca 2 +$ ions, which suggests that the increase in the propensity of direct K12–surface contact is a nonlocal effect. K12 was found to coordinate with the $TiO 2$ surface via its ammonium group, as shown in Fig. 6. This reflects the strong binding of lysine analogs at the negatively charged aqueous titania interface⁶⁵ and previously reported simulations of the RKLPDA hexapeptide on amorphous titania.³³ In addition to K12, Ti-1 featured two other contact sites, S9 and the N-terminus which, respectively, showed a small (15%) and an appreciable (28%) degree of direct contact in $CaCl 2$ (Fig. 5). The direct S9–surface contact was due to the coordination between the hydrogen of the Ser hydroxyl group and a bridging oxygen on the surface or the oxygen of a surface hydroxyl [Fig. 6(b)]. While the binding affinity of the analog of Ser to the $TiO 2$ surface was predicted to be weak on its own, we propose that a sufficiently high number of polar, uncharged residues such as Ser, may result in a collectively strong peptide– $TiO 2$ affinity. This suggestion is supported by the fact that many $TiO 2$-binding peptide sequences identified via biocombinatorial techniques were found to be rich in Ser residues.^66,67

The negatively charged Asp residue (D8) in Ti-1 was found to adsorb to the negatively charged $TiO 2$ interface via either mono- or bidentate bridging coordination with a $Ca 2 +$ ion adsorbed directly at the surface. Despite the predicted favorable affinity for Asp analog (methanoate ion) to the negatively charged $TiO 2$ surface (where the molecule was found to adsorb directly to surface-exposed positively charged Ti sites⁶⁵), D8 was not found to be in direct contact with the surface in NaCl solution,⁵⁹ which was thought to be due to its proximity to a substantial hydrophobic contact. Inspection of all frames corresponding to the case where D8 was located within 5–6.6 Å from the surface revealed that in this distance range, D8 was in $Ca 2 +$-mediated contact with the surface, where this residue coordinated to a $Ca 2 +$⁠, located in the second or third (Fig. 7) interfacial water layer, via the carboxylate group of Asp chain. This $Ca 2 +$-mediated Asp– $TiO 2$ interaction is in agreement with observations reported for the interaction of Glu (in the KEK tripeptide)⁶⁰ and Asp (in the RGD tripeptide)^40,61 at the negatively charged aqueous rutile $TiO 2$ (110) interface.

We found two dominant configurations. The first was a predominantly monodentate Asp– $Ca 2 +$ coordination, where one oxygen of the carboxyl group of Asp coordinated with a $Ca 2 +$ ion directly adsorbed to the surface [Fig. 7(a)]. Calcium ions adsorbed directly to the surface were found to bind predominantly in the region between surface bridging oxygen “rails” (along the [001] direction) and surface hydroxyls when in proximity to the latter, or atop the bridging oxygen rails, displacing water molecules in the first interfacial water layer (shown in Fig. 8). The second was a predominantly bidentate D8– $Ca 2 +$ coordination where both oxygens of the carboxylate group of Asp interacted with a $Ca 2 +$ ion located in the second or third interfacial water layer [Fig. 7(b)]. Both mono- and bidentate Asp– $Ca 2 +$– $TiO 2$ coordination have been reported by Monti et al.⁶⁰ and Wu et al.^40,61 for the adsorption of the KEK and RGD tripeptides, respectively, to the charged aqueous rutile $TiO 2$ (110) interface. In particular, the findings of Monti et al.⁶⁰ suggested that the adsorption of $COO −$ groups to the charged $TiO 2$ surface can occur via $Ca 2 +$ ions acting as a bridge between the peptide and the surface.⁶⁰ 

Furthermore, in all peptide configurations involving an Asp– $Ca 2 +$– $TiO 2$ interaction, the peptide’s other carboxyl group (i.e., the C-terminus) was found in a bidentate coordination with a $Ca 2 +$ ion that was adsorbed to the interfacial water layers or located further from the surface. The greater propensity of the C-terminus to remain distant (and yet still bind via $Ca 2 +$⁠) from the surface in $CaCl 2$ compared with NaCl could be due to the greater K12–surface direct contact in the presence of $Ca 2 +$ ions, which is possibly due to the considerable $Ca 2 +$-mediated adsorption of D8 to the $TiO 2$ surface. However, the correlation between the degree of surface-contact of D8 and that of other residues is not captured in the analysis of individual residue–surface contact (Fig. 5). To capture this, we calculated conditional probabilities of residue binding and found a 49% propensity for D8 and K12 to both adsorb at the same time to the surface (via $Ca 2 +$-mediated and direct contact, respectively). Indeed, our analysis further showed that when D8 was adsorbed, K12 was also adsorbed in 99% of those frames. However, when D8 was not in contact with the surface, K12 only had a 32% propensity to adsorb directly to the surface. This propensity is only slightly different from that calculated for direct K12–surface contact in NaCl (40%). While these data are not necessarily definitive evidence that the adsorption of one residue (D8 or K12) to the surface drives the adsorption of the other, the results clearly demonstrate that the contact of these two residues to the $TiO 2$ surface in $CaCl 2$ is correlated.

However, for Ti-2, the number of contact residues/points with at least a small (10%) direct contact propensity in $CaCl 2$ (Fig. 5) was fewer compared with NaCl (Fig. 5). In $CaCl 2$⁠, Ti-2 was found to bind directly to the surface mostly via G1, H4, and R9, where the considerable (58%) degree of direct G1–surface contact could be due to the nearby N-terminus. All three of these contact points (N-terminus, H4, and R9) featured a positively charged group, and thus their marked propensity to directly adsorb to the surface was expected given the favorable binding predicted between positively charged molecules and the negatively charged $TiO 2$ surface.⁶⁵ Not surprisingly, all three were found to predominantly adsorb via coordination to the bridging oxygen rails on the surface. As was found for Ti-2 adsorption in NaCl,⁵⁹ R9 pinned the peptide to the surface through a bidentate coordination to two neighboring bridging oxygen atoms (Fig. 4), consistent with the binding mode observed for the Arg analog.⁶⁵ Furthermore, H4, which was the only His in the Ti-2 sequence that was modeled in the protonated form (i.e., H2 and H6 were modeled as charge-neutral), showed the lowest surface-contact propensity, likely due to its proximity in the sequence to hydrophobic residues (H2–T3 and Y5–V8). These hydrophobic residues were expected to result in repulsive interaction with the surface (Fig. 5) as previously predicted.⁶⁵ An in-depth exploration of the impact of His protonation on Ti-2 adsorption to the aqueous $TiO 2$ interface will be the subject of a future investigation.

Solvent-mediated contact between Ti-2 residues and the surface was found to be greater in $CaCl 2$ compared with NaCl, with all but three residues (R9, T10, and T12) showing at least a small (greater than 10%) propensity to adsorb to the surface via the first interfacial water layer in $CaCl 2$ (Fig. 5). In their investigation of the adsorption of the tripeptide RGD to the aqueous $TiO 2$ interface using MD simulations, Wu et al.^40,61,68 reported that cations such as $Na +$ and $Ca 2 +$ compete with peptide side chains for adsorption sites on the $TiO 2$ surface. Not surprisingly, such competition should be stronger for divalent cations such as $Ca 2 +$ than monovalent cations such as $Na +$⁠.^40,61 If such competition exists between cations and Ti-2 residues, this may explain the more extensive solvent-mediated contact between Ti-2 and the surface in $CaCl 2$ solution compared with that in NaCl solution. In contrast, for Ti-1, the most significant solvent/ion-mediated contact with the surface (Fig. 5) was due to D8 (50%) and the C-terminus (41%) which, as argued earlier, appeared to be $Ca 2 +$-mediated adsorption.

As articulated in previous work, there are limitations to studying the bio- $TiO 2$ interface using a pseudocrystalline $TiO 2$ surface model, because, in reality, the $TiO 2$ surface is expected to be amorphous or semicrystalline. However, our understanding of the phenomena governing the complex biomaterial interface is limited, and thus the crystalline $TiO 2$ surface model provides a necessary stepping stone to further advance our knowledge in this field. Because of the differences between crystalline and noncrystalline surfaces, we are mindful of the discrepancies that may arise when comparing experimental data with molecular simulation results. Furthermore, while the great majority of simulations of the peptide–material interface, including the work presented here, model the adsorption of a single peptide chain, under experimental conditions the interface may contain the presence of laterally interacting adsorbed peptides. Therefore, some differences between simulation results and experimental data are expected, because in the experiment, the adsorption affinity/behavior may be influenced by inter-peptide interactions at the interface. Such possible influences have been explored using multipeptide interfacial REST-MD simulations for Au nanoparticles⁶⁹ and graphene surfaces.⁷⁰ Such multichain REST-MD simulations require extensive computational resources, and it remains for future work to determine if this is an influential factor for titania interfaces. It must also be noted that, while the $CaCl 2$ simulations presented in this chapter utilized the standard CHARMM22* parameters to describe $COO −$– $Ca 2 +$ interactions, recent evidence⁷¹ indicates that the CHARMM22* parameters slightly over-estimate the strength of this interaction. Church et al. used a close comparison of free energy calculations and experimental data to re-fit this interaction within the CHARMM22* framework.⁷¹ We recommend that future simulations of biomolecular adsorption involving $Ca 2 +$ take into account these revised parameters.

Our MD simulations suggest that the presence of $Ca 2 +$ has an impact on peptide structure and peptide–surface contact at the aqueous rutile TiO $2$ (110) interface. This was explored using two experimentally determined Ti-binding peptides: Ti-1 (QPWLFATDSLIK) and Ti-2 (GHTHYHAVRTQT). In $CaCl 2$⁠, both peptides supported a less diverse ensemble of conformations and fewer similarities between adsorbed and in-solution conformations were predicted, compared with counterpart data obtained for NaCl solution.⁵⁹ The presence of $Ca 2 +$ ions was found to enhance the binding of Asp8 in Ti-1 to the surface by acting as a bridge between the negatively charged residue and the negatively charged surface. Also, this enhanced surface-contact propensity of Asp8 was found to be correlated to an increased surface contact of Lys12. In contrast, Ti-2 featured fewer direct contact residues in $CaCl 2$ solution compared with NaCl solution but maintained a strong degree of direct surface contact via Arg9 and the N-terminus. Ti-2 instead featured a more extensive degree of solvent-mediated contact, involving most residues and the first interfacial water layer. Our findings can offer constructive insights into the manipulation of peptide– $TiO 2$ interactions at the molecular level, amenable to utilization in biomedical applications.

This material is based upon work supported by the Air Office of Scientific Research, grant number FA9550-12-1-0226. We gratefully acknowledge the Victorian Life Sciences Computation Facility (VLSCI) and the National Computational Infrastructure (NCI) for provision of computational resources. AMS thanks Deakin University for a DUPR scholarship.

Tiffany R. Walsh earned her Ph.D. degree in theoretical chemistry from the University of Cambridge, UK, as a Cambridge Commonwealth Trust scholar, after graduating with a B.Sci (Hons) from the University of Melbourne. Following a Glasstone Fellowship in the Department of Materials at the University of Oxford in the UK, she joined the faculty of the University of Warwick in the Department of Chemistry and the Centre for Scientific Computing. In 2012, she returned to Australia to the Institute for Frontier Materials at Deakin University, Geelong, where she is currently Professor of Bio/Nanotechnology.

Her research interests focus on computational materials science, chiefly modeling interfaces between soft matter and solid surfaces, including nanoparticles, using both first-principles calculations and molecular mechanics-based simulations. In particular, she has pioneered advanced molecular simulation approaches and novel analyses for predicting structure/property relationships of biointerfaces; these approaches have since been adopted by numerous research groups worldwide. She is also the co-creator of atomistic force-fields for describing biointerfaces with gold, silver, and graphene substrates.

Reflections: Tread your own path in your early career. It is not necessarily an easy journey compared with the potential protection and privilege that comes with being a junior partner of a big-name senior investigator. However, you will have freedom to establish and consolidate your own area of expertise from the get-go, you might end up more resilient and resourceful for it, and you won’t have to worry about stepping out from underneath someone else’s huge shadow.

Anas M. Sultan completed a B.Eng. (Hons) degree in Biochemical-Biotechnology at the International Islamic University Malaysia. He then earned an M.Sc. (Biotechnology Engineering) from the same university, researching the use of molecular dynamics simulations to improve the thermal stability of phytases, under the supervision of Prof Ibrahim Noorbatcha. In 2017, he was awarded a Ph.D. from Deakin University for his work with Prof Tiffany Walsh at the Institute for Frontier Materials. His research utilized molecular dynamics simulations and first-principles calculations to investigate biomolecular adsorption at the aqueous interface of medical implant materials, including titania and PEDOT.

Zak E. Hughes is currently a lecturer in theoretical and computational chemistry at the University of Bradford, UK. He obtained his MChem and PhD from Durham University, UK, under the supervision of Prof. Mark Wilson. After completing his Ph.D., he spent time at Curtin University, Australia, and Deakin University, Australia. His research focuses on the molecular simulation of soft matter and interfacial systems as well as the development/refinement of force-fields for such simulations.