Unraveling the Interaction between Histidine Side Chain and the Au

Aug 12, 2008 - Sayantani Chall , Soumya Sundar Mati , Indranee Das , Amrita Kundu , Goutam .... Prachi Joshi , Vasundhara Shewale , and Ravindra Pande...
1 downloads 0 Views 847KB Size
13540

J. Phys. Chem. C 2008, 112, 13540–13545

Unraveling the Interaction between Histidine Side Chain and the Au(111) Surface: A DFT Study Francesco Iori,*,† Stefano Corni,‡ and Rosa Di Felice‡ Department of Physics, UniVersity of Modena and Reggio Emilia, Modena, Italy, and INFM-CNR National Research Center on nanoStructures and bioSystems at Surfaces (S3), Modena, Italy ReceiVed: February 21, 2008; ReVised Manuscript ReceiVed: June 3, 2008

The interaction between proteins and the surfaces of inorganic materials is of great importance in natural systems and a long studied topic. What is yet to be fully understood is the mechanism that determines such interactions, in particular, for a given surface, which aminoacid, if any, binds and, in that case, what is the nature of the binding. The Au(111) surface, due to its utilization in several bioelectronics and biological applications, is one of the most commonly used substrates for experimental investigations. Among others, various experimental results have shown a clear affinity between the natural aminoacid histidine and the gold surface. Because the main contribution to the protein-surface interaction is likely mediated by the solvent accessible sidechains, we have focused our attention on imidazole, the molecule representing histidine side chain. Plane waves DFT calculations, in the supercell approach, have been performed to evaluate the nature of the interaction. Our results indicate the presence of a “on top” adsorption geometry with a covalent contribution to the binding between the unprotonated nitrogen atom and the superficial gold atom directly below it. Introduction The scientific knowledge on the interaction between biomolecules and metal surfaces is of great importance in many fields, from technological applications to the investigation of natural systems. In this context, the interaction between proteins and the surfaces of inorganic materials is of paramount importance, and it has been studied for a long time.1,2 Recently, it has been demonstrated that different combinatorial biotechniques, such as phage or cell-surface display, are efficient to select proteins able to specifically bind to a given inorganic surface.3 However, to date, the demonstration of specific proteinsurface associations has not been accompanied by an understanding of the mechanisms that determine the partnership and the resulting function. Moreover, while the ongoing growth of computational power and new theoretical methods have allowed classical molecular dynamics studies to investigate very large systems (several thousands of atoms), ab initio calculations involving even a small protein and an inorganic surface still require extreme supercomputing facilities. As initial working hypothesis, it is intuitive that the interaction between a protein and a surface occurs mainly through the aminoacid side chains, and this is our assumption. Thus, as a first important stage to understand the whole protein/surface complex, one can investigate the coupling between amino acid functional groups and the surfaces,4 a problem more manageable from both experimental and theoretical points of view. The Au(111) surface is one of the most commonly used substrates for theoretical and experimental investigations, due to its utilization in several bioelectronics5 and biological applications.6 The computing power required for studying, with ab initio methods, a single amino acid interacting with a gold * To whom correspondence should be addressed. E-mail: francesco.iori@ unimore.it. † University of Modena and Reggio Emilia. ‡ INFM-CNR.

surface can be considerable, especially when different conformations must be taken into account. For this reason, computational studies so far have dealt with aminoacids such as alanine,7 tyrosine,8 phenylalanine,8,9 glycine,10 or sulfur containing amino acids such as cysteine,11 the latter case aimed at the investigation of the S(thiolate)-Au bond, a well-known interaction with available experimental data. A similar computational approach has also been successfully applied to the study of DNA nucleobasis on the Au(111) surface by Piana and co-worker.12 Experimental studies by Zubavichus and co-workers13 on the adsorption of the amino acid histidine on Au(111) have suggested a variety of adsorption geometries. In particular, vacuum deposition of a histidine monolayer resulted in adsorption in the anionic form, in which the molecule is strongly coordinated to the surface via the combination of all the three possible coordinants: the amino group, the carboxylic group, and the unprotonated nitrogen atom of the side chain. In the work of Belcher and co-workers,14 performed with the yeast surface display technique, a histidine homohexapeptide has shown distinct binding affinity toward a gold film. In a protein structure, the aminoacidic carboxyl and the amino groups are condensed into the peptide bond between amino acids, resulting in a change in the chemical properties of the groups and in a reduction of its accessibility to an interacting surface, especially when compared to a side chain pointing outward. We can then hypothesize that the affinity of histidine-rich proteins for a gold surface is mainly due to the interaction between the two tautomeric states of its side-chain ring, methylimidazole and the substrate. In addition, the affinity of the imidazole ring for gold is further supported by the recent results that indicate that the adsorption of imidazole on gold nanoparticles results in selfassembled Au-imidazole complexes.15 In this paper, by means of first principles calculations in the framework of plane-wave pseudopotential density functional theory (DFT), we investigate imidazole interaction geometry and bonding properties on the Au(111) surface, assuming that

10.1021/jp801542s CCC: $40.75  2008 American Chemical Society Published on Web 08/12/2008

Histidine Side Chain and the Au(111) Surface the molecule is in its neutral configuration and without making any a priori assumption on the geometry of the imidazole/gold interface. When choosing a model for a real system, one should take into account what can be learnt from it and whether a more complex model would provide more information or just hinder the calculations. Based on this consideration we chose to further reduce histidine side chain (methylimidazole) to the simpler and more symmetrical imidazole. We are motivated by the necessity to focus our study on its more reactive part: indeed, the wellknown low reactivity of gold,16 combined with the poor description of van der Waals interactions by DFT methods, the only reasonable interaction between methyl and gold, would make its inclusion an unnecessary complication. Studies on similar systems resulting from molecular adsorption on gold have been performed with analogue plane-wave DFT methods, demonstrating the suitability of such theoretical approaches. Reimers and co-workers17,18 investigated the adsorption of ammonia and pyridine on a Au(111) surface; they have obtained a favored adsorption site on top of a surface atom for the nitrogen atom of both molecules. Ford and co-workers19 investigated the adsorption of a set of seven different amines on a Au(111) surface, in the presence and absence of an adatom: a direct interaction with the adatom, whenever present, is found to be preferable, while, in its absence, a surface gold atom is found to be the adsorption point. We aim here at unravelling the most favorable adsorption geometry in the case of imidazole adsorption on Au(111). Furthermore, our main target is the understanding of the resulting electronic properties, with a focus on the molecule/ metal hybridization. Methods In analogy with several works devoted to the investigation of molecules covalently bonded to metal surfaces,13,17–19 the unreconstructed 1 × 1 gold structure has been used to model the Au(111) surface. The rationale under this condition is that the energy gain upon formation of new bonds lifts the weaker driving forces toward the small displacements typical of the (3 × 23) herringbone reconstruction.20 The assumption is a-posteriori supported by our finding of “covalent” moleculesubstrate bonding. Furthermore, consider also that the tiny 0.3% displacements of the herringbone reconstruction are within the precision of the computational approach,21 so that a higher computational accuracy would be unlikely reached even by considering the gold reconstruction. Therefore, we have chosen a working compromise that is physically viable and at the same time allows an agile computational feasibility, given the fact that the large supercells needed to simulate the (3 × 23) surface would enormously increase the computational burden without bringing a significantly higher precision. A slab of four atomic layers with a periodically repeated (23 × 3) 2D supercell was used, corresponding to 12 atoms per layer: such a system was considered large enough to reproduce the adsorption of cysteine on a Au(111) surface with half the coverage of a “wet” monolayer.11a Periodic boundary conditions in all the three Cartesian directions have been applied. A vacuum thickness of 10 Å was used in calculations for structural relaxation. Each structural relaxation was followed by a selfconsistent calculation at the optimized geometry, with a supercell containing 25 Å of vacuum for a refinement of the total energy and the electronic structure and for the evaluation of the formation energy of the interface relative to the system composed of the isolated molecule and clean surface.

J. Phys. Chem. C, Vol. 112, No. 35, 2008 13541 The surface relaxation and all the necessary calculations have been performed in the frame of Density Functional Theory, using the gradient-corrected PBE22 exchange-correlation functional, with ultrasoft pseudopotentials23 to describe electronion interactions. Plane waves up to a kinetic energy cutoff of 25 Ry for electron wave functions and 200 Ry for electron density were included in the basis set. Such a basis set is sufficient to accurately describe the bond lengths of imidazole and the structural parameters of gold. A 4 × 4 × 1 MonkhorstPack k-point mesh was used for the Brillouin zone sums. All the atomic positions were relaxed until each atomic force component was smaller than 0.026 eV/Å. The geometry optimization for each of the various investigated structures was performed with a two-step scheme: (1) from the starting configuration, only the coordinates of the imidazole atoms were optimized, while the gold atoms were kept frozen; (2) then a second optimization, with all the atoms free to move, followed. Preliminary tests proved that this choice produces the same results as letting all the atoms relax from the starting guess, but with increased computational efficiency. The interaction energy between imidazole and the gold surface was calculated as the difference between the total energy of the optimized geometry of the molecule/gold system and the total energy of the system composed of a free molecule and a clean surface, simulated by placing the molecule at 10 Å from the surface, with the same orientation and in-plane location relative to the substrate as in the coupled interface. These total energies were computed at fixed optimized geometries with thicker supercells (containing 25 Å instead of 10 Å of vacuum) and with a finer 6-8 Monkhorst-Pack k-point mesh than those used in structural relaxations, to increase the computational accuracy. All the calculations were performed with the PWscf24 suite of programs, while graphical analysis and pictures have been obtained with Xcrysden.25 Results and Discussion Adsorption Energy and Geometry. Test calculations performed on the isolated imidazole molecule resulted in a good agreement (within ≈ 1%) with bond lengths and angles determined from experimental and other theoretical data.26 Several initial geometries have been taken into account for interface optimization, bearing in mind that the unprotonated nitrogen atom (N1 of Figure 1b), with its lone pair not involved in the aromatic π system, is the most reactive site for imidazole. We have considered structures with this atom in the proximity of the gold lattice hollow, bridge and top sites, highlighted in Figure 1a, with the molecular plane either parallel or perpendicular to the gold plane, as shown in Figure 1d-f. The first noticeable result of our calculations resides in the observation that, regardless of the initial guess, including the ones with N2 close to a gold atom, all the optimizations evolved toward the same geometry, with only one nitrogen atom (precisely N1) directly interacting with the surface. The imidazole ring in the relaxed structure is indeed positioned almost perpendicular to the surface plane with the N1 atom (N1int, henceforth) above a top site of Au(111), interacting directly with one Au atom (Auint, henceforth) as can be seen in Figure 1c. This means that the two initial geometries of Figure 1d,e, with the imidazole plane parallel to the (111) planes, and both nitrogen atoms close to the surface are not even metastable. The optimized imidazole/Au(111) interface is characterized by a N1int-Auint distance of ≈2.3 Å, with an 85° angle between the imidazole molecular plane and the metal surface plane. The

13542 J. Phys. Chem. C, Vol. 112, No. 35, 2008

Iori et al. TABLE 1: Interface Formation Energy and Distance for various Molecule/Metal Interfaces molecule a

imidazole pyridineb ammoniab ammoniac methylaminec dimethylaminec cysteined

kJ/mol

Å

45.6 30.5 32.2 21.0 28.0 28.0 74.5

2.30 2.46 2.45 2.45 2.46 2.55 2.51

x-c functional GGA/PBE GGA/PW91 GGA/PW91 GGA/PBE GGA/PBE GGA/PBE GGA/PW91

a This work. b Reimers et al.17,18 c Ford et al.19 d Absolute energy value for the reaction 2RS-H(g) + 2Au(111) f 2RS-Au(111) + H2(g), Di Felice et al.11a

Figure 1. (a) Top view of the Au surface top layer with highlights of the three typical low-energy adsorption sites in the hexagonal lattice. (b) Structure of imidazole. (c) Side view of the optimized geometry: N1 is above a top Au site, one-fold coordinated with Au, and the imidazole plane have an orientation of 85 degrees relative to the (111) substrate planes. (d) Top view of a starting configuration in which the N1 atom is located above a top Au site (N1 is one-fold coordinated with Au) and the imidazole plane lays parallel to the (111) substrate planes. (e) Top view of a starting configuration in which the N1 atom is located above a bridge Au site (N1 is 2-fold coordinated with Au) and the imidazole plane lays parallel to the (111) substrate planes. (f) Top view of a starting configuration in which the N1 atom is located above a top Au site (N1 is one-fold coordinated with Au) and the imidazole plane lays perpendicular to the (111) substrate planes.

adsorption energy gain is 45.6 kJ/mol. No significant distortion is observed on imidazole upon adsorption, while the interacting Auint atom moves outward by about 0.1 Å from the initial inplane position. Our result is also in agreement with the work of Holze,27 who obtained a perpendicular orientation upon the adsorption of molecular imidazole on a gold electrode. To deepen our understanding of the adsorption features, the formation energy was evaluated also for other imidazole/ Au(111) interfaces, obtained from that of Figure 1c by maintaining the azimuthal orientation of the molecular plane containing the [111] substrate direction and rotating it with respect to the other in-plane axes [11j0] and [112j]. The formation energy profile for this rotation is practically flat, with a maximum of ≈2 kJ/mol, corresponding to the closest distance between the lowermost hydrogen atom and the gold surface. This trend indicates that the main contribution to the energy comes from the interaction between the unprotonated nitrogen and the closer gold atom, while other geometrical/sterical effects do not play a dominant role. Due to the well-known shortcomings of DFT calculations in describing the nonbonding contributions to the formation energy, the value of the energy barrier reported here falls within the computational accuracy and should be taken with care. However, this fault does not relevantly invalidate our main conclusion here, namely that the axial rotation does not affect the interface formation energy. We demonstrate more closely the covalent nature in the next section by presenting and discussing the electronic structure. In Table 1 the adsorption energies and distances from the topmost surface plane for pyridine and the most strongly interacting amines, obtained with DFT calculations, are reported for comparison with imidazole. In all these cases, except for cysteine, the molecule resides with its most reactive atom (N for pyridine and amines) above a top site of the gold lattice. Imidazole has the smallest height above the surface among the

molecules in Table 1, and in particular with respect to its closest relative pyridine. However, to a small decrease (∼6%) of the imidazole height relative to pyridine corresponds a much larger increase of the adsorption energy (∼30%). Being evident that in both pyridine and imidazole cases the vicinity of a nitrogen lone pair to a surface gold atom is the key factor of the adsorption, it is nevertheless interesting to observe that the difference in interaction energy between the two molecules follow the trend of the increase in the π-system electron richness (i.e., the number of π-electrons per ring nuclear charge), from pyridine to imidazole. Such augmented electron availability may be a factor that helps the formation of an interaction of dative nature between the unprotonated nitrogen atom (N1 in Figure 1b) of imidazole and the surface gold atom on which it resides. Electronic Structure. To elucidate the nature of the molecule-surface interaction, we now present and discuss our results for the electronic structure of the imidazole/Au(111) interface. All the plots presented in this section are limited to the energy range comprehending Au valence states, where hybridization of the imidazole orbitals with the Au d band occurs: at lower energy, the molecular orbitals maintain a purely molecular character, as in the isolated molecules. In addition, to achieve a quantitative comparison of the shifts upon adsorption, the energy values of the noninteracting system have been shifted, aligning the deepest level to the corresponding one of the interacting system. This procedure is meaningful because we have verified that the deepest imidazole orbital remains itself upon adsorption. The Fermi energy of the interacting system has been set as the origin of the scale. The density of states (DOS) curves of the imidazole/Au(111) interface (black) and of the Au(111) clean surface (red), projected on all the orbitals of the Auint atom that interacts with N1int in the case of the interface, are presented in Figure 2. The changes in the peaks are a clear mark of the hybridization between orbitals of the two interacting species. This fact can be confirmed by comparing the density of states, projected (PDOS) onto the atomic orbitals centered on the atoms involved in the adsorption, in the interacting and noninteracting systems. By plotting the projection on each orbital separately, we obtain a finer understanding of the orbitals whose hybridization is responsible for the strength of adsorption. The PDOS curves on the s and pz orbitals of the N1 lone pair are shown in the top and middle panels of Figure 3 for the interacting and noninteracting systems. Instead of the sharp peak at -3 eV, typical of the isolated molecule (red), broad peaks appear upon adsorption (black). In particular, for s and pz orbitals, two main peaks can be identified, each of them the convolution of smaller peaks. One peak centered at ≈-5 eV is

Histidine Side Chain and the Au(111) Surface

J. Phys. Chem. C, Vol. 112, No. 35, 2008 13543

Figure 2. Projected DOS on the gold atom directly bonded to the imidazole N1int atom (black line), compared to the projected DOS on the same atom at the clean surface (red line). The DOS for the clean surface was extracted from the calculation for the noninteracting imidazole/gold system in which the molecule is placed 10 Å away from the surface.

down-shifted by -2 eV with respect to the free-molecule peak, and the other one centered at ≈-1.5 eV is up-shifted by 1.5 eV. Based on the knowledge of the adsorption mechanisms of thiols on gold,11a these features are likely to be associated with bonding and antibonding orbitals, respectively, and this is demonstrated by showing isosurface plots of relevant electron states. We remark that they are both related to occupied states because they are centered at energies that are below the Fermi level. In the bottom panel of Figure 3 it is possible to have a closer look at the effect of adsorption on the PDOS of Au: the peaks associated with dz2 noninteracting orbitals (red), located around -4 and -2 eV, show an evident shift upon adsorption (black) by -1.5 and 0.5 eV, respectively. The resulting energy values for these peaks are perfectly aligned to the energy values of N1int s and pz orbitals, indicating that at those energy values (≈-5 eV and ≈-1.5 eV) the interface system may be characterized by orbitals that possess both surface and adsorbate character. The symmetry of the interacting orbitals analyzed in Figure 3 points to a σ-like interaction between gold and imidazole. To inquire if a π-like contribution to the interaction is also present, we analyzed the relationship between imidazole px and gold dzx orbitals. The PDOS curves on the N1int px and Auint dzx orbitals are shown in Figure 4. Again, we find peaks centered at -4.0 and -2.2 eV peaks with a N1int character (top) and peaks with a Auint character, which is possibly associated to the appearance of peaks that have simultaneously both characters and thus indicating that also a π-like coupling contributes to the imidazole-gold interaction. Finally, the peak at ≈-7.5 eV, which appears with various intensities in all reported PDOS curves, is associated with orbitals located below the gold surface layer and orbitals located along imidazole covalent bonds. On the basis of the data presented so far, the small N1int-Auint distance and the indications from the DOS analysis of moleculesubstrate orbital mixing in the optimized imidazole/Au(111) interface geometry emerge as evidence of binding. However, the adsorption energy is smaller than those typical of covalent bonds, such as the bond occurring between cysteine or other thiols and the Au(111) surface.11 On the other hand, the imidazole adsorption energy is sensibly stronger than that of any other amines that have been studied with theoretical methods so far (see Table 1);17–19 thus, it is reasonable to consider the

Figure 3. PDOS curves on the N1 s (top), N1 pz (middle), and Au dz2 (bottom) orbitals. The black (red) line in all panels refers to the interacting (noninteracting) imidazole/Au(111) system. The Fermi energy is set as the origin of the energy scale.

existence of a weak covalent contribution to the bonding. To assess such a covalent contribution, one should find unoccupied orbitals with an antibonding character. The analysis of the PDOS’s reveals the splitting of the molecular orbitals, likely indicating bonding and antibonding. However, only the systematic analysis of the one-particle electron states can reveal the formation of bonding and antibonding orbitals. We have inspected several isosurface plots of orbitals of the interacting system falling at energies below the PDOS peaks that we have identified as possibly bonding and antibonding peaks, and few relevant ones are presented in Figures 5 and 6. As can be seen in Figure 5, isosurface plots of orbitals corresponding to energy values of -5.18 eV (a) and -4.45 eV (b) show bonding orbitals of π-like and σ-like nature, respec-

13544 J. Phys. Chem. C, Vol. 112, No. 35, 2008

Iori et al. TABLE 2: Lo¨wdin Charges of the Atoms Involved in the Adsorption, in the Isolated and Interacting Imidazole/ Au(111) Systemsa isolated interacting ∆e a

Figure 4. PDOS curves on the N1 px (top) and Au dzx (bottom) orbitals. The black (red) line in both panels refers to the interacting (noninteracting) imidazole/Au(111) system. The Fermi energy is set as the origin of the energy scale.

Figure 5. Isosurfaces plots for the imidazole/Au(111) interface. (a) Bonding orbital with π-like shape with energy -5.18 eV. (b) Bonding orbital with σ-like shape with energy -4.45 eV.

Figure 6. Isosurfaces plots for the imidazole/Au(111) interface. (a) Antibonding orbital with π-like shape with energy -0.1 eV. (b) Antibonding orbital with σ-like shape with energy 0.75 eV.

tively. In general, hybrid orbitals with energy between -5.7 and -4.2 eV, considering the origin of the scale as the Fermi energy, display bond-like shapes. Starting from an energy value of -4.1 eV to values above the Fermi energy, a number of orbitals of π-like and σ-like antibonding nature can be observed. Two such states at the energies of -0.1 and 0.75 eV are illustrated with isosurface plots in Figure 6a and b, respectively. The existence of antibonding states below the Fermi energy appears to be the main reason for the considerable weakening

Au

N1

imidazole

0.1 0.0 -0.1

-0.3 -0.1 0.2

0.0 0.3 0.3

Values are in units of the absolute value of electron charge e.

of the interaction between gold and imidazole. However, the presence of hybrid orbitals of antibonding nature for values higher than the Fermi energy (i.e., unoccupied) suggests that the interaction energy has a net bonding contribution. In particular, it is interesting to observe that it was not possible to clearly identify any unoccupied antibonding orbital of π-like nature. This occurrence suggests that the sole contribution to the bonding is of σ nature, a conclusion coherent with the low energy barrier that is present for imidazole azimuthal rotation. To complete the analysis and verify the hybridization effect on the charge distribution, the electronic charges localized on the atoms involved in the interaction and on the whole imidazole molecule have been calculated with the Lo¨wdin’s scheme28 and the results are shown in Table 2. This method implies the building of an auxiliary basis set with atom-centered Slater functions, with exponents suitable for isolated atoms, unto which the projection of the molecular orbitals is performed. The accuracy of the auxiliary basis set was verified: the norm of the projected molecular orbitals here discussed is