We present the results of modelling studies aimed at the understanding of the interaction of a 7 nm sized water droplet containing a negatively charged globular protein with flat silica surfaces. We show how the droplet interaction with the surface depends on the electrostatic surface charge, and that adhesion of the droplet occurs when the surface is negatively charged as well. The key role of water and of the charge-balancing counter ions in mediating the surface-protein adhesion is highlighted. The relevance of the present results with respect to the production of bioinorganic hybrids via encapsulation of proteins inside mesoporous silica materials is discussed.
The stabilization of biologically active molecules under conditions different from the physiological ones can be considered a key issue in modern material science . In this scenario, protein immobilization onto inorganic inert matrices is an effective route for the formation of new composite systems characterized by catalytic efficiency and selectivity comparable with that of in vivo enzymatic systems . The realization of such hybrid materials is by no means a trivial task because strong host–guest interactions may induce protein dehydration and denaturation, thus leading to inert adducts. On the other hand, progress in the synthesis of mesoporous silica matrices , characterized by ordered pore arrays, high surface areas, controlled morphology, tunable surface polarity and chemical stability, has made such materials highly attractive as media for the encapsulation, immobilization and confinement of biopolymers . Moreover, owing to the nanometric scale of the pores dimension (2–50 nm), encapsulation may reduce protein aggregation, thus enhancing the enzyme stability . Also, owing to their biocompatibility, ordered mesoporous silica-based composites can be exploited as drugs carriers, bioactive agents and in tissue engineering for different nanomedicine applications [1–4].
Hybrid fabrication protocols are time consuming, and post-encapsulation procedures that reduce the pore openings, such as grafting of organosilane molecules, are commonly employed in order to avoid enzyme leaching, which would be detrimental for the composite activity . The presence of water in the hybrids is essential for maintaining the biopolymer's activity; moreover, a proper pH value is mandatory as well because, in general, maximum activity is attained in a specific pH range. Since the protonation state of protein residues is a function of pH, the sign of the total electrostatic charge of the biomolecule is pH dependent as well. Furthermore, also the total charge of the inner surfaces in mesoporous silica strongly depends on pH: the isoelectric point (pI) of the popular Santa Barbara amorphous (SBA-15) material is reported to be in the 2–3 range , and, within such an interval, the SBA-15 surface charge may change sign as a function of pH. Rather surprisingly, experience has shown that maximum protein loading in mesoporous SBA-15 is achieved when both the protein and the SBA-15 matrix have total charge of the same sign. For example, maximum loading of the globular protein pepsin (pI=1) occurs at pH 3.6, when both the protein and the matrix are negatively charged,  while lysozyme (pI=11) is optimally adsorbed inside chemically functionalized SBA-15 at pH values where both lysozyme and modified SBA-15 are positively charged . In this latter case, experiments indicate that the properties of the lysozyme–SBA-15 composite may depend on the nature of the charge-balancing ions .
Computational approaches  can be of overwhelming relevance for the development of hybrid materials. Actually, the investigation of protein–solid surfaces interactions via simulations, in spite of being still in its pioneering phases, has been of great help in unravelling many aspects of the general problem , especially concerning interactions with flat surfaces. For instance, lysozyme on hydrophylic silica and graphite surfaces [9,10], and albumin subdomains on chrysotile and graphite [11,12] were investigated, evidencing conformational alterations and/or partial unfolding of the protein upon adhesion. Owing to the interest in the field, a growing number of simulation studies about the surface adsorption of proteins has been performed in the last few years [13,14]. Moreover, an accurate picture of the interaction with silica surfaces can be obtained by considering amino acid molecules, for which first principle methods can be employed. Both ab initio and experimental investigations have been performed, indicating that amino acids interact with silica surfaces [15–17]. In addition, the confinement of organic molecules in crystalline nanoporous systems has been investigated by density functional theory approaches, thus enabling important features of the relationships between stability of composite materials and host–guest interactions to be uncovered [18–20]. Nevertheless, determining the leading interactions in biopolymers/mesoporous silica hybrids is still a challenging task for modelling approaches. Besides the intrinsic complexity of these systems, difficulties stem from the fact that biopolymer dimensions are often very close to the inner pore diameters.
Let us consider the case of pepsin encapsulation in SBA-15. Pepsin, a single-chain enzyme consisting of 326 residues (M=34 623 kDa), is characterized by two topologically similar domains forming a bi-lobe structure [21–23], and by a secondary structure comprising both α-helices and β-sheets. The catalytic site, formed by two aspartate (Asp) residues, is located at the junction of the two lobes and close to a flexible loop region (residues 70–82), the ‘flap’, which is supposed to favour positioning of the substrate during catalysis. The enzyme has optimal activity in the 1–4 pH range, minimal activity above pH 5 and is denatured at pH 6–7 [24,25]. Molecular dynamics simulations on substrate-free pepsin in a pH=3.6 aqueous solution indicated stability of the enzyme in the medium adopted for the preparation of the pepsin–SBA-15 composite: the active site structure is preserved, relevant residues maintain relative positioning consistent with crystallographic data, and only small fluctuations characterize the pepsin surface charge distribution . Moreover, at the hybrid's fabrication conditions, the average pepsin dimensions amount to approximately 4.5×5.0×6.6 nm [5,26]. As far as the inorganic matrix is concerned, diffraction studies established that the cylindrical pores of the adopted SBA-15 samples have approximately 7 nm diameter . These data indicate that the motion of the biopolymer inside the matrix can be hindered by geometrical confinement, and that the centre of mass of the protein cannot be at a distance larger than approximately 3.5 nm from any atom of the cylindrically shaped pore surface. This very fact might hinder the understanding of the efficacious interactions because energetic effects overlap with geometric confinement effects and disentangling the role of the different contributions may become an extremely difficult problem. Moreover, one should consider that the catalytic activity of pepsin–SBA-15 adducts is practically unaffected with respect to the physiological conditions, indicating that encapsulation has not altered the protein structure–functionality relationship . Such a finding suggests that protein–silica interactions should not be strong; otherwise, protein denaturation would have been detected.
This work is aimed at the understanding of the main interactions leading to stable hybrids. To accomplish this goal, we designed a model system free from the confining effects that would be induced by the presence of a real pore model, and studied how solvated pepsin behaves when in contact with silica surfaces with polarity close to typical SBA-15 values. Our results reveal that the efficacious interactions between the hydrated enzyme and the inorganic matrix are weak, governed by the silica surface polarity and strongly affected by the nature of the co-solvating ions.
It is well known that a nanosized water system in the vacuum assumes a spherical shape (a droplet) unless strong water–interface interactions are present, and that this morphology is maintained, even with weakly interacting surfaces . Indeed, silica surfaces with moderate polarity (low silanol concentration) are considered hydrophobic because wetting does not occur and the droplet morphology is lost only at very high surface polarity. Here, we exploit such a property to study the interaction of a hydrated protein in contact with flat silica surfaces characterized by different total electrostatic charge. It should be noticed that the circular section of the water droplet is similar to the cylindrical section of the SBA-15 pores. Therefore, by modelling pepsin inside an approximately 7 nm diameter ‘droplet’, the protein is embedded in an environment that can be considered geometrically very close to that experienced by the hydrated protein in the SBA-15 pores . Overall, the simulation system adopted here enables one to establish whether there is effective attraction between a hydrated protein and a silica surface without imposing the geometrical constraint of the pore structure. These features make our approach a promising general strategy for the investigation of weak host–guest interactions in protein–silica hybrid composites. In the specific case of pepsin, use of this approach evidenced that the negatively charged biopolymer effectively interacts only with negatively charged silica surfaces. Such an unexpected interaction is mediated by the solution (water and co-solvating cations) and may be tuned by varying the nature of the cation.
2. Models and methods
Molecular dynamics studies were performed on model systems composed of a hydroxylated silica surface and a water droplet containing the globular protein pepsin. The starting configurations of the enzyme were based on the 4PEP crystallographic structure of porcine pepsin at 1.8 Å resolution available from the protein data bank . The total electrostatic charge of pepsin, −6|e|, was determined previously by calculating the pKa of the protein side chains at pH=3.6 , where the protein loading in SBA-15 is maximal. In particular, His53, Lys319, Arg307 and Arg315 were modelled as positively charged residues, whereas 10 Asp residues (11, 87, 118, 138, 171, 195, 215, 242, 290, 314) were unprotonated (negatively charged).
Two model silica surfaces were investigated, one electrically neutral, and one with surface charge of −4|e| (figure 1). Electroneutrality of the systems was obtained by adding K+ or Na+ cations (six cations for the neutral surface models, and 10 cations in the case of the negatively charged surface models). The cation concentration for the two model systems amounts to approximately 4×10−2 M and approximately 7×10−2 M, respectively, in line with available data [5,6].
The co-solvating ions in the solution adopted for the preparation of the pepsin–SBA-15 hybrids were potassium cations (a potassium acetate buffered solution was used) . A flat slab of amorphous-hydrated silica with formal stoichiometry [Si1600O3200⋅(H2O)200] and size 10.04×10.04 nm was modelled [28–30]. Surface water was assumed to be chemisorbed and modelled as surface Si–OH groups; the actual stoichiometry of the model surface is therefore [Si1600O3000⋅(OH)400]. The surface OH groups were treated within the united atom approximation. A refined surface model that takes the silanol protons explicitly into account is currently in progress. From preliminary data, we can anticipate that the main results presented here are not affected by such a model upgrading.
In the simulations presented herein, only one side of the silica slab was hydroxylated. The average slab thickness was 1 nm, and the silanol group concentration on the hydroxylated surface amounted to 3.97 OH nm−2, a value typical of mesoporous silica systems [5,28–30]. Periodic boundary conditions were applied in three dimensions on a simulation box of 10.04×10.04×10.04 nm. The simulation cell consisted of a total of 30 649 atoms (including 6974 water molecules) for the neutral surface systems and of 30 641 atoms (6970 water molecules) for the charged surface models (figure 2). Such a choice of the number of water molecules enables one to build up an approximately 7 nm diameter droplet. The silica slab thickness, along with a cubic box size of approximately 10 nm, allows for a 9 nm vacuum: therefore, depending on the actual interaction, the pepsin-containing water droplet is free to adhere or not to the silica surface. The simulations were performed with the NAMD program package . Pepsin atoms and the charge-balancing counter ions (Na+ or K+) were described by the CHARMM22 force field , the rigid TIP3P model was used for water , and long-range interactions were described by using a particle-mesh Ewald summation scheme . The amorphous silica slab was built on the basis of an earlier atomic model [28–30] and the CVFF_AUG force field parameters were adopted for the non-bonded interactions of the Si and O slab atoms , which were kept fixed. The charges of the silica surface atoms were +2.4|e| for Si, −1.2|e| for O and −0.6|e| for the united atom OH groups. The charge of the deprotonated surface oxygen atoms was −1.6|e|. The four negatively charged silanol groups (silanolates) were arranged in a square grid with O−⋯O− separation of approximately 2 nm. A picture of the silanol distribution in the adopted slabs, evidencing the positioning of the silanolates, is shown in figure 1. The resulting surface charge density amounts to −6.4 10−7 μC cm−2 and is compatible with a pH of 3.6, at which maximum loading of the enzyme in SBA-15 was experimentally detected . The simulations were performed in the constant number of particles N, volume V and temperature T (NVT) ensemble at 300 K with an integration time step of 1 fs. Equilibration runs of 2 ns were performed, followed by production runs of 2.5 ns. Equilibration was checked by monitoring protein dimensions, interatomic distances between selected residues and distances between residues and the slab atoms; small standard deviations were obtained in all cases. As surface-induced protein relaxation processes can be considered converged on a time scale of 2–3 ns , the length of the reported trajectories is appropriate for the investigation of the silica-solvated enzyme interactions.
The simulations with the neutral surface, independently of the nature of the charge-balancing counter ions (Na+ or K+), provided very similar results. Along both simulations, the negatively charged protein and the positive counter ions stay immersed in the water droplet and can therefore be considered fully solvated. The droplet gets closer to the hydroxylated side of the silica slab, which is the most polar face of the matrix, however no stable interaction between the droplet and the surface was detected. Indeed, the water droplet is ‘floating’ over the polar side of the slab, as clearly evidenced in figure 3.
Relevant radial distribution functions (RDFs), reported in figure 4, confirm that adhesion of the solvated enzyme does not occur on neutral silica surfaces: The RDFs for the surface silanol oxygen–water oxygen pairs, shown in figure 4a, indicate a very weak interaction independently of the nature of the co-solvating cation. In addition, the shape of the RDFs curve for the protein atoms and the surface silanol oxygen atoms (figure 4b) indicates that there is no interaction between surface and protein. Interestingly, at least four water shells surrounding the protein are detectable in the protein atom–water oxygen atom RDFs (figure 4c), thus demonstrating that, in these conditions, the protein is fully solvated independently of the type of counterion. Moreover, in both cases, the cations remain inside the droplet, and, as shown in figure 4d, they do not interact with the neutral silica surface.
Such a picture drastically changes in passing to the charged silica surfaces, where adhesion of the protein-containing droplet occurs and the interaction depends on the co-solvating cations. In particular, the trajectory snapshots reported in figure 5 reveal that the droplet adheres to the surface region in which the negatively charged silanolate oxygens are located. As evidenced by the RDFs in figure 6a, water molecules are at hydrogen bond distances from the deprotonated silanols with both K+ and Na+ cations. Interestingly, the cation–silanolate oxygen RDFs, represented in figure 6d, highlight a qualitatively different behaviour for the two co-solvating cations: K+ are coordinated to the silanolates, while Na+ ions are not. In particular, three out of four silanolates are steadily coordinated by K+. The protein atom–water oxygen atom RDFs (figure 6c), are nearly indistinguishable from the corresponding RDFs calculated for the neutral slab case, indicating that the protein keeps on being fully solvated, even when interacting with charged surfaces. On the other hand, the RDFs calculated for the protein atom–surface oxygen pairs (figure 6b) display remarkable differences with respect to those obtained for the uncharged surface models. In particular, the RDFs involving the silanolates show a broad feature in the 2–4 nm range, which proves the existence of efficacious attractive interactions with the surface. As indicated by the peak at 13 Å, detected only in the case of K+, the enzyme gets closer to the surface when potassium is present (figure 6b). Remarkably, no protein atom is detected at distances smaller than 0.5–1.0 nm from the surface atoms, indicating therefore the absence of direct protein–silica surface interactions and suggesting a key role of water molecules and co-solvating cations in mediating the adhesion of the negatively charged protein to the negatively charged inorganic matrix.
This picture is corroborated and rationalized by the pair interaction energies calculated for the different components of the model systems: protein, surface, water and cations (table 1). From the energies in the last line, which provide an estimate of the interaction between the inorganic matrix (slab) and the solvated protein (pep+H2O+ions), it is evident that the system with the charged surface and K+ is the one characterized by the strongest interaction, whereas no significant interaction is detected with the neutral slab. As expected, pepsin mainly interacts with water, and interaction with K+ is favoured over that with Na+ (lines one to four). Remarkably, Na+ shows a higher affinity for water with respect to K+, and, correspondingly, a weaker interaction with the charged surface.
To detect any modification of the secondary and tertiary structure of pepsin as a consequence of its interaction with the surface, we calculated the root mean square distance (r.m.s.d.) of the protein atoms with respect to the initial configuration along the four trajectories. The r.m.s.d. obtained for the whole protein interacting with the charged surfaces (figure 7, black lines), have an almost horizontal character and are below 2 Å. Similar trends are obtained in the case of the neutral surfaces (not shown). The largest contribution to the protein's r.m.s.d. is due to the dynamics of the residues in the flap region (grey/red lines). Keeping into account that flap flexibility is needed for the catalytic efficiency of the enzyme, this result suggests that pepsin catalytic properties are maintained. The calculated average distances between the two Asp in the catalytic dyad, close to the crystal structure value and characterized by small standard deviations, corroborate such a hypothesis (table 2). Moreover, the size and shape of the protein are also subjected to very small variations along the simulations, as can be deduced from table 2 and from the superposition of the 0–2.5 ns configurations (figure 8).
Orientation of the protein with respect to the surface has been monitored along the four trajectories, and some representative configurations are shown in figure 9 for the K+/charged surface model. Interestingly, owing to the weak surface–protein interaction, the protein shows a significant rotational mobility: indeed, different residues are found in closest contact with the surface along the trajectory and significant variations of the protein orientation are detected. It is to point out that pepsin presents a very complex charge distribution on its surface, with a predominance of negatively charged regions, but also with positively charged patches . The non-uniform charge distribution on pepsin, together with its water–cations-mediated interaction with silica, may therefore explain the absence of a preferred orientation of the enzyme with respect to the surface.
4. Discussion and conclusion
In this work, we have presented modelling studies aimed at the understanding of the efficacious interactions between a hydrated protein and a silica surface. The simulation model was set up in order to describe a weakly interacting host–guest system as can be found in the encapsulation process of proteins in mesoporous silicas, where the pores inner surfaces have surface silanol concentration in the 2–4 OH groups per square nanometre range. As the pores diameter in, for example, SBA-15 mesoporous materials can be comparable to the protein size, it might be difficult to disentangle the presence of effective interactions from geometrical confinement effects. Therefore, in this work, we have simulated a geometrically unconstrained protein-containing water droplet with diameter comparable with the SBA-15 pore section. Such a biodroplet was allowed to interact with a flat silica slab characterized by a surface silanol concentration close to that typical of mesoporous silica materials. Two surface models were considered, an electrically neutral slab and a negatively charged slab. Owing to the negative charge of pepsin, co-solvating cations were needed to ensure system electroneutrality. We have studied the effects of different cations, namely Na+ and K+.
The models allowed us to highlight that the silica electrostatic charge, owing to silanol deprotonation, influences the adhesion of a protein-containing water droplet onto the solid surface. Our results clearly indicate that a negatively charged globular protein, pepsin in this study, can be attracted by a negatively charged silica surface while keeping its hydrated environment. Indeed, adhesion of the hydrated protein to the charged surface patch occurs without any direct contact among protein atoms and surface atoms. The efficacious interaction is mediated by the solvent and strongly influenced by the nature of the counter ions. In the case of Na+, the water molecules stick, via hydrogen bonding, to the silanolates, while the cations are immersed in the droplet. A different adhesion mechanism has been evidenced in the case of potassium: K+ ions coordinate the negative surface sites and the adsorbed cations drag the protein-containing water droplet. In both cases, the droplet is attracted (immobilized) on top of the negative surface patch.
Our investigation will serve as a starting point for further modelling studies where the structure of a mesopore will be explicitly described, and where one might investigate the joint effects of the pore geometrical confinement and of the efficacious interaction here unravelled for the first time.
Finally, the results presented in this work are in line with the experimental observation that pore surface and protein need to have electrostatic charges of the same sign in order to maximize protein loading. Also in line with experimental evidence is our finding that the nature of the counter ions is an important factor for the optimization of protein encapsulation procedures inside mesoporous silica matrices.
One contribution of 11 to a Theme Issue ‘Structure and biological activity of glasses and ceramics’.
- This journal is © 2012 The Royal Society