Density Functional Theory Study of the Binding of Glycine, Proline

Mar 18, 2009 - ... ‡Department of Earth Sciences, University of Cambridge, Downing Street, ... apatite Ca10(PO4)6(OH)2 (HA),4 a protein matrix, and ...
0 downloads 0 Views 4MB Size
Download PDF
pubs.acs.org/Langmuir © 2009 American Chemical Society

Density Functional Theory Study of the Binding of Glycine, Proline, and Hydroxyproline to the Hydroxyapatite (0001) and (0110) Surfaces Neyvis Almora-Barrios,*,† Kat F. Austen,‡ and Nora H. de Leeuw† †

Department of Chemistry, University College London, 20 Gordon Street, London WC1H 0AJ, United Kingdom and ‡Department of Earth Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EQ, United Kingdom Received November 20, 2008. Revised Manuscript Received February 6, 2009 In view of the importance of the hydroxyapatite/collagen composite of both natural bone tissue and in synthetic biomaterials, we have investigated the interaction of three constituent amino acids of the collagen matrix with two major hydroxyapatite surfaces. We have employed electronic structure techniques based on the density functional theory to study a range of different binding modes of the amino acids glycine, proline, and hydroxyproline at the hydroxyapatite (0001) and (0110) surfaces. We have performed full geometry optimizations of the hydroxyapatite surfaces with adsorbed amino acid molecules to obtain the optimum substrate/ adsorbate structures and interaction energies. The calculations show that the amino acids are capable of forming multiple interactions with surface species, particularly if they can bridge between two surface calcium ions. The binding energies range from 290 kJ mol-1 for glycine on the (0001) surface to 610 kJ mol-1 for hydroxyproline on the (0110) surface. The large adsorption energies are due to a wide range of interactions between the adsorbate and surface, including proton transfer from the adsorbates to surface OH or PO4 groups. Hydroxyproline binds most strongly to the surfaces, but all three amino acids should be good sites for the nucleation and growth of the hydroxyapatite (0110) surface at the collagen matrix.

1. Introduction A wide variety of materials have been used for hard tissue replacement in dental and orthopaedic applications, ranging from metals to ceramics,1-3 where they are often designed to resemble as closely as possible the natural tissue. Bone itself is a highly hierarchical collagen/mineral composite, containing nanosized mineral platelets, essentially carbonated hydroxyapatite Ca10(PO4)6(OH)2 (HA),4 a protein matrix, and water.5 The protein in bone is predominantly type I collagen, the structure of which is complex and which self-assembles in a multistep process.6 In summary, a triple helix is formed from three polypeptide chains with a highly repetitive amino acid sequence with glycine (NH2-CH2-COOH) in every third position [GLY-X-Y]n, where X and Y are commonly proline (NH-C4H7-COOH) and hydroxyproline (NHC4H6(OH)-COOH), respectively. Experimental studies have demonstrated that the growth process of HA is controlled by the interaction between the organic matrix and the HA crystal,7-11 which determines the *To whom correspondence should be addressed. E-mail: n.barrios@ ucl.ac.uk. (1) Elbert, D. L.; Hubbell, J. A. Annu. Rev. Mater. Sci. 1996, 26, 365–394. (2) Klee, D.; Hocker, H. Biomed. Appl./Polym. Blends, 1999 1999, 149, 1–57. (3) Griffith, L. G. Acta Mater. 2000, 48(1), 263–277. (4) Ducheyne, P.; Qiu, Q. Biomaterials 1999, 20(23-24), 2287–2303. (5) Fratzl, P.; Gupta, H. S.; Paschalis, E. P.; Roschger, P. J. Mater. Chem. 2004, 14(14), 2115–2123. (6) Stevens, M. M.; George, J. H. Science 2005, 310(5751), 1135–1138. (7) Du, C.; Falini, G.; Fermani, S.; Abbott, C.; Moradian-Oldak, J. Science 2005, 307(5714), 1450–1454. (8) Kirkham, J.; Brookes, S. J.; Shore, R. C.; Wood, S. R.; Smith, D. A.; Zhang, J.; Chen, H. F.; Robinson, C. Curr. Opin. Colloid Interface Sci. 2002, 7(1-2), 124–132. (9) Minton, A. P. Biophys. J. 1999, 76(1), 176–187. (10) Smith, C. E.; Nanci, A. Anat. Rec. 1996, 245(2), 186–207. (11) Wen, H. B.; Fincham, A. G.; Moradian-Oldak, J. Matrix Biol. 2001, 20(5-6), 387–395.

5018

DOI: 10.1021/la803842g

eventual morphology of the HA platelets in the bone. The apatite platelets are very thin indeed, typically 2-4 nm,5 and these crystals are arranged in an ordered fashion within and around the collagen strands to form mineralized collagen fibrils. The apatite mineral is aligned with its c-axis, the [0001] direction, along the fibril. These mineralized fibrils are then arranged in parallel arrays, and the arrays are further organized into three-dimensional structures to form the bone tissue.12 The precise structure and orientation of the mineralized collagen fibril in the bone largely depends on its function; regions experiencing tensile loading are inclined to have more longitudinally oriented fibers, while regions under compressive loading typically contain more transverse fibers.12-14 As the mineral component of bone is predominantly carbonate-substituted hydroxyapatite, HA is now used as an effective, commercially available bone implant material. The implant can be built as an interconnected network of channels, as a porous structure encourages bone cell ingrowth. HA/collagen composite materials are biodegradable and good substrates for bone cell attachment and proliferation as well as new bone formation.15 Understanding at the atomic level the binding of proteins to inorganic surfaces obviously plays an important role in the development of these materials for biocompatible medical implants, and quantitative knowledge of the collagen/HA interface is required to understand the nucleation and directed growth of hydroxyapatite at the collagen matrix.

(12) Vanderby, R. J. Biomech. 2003, 36(10), 1523–1527. (13) Martin, R. B.; Burr, D. B.; Sharkey, N. A. Skeletal tissue mechanics; Springer: New York, 1998. (14) Cui, F. Z.; Li, Y.; Ge, J. Mater. Sci. Eng., R 2007, 57(1-6), 1–27. (15) Tirrell, M.; Kokkoli, E.; Biesalski, M. Surf. Sci. 2002, 500(1-3), 61–83.

Published on Web 3/18/2009

Langmuir 2009, 25(9), 5018–5025

Almora-Barrios et al.

Article

The influence of amino acids on hydroxyapatite crystallization has been well studied experimentally,16-19 reporting that amino acids are effective regulators in processes involving both the nucleation and crystal growth of HA. However, detailed structural information such as the binding sites and adsorption modes of these biomolecules on the HA surfaces is still largely lacking. The present paper therefore reports a detailed computational study of the interaction of the three amino acids, glycine, proline, and hydroxyproline (GLY, PRO, HYP), that largely make up the collagen matrix with two major hydroxyapatite surfaces. We have employed electronic structure techniques based on the density functional theory (DFT) to study different modes of adsorption of the individual amino acids at the HA surfaces. In particular, we have studied the functional groups of the three amino acids, which are accessible to the hydroxyapatite mineral when it is deposited at the collagen triple helix. These calculations provide detailed quantitative information on the interfacial structure, binding energies, and chemical processes taking place between the adsorbates and the major surface features of the apatite mineral.

2.

Theoretical Methods

The simulations were performed using the SIESTA code,20 which employs DFT,21 norm-conserving pseudopotentials, and linear combinations of numerical atomic orbitals (LCAO) to calculate the total energy of the system. We have used the Perdew-Burke-Ernzerhof 22 generalized gradients approximation for the exchange-correlation functional, and pseudopotentials for all atoms were generated in the Troullier-Martins manner.23 The basis sets used for all atoms were of the DZP type (double ζ with polarization), apart from the oxygen and hydrogen in the hydroxy group, where the basis sets were obtained from the optimization of water at 0.2 GPa.24,25 For the oxygen, carbon, nitrogen, and hydrogen in the amino acids, the basis sets were optimized with respect to the interaction of guanine-cytosine at 0.2 GPa.26 The unit cell was optimized using a cutoff energy of 250 Ry, sampling was taken at the Γ point, and a force tolerance of 0.01 eV A˚-1 was used. The suitabilities of cutoff energy, k points, and force tolerance were evaluated by monitoring the convergence of the total energy with respect to the various parameters and validated against the structural properties of the hydroxyapatite. The crystal structure of hydroxyapatite has the P63/m space group with the 4e Wyckoff positions occupied by two hydroxy oxygen atoms, each with 1/2 occupancy.27 In order to (16) Koutsopoulos, S.; Dalas, E. Langmuir 2000, 16(16), 6739–6744. (17) Koutsopoulos, S.; Dalas, E. J. Cryst. Growth 2000, 216(1-4), 443–449. (18) Pan, H. H.; Tao, J. H.; Xu, X. R.; Tang, R. K. Langmuir 2007, 23(17), 8972–8981. (19) Jack, K. S.; Vizcarra, T. G.; Trau, M. Langmuir 2007, 23(24), 12233–12242. (20) Ordejon, P.; Artacho, E.; Soler, J. M. Phys. Rev. B 1996, 53(16), 10441–10444. (21) Kohn, W.; Sham, L. J. Phys. Rev. 1965, 140(4A), 1133. (22) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77(18), 3865–3868. (23) Troullier, N.; Martins, J. L. Phys. Rev. B 1991, 43(11), 8861–8869. (24) Anglada, E.; Soler, J. M.; Junquera, J.; Artacho, E. Phys. Rev. B 2002, 66(20), 205101. (25) Junquera, J.; Paz, O.; Sanchez-Portal, D.; Artacho, E. Phys. Rev. B 2001, 64(23), 235111. (26) Fernandez-Serra, M. V.; Junquera, J.; Jelsch, C.; Lecomte, C.; Artacho, E. Solid State Commun. 2000, 116(7), 395–400. (27) Stork, L.; Muller, P.; Dronskowski, R.; Ortlepp, J. R. Z. Kristallogr. 2005, 220(2-3), 201–205.

Langmuir 2009, 25(9), 5018–5025

Figure 1. Plan and side views of the hydroxyapatite structure, showing the OH- groups in hexagonal channels surrounded by triangles of Ca ions (Ca = green, O = red, P = purple, and H = white). translate this structure into a model with full occupancies for the hydroxy groups, as is required to carry out the calculations, we have assigned alternate 0 and 1 occupancies to these sites. We have also doubled the unit cell in the b-direction to obtain two OH columns per unit cell, which alternate in the direction of the hydroxy groups (Figure 1).28 This arrangement, shown in Figure 1b, where all hydroxy groups within a column are pointing in the same direction but with alternating columns in the b-direction, has been calculated to be the most stable arrangement within the hexagonal HA crystal structure28 and is also the arrangement of the hydroxy groups in pure, synthetic HA, which crystallizes as an ordered monoclinic structure with a double unit cell compared to the hexagonal structure.27 However, as natural HA has the hexagonal structure, we have used this structure for our calculations, thus allowing direct comparison of the simulations with experimentally determined surfaces. The HA surfaces were modeled as slabs of 10 A˚ thick, perpendicular to the surface, and infinitely repeated laterally through three-dimensional periodic boundary conditions. The slabs were separated from their images in the neighboring cells by a vacuum region of approximately 50 A˚, thus avoiding interactions between repeating slabs. The stabilities of the surfaces are described by their surface energy, which experimentally is the energy required to cleave the bulk crystal exposing the surface, given by γ ¼ ðEsurf -Ebulk Þ=A

ð1Þ

where Esurf is the energy of the simulation cell containing the surface, Ebulk is the energy of an equivalent number of bulk stoichiometric units, and A is the surface area. A low positive (28) de Leeuw, N. H. Chem. Commun. 2001, 17, 1646–1647.

DOI: 10.1021/la803842g

5019

Article

Almora-Barrios et al.

Figure 2. Calculated lowest-energy structures of (a) glycine, (b) proline, and (c) hydroxyproline, showing intramolecular hydrogen bonding (C = gray, O = red, N = blue, and H = white). value for γ indicates a stable surface. We have therefore used a (2  2  1) simulation supercell for all the calculations. Tests of the convergence of the total energy and properties with respect to the slab thickness showed that the surface energies are not converged entirely but are within 0.05 J m-1, whereas on increasing the slab thickness the adsorption energies and indeed the structures of the surfaces were not affected significantly (within 1 kJ mol-1) (see Tables S1 and S2 of the Supporting Information). The amino acid molecules glycine, proline, and hydroxyproline were optimized as neutral molecules and as zwitterions, which would be their preferred configuration in water. However, we found that in isolation the neutral forms of the molecules were preferred over the charge-separated configuration, which rearranged upon geometry optimization. The optimum structures of the free amino acids are largely due to the formation of intramolecular hydrogen bonds, which are shown in Figure 2. The instability of the zwitterionic structure within gas phase calculations has been reported previously for glycine in detailed ab initio investigations,29,30 and it is therefore not surprising that the same is found for proline and hydroxyproline. In any case, in the protein, the amino acids would not be charged either and previous simulations of the adsorption of organic molecules with amine and carboxylic acid groups to apatite and other mineral surfaces have shown that the use of neutral molecules provides quantitative insight into adsorption modes and strengths, especially in a comparison between different surfaces and adsorbates.31-33 Thus, the processes including interaction with water molecules were not considered in this study, although future work will build on (29) Ding, Y. B.; Kroghjespersen, K. Chem. Phys. Lett. 1992, 199(3-4), 261–266. (30) Yu, D.; Armstrong, D. A.; Rauk, A. Can. J. Chem. 1992, 70(6), 1762–1772. (31) de Leeuw, N. H.; Cooper, T. G. Cryst. Growth Des. 2004, 4(1), 123–133. (32) Cooper, T. G.; de Leeuw, N. H. Langmuir 2004, 20(10), 3984–3994. (33) Mkhonto, D.; Ngoepe, P. E.; Cooper, T. G.; de Leeuw, N. H. Phys. Chem. Miner. 2006, 33(5), 314–331.

5020

DOI: 10.1021/la803842g

this study to use classical methods to improve our understanding of the effect of water on the adsorption of amino acids at the hydroxyapatite interface. In our investigation of the sorption of the amino acids at the HA surfaces, we have adsorbed one molecule per surface simulation cell at a variety of different surface sites to identify the energetically most favorable location. Experimental studies report that the primary association between amino acids and the hydroxyapatite surface is via interactions between the carboxylic acid groups and the surface calcium ions. Furthermore, amine groups could interact with oxygens of phosphate groups.16,17,34 This information was useful to identify likely adsorption sites in both (0001) and (0110) surfaces as starting positions for our simulations. In addition, the amino acids were adsorbed at the surfaces in their optimized geometries in a number of different initial configurations, before the complete surface/adsorbate systems were geometry optimized again and the adsorption energies calculated. In each case, the surface simulation cells were supercells of sufficient size (262-354 A˚2), containing 176 atoms and a vacuum gap of at least 50 A˚, to ensure the absence of any computational artifacts due to the periodic boundary conditions parallel to the surface or across the vacuum gap. If the cells had been too small, the adsorbed amino acids could have interfered with their images in the periodically repeated surface cell, which would affect the geometries and energies. The adsorption energies due to the interaction of the amino acids with the HA surfaces were calculated according to eq 2: Eads ¼ Esystem -ðEsurf þ Ea:a: Þ

ð2Þ

where Esystem is the energy of the surface with adsorbed amino acid, Esurf is the energy of the simulation cell containing the surface only, and Ea.a. is the self-energy of the free amino acid molecule in its optimized geometry, calculated using the same conditions. A negative adsorption energy thus indicates that adsorption of amino acids at the surfaces is thermodynamically favorable.

3. Results and Discussion First, we have geometry optimized the bulk HA structure, where the cell parameters of the relaxed structure are calculated as a = 9.403 A˚, 2b = 18.812 A˚ (b = 9.406 = a), c = 6.954 A˚, R = β = 90°, and γ = 120° (Figure 1), in clear agreement with experiment a = b = 9.432 A˚, c = 6.881 A˚, R = β = 90°, and γ = 120°.35 The calcium and phosphate ions respond little to the location of the hydroxy groups, which are stacked in a regular column within the hexagonal channels formed by triangles of Ca ions, although the direction of the OH groups may differ randomly between neighboring channels without an energetic penalty.28,36 The PO4 groups display little variation in size and shape, and the calculations yield an average P-O bond length of 1.596 A˚ and average O-P-O angle of 109.41°, as compared with the corresponding experimental averages of 1.535 A˚ and 109.45°, respectively. 3.1. HA Surfaces. Having optimized the HA bulk structure, we have used the METADISE code37 to create (34) Koutsopoulos, S.; Dalas, E. Langmuir 2001, 17(4), 1074–1079. (35) Posner, A. S.; Perloff, A.; Diorio, A. F. Acta Crystallogr. 1958, 11(4), 308–309. (36) de Leeuw, N. H. Phys. Chem. Chem. Phys. 2002, 4(15), 3865–3871. (37) Watson, G. W.; Kelsey, E. T.; deLeeuw, N. H.; Harris, D. J.; Parker, S. C. J. Chem. Soc., Faraday Trans. 1996, 92(3), 433–438.

Langmuir 2009, 25(9), 5018–5025

Almora-Barrios et al.

the surfaces, which were subsequently optimized using the SIESTA method. In this study, we have concentrated on two of the most significant surfaces of the apatite bone mineral, namely, the (0001) and (0110) surfaces. These two surfaces are important in the morphology of the apatite mineral platelets, and there is experimental evidence that these faces act as the binding site for many ionic species, including small molecules, polymers, and anionically modified cell surfaces.38,39 The basic building block of the (0001) surface consists of three phosphate groups and three calcium ions, which is translated along the a and b axes to form the infinite surface. Previous energy minimization calculations of the different terminations of a number of HA surfaces40 have shown that this (0001) termination is the most stable, hence its use in this work. It is not possible to construct a HA slab of the (0110) surface with the same surface termination on both sides, even though the total dipole moment across the slab perpendicular to the surface is zero, which is the requisite for a stable surface. As a result, the bottom plane of the slab contains two calcium ions only, whereas the top plane has an extra two hydroxy groups. The (Ca-Ca) termination is the more stable, but the (OH-OH-Ca-Ca) termination has a greater variation of surface species, which could form different interactions with the adsorbates,40 and both planes will occur during the HA growth process. The surface energy of this slab will be an average of the two (0110) terminations, but the adsorption energy will be specific to the particular plane used as substrate. For our studies, we have used the more reactive (OH-OH-Ca-Ca) termination, as this provides a wider variety of adsorption sites and hence will provide more insight into the different modes of adsorption. The calculated surface energies of the (0001) and (0110) surfaces are 1.01 and 1.32 J m-1, respectively, indicating that the (0001) surface is more stable than the (average) (0110) surface, in agreement with classical calculations.40,41 The less stable surface will be more reactive and may quickly accumulate more HA material to eventually grow out of the crystal morphology. However, its higher reactivity may also lead to a stronger interaction with the amino acids, which is important for the hydroxyapatite formation at the collagen matrix. The relaxed (0001) and (0110) hydroxyapatite surfaces are shown in panels (a) and (b), respectively, of Figure 3. After relaxation, the symmetry is broken due to the reorientation of the hydroxy groups within the channel in both surfaces. Upon geometry optimization of the bulk crystal, the hydroxy groups remain lined up within the columns, where the OH 3 3 3 OH hydrogen bond distance averages 2.49 A˚, in agreement with previous DFT calculations using a planewave pseudopotential approach.28 However, when a surface is present, the now undercoordinated surface hydroxy groups rotate to increase their interaction with each other and with phosphate oxygen atoms, as shown in Figure 3. In the (0110) surface, the disordering of hydroxy groups is more significant than in the (0001) surface, probably due to their initial orientation parallel to the surface plane, although the surface area of the (0001) surface is greater than that of (38) Wierzbicki, A.; Cheung, H. S. THEOCHEM 2000, 529, 73–82. (39) de Leeuw, N. H. Phys. Chem. Chem. Phys. 2004, 6(8), 1860–1866. (40) Filgueiras, M. R. T.; Mkhonto, D.; de Leeuw, N. H. J. Cryst. Growth 2006, 294(1), 60–68. (41) de Leeuw, N. H.; Rabone, J. A. L. CrystEngComm 2007, 9(12), 1178–1186.

Langmuir 2009, 25(9), 5018–5025

Article

Figure 3. Relaxed surfaces of hydroxyapatite: (a) PO4-Ca-PO4-

PO4-Ca-Ca termination of the (0001) surface and (b) Ca-Ca in the bottom and (OH-OH-Ca-Ca) in the top of the (0110) surface (Ca = green, O = red, P = purple, and H = white).

the (0110) surface and will therefore allow more surface relaxation, if required. In the (0001) surface, the surface hydroxy groups remain in their original orientation, but they relax into the bulk material. The OH 3 3 3 OH hydrogen bond distances shorten to an average value of 2.09 A˚, and hydrogen bonds between OH 3 3 3 OPO3 are formed at about 2.44 A˚. In the (0110) surface, the hydroxy groups remain more or less parallel to the surface plane, although half of the groups rotate to interact more closely with neighboring OH groups with average OH 3 3 3 OH distances of 2.31 A˚, whereas the resulting OH 3 3 3 OPO3 distances are about 2.13 A˚. There is no significant movement of the surface calcium ions or subsurface phosphate groups. 3.2. Adsorption of Amino Acids. We next placed the optimized amino acids, above the surfaces, in a range of different starting positions and surface sites. Once located near the surface, the complete system was geometry-optimized once again, and the molecules were allowed to translate, rotate, and distort freely during the relaxation to optimize their interaction with the surfaces. However, for the amino acid molecule to distort significantly from its ideal structure, clearly the strength of binding to the surface will have to outweigh the loss of energy due to the breaking of the intramolecular hydrogen-bonded network in the free molecule. We report here the lowest-energy surface/adsorbate structures, but details of less favorable adsorption geometries are reported in the Supporting Information. The lowest-energy geometry of the glycine adsorbed at the (0001) surface is shown in Figure 4a (GLY-1), where we see that the glycine molecule adsorbs into the gap in the surface, which is an undercoordinated and hence reactive site. The amine group is located above a surface phosphate group, where in the bulk material a calcium ion would be sited, and the amine group’s hydrogen atoms form a number of DOI: 10.1021/la803842g

5021

Article

Almora-Barrios et al.

Figure 4. Geometry-optimized structures of glycine adsorbed at the hydroxyapatite surfaces, showing interatomic distances: (a) overall and (b) detailed view of the (0001) surface and (c) overall and (d) detailed view of the (0110) surface (Ca = green, O = red, P = purple, H = white, C = gray, and N = blue). hydrogen-bonded interactions to the oxygen atoms of the phosphate group with (NH 3 3 3 OPO3) distances ranging from 1.47 to 1.54 A˚. In addition, the oxygen atoms of the carboxylic acid group form a number of interactions with the topmost calcium ions (Ca 3 3 3 O = 2.29-2.33 A˚), and as a result the hydrogen atom of the carboxyl group migrates to the amine group. As such, the amino acid adsorbs as a zwitterion in the energetically preferred mode (Figure 4b). The electrostatic interactions between the carboxylic acid group and calcium ions and the hydrogen bonds between the amine group and surface oxygen atoms of the phosphate group give rise to strong binding of the glycine to the (0001) surface of hydroxyapatite. The main difference between this structure and GLY-2 (shown in the Supporting Information) is the proton transfer from the carboxylic acid group to the amine group of glycine, rather than to the surface phosphate group, whereas GLY-3 and GLY-4 have lost the interaction between the carboxylic acid group and the surface calcium ions, as shown in the Supporting Information. In summary, glycine is adsorbed at the hydroxyapatite (0001) surface in its zwitterionic form, which is in agreement with the recent computational study of Rimola et al.42 On the (0110) surface, the most favorable mode of adsorption for glycine is by bridging between three calcium atoms through coordination by a carboxylic acid group (42) Rimola, A.; Corno, M.; Zicovich-Wilson, C.; Ugliengo, P. J. Am. Chem. Soc. 2008, 130(48), 16181–16183.

5022

DOI: 10.1021/la803842g

and by its nitrogen, shown in Figure 4c and d (GLY-1). Upon adsorption to the surface, the hydrogen atom of the carboxylic acid group migrates to the oxygen atom of a surface phosphate group, leaving the glycine molecule with a carboxylate group, which interacts strongly with surface calcium ions as described below. One carboxylate oxygen is coordinated to one calcium at a distance of 2.43 A˚ and to another calcium at a distance of 2.39 A˚, whereas the other carboxylate oxygen is coordinated to a calcium at a distance of 2.37 A˚. The nitrogen atoms interact with calcium at a distance of 2.55 A˚. In addition, a range of hydrogen bonds between the hydrogen atoms of the amine group and oxygen atoms of the surface hydroxy and phosphate groups further enhance the binding between the glycine molecule and the (0110) surface (NH 3 3 3 OPO3 = 2.07-2.76 A˚). The GLY-2 structure (see the Supporting Information) is very similar to GLY-1 (difference in energy is only 7 kJ mol-1), with the main difference being the more numerous interactions between the carboxylic acid group and surface calcium ions and the hydrogen bonding between the carboxylate group and the hydroxy group present in the GLY-1 model. However, in GLY-3, the proton migrates to a surface hydroxy group, but this process is energetically less favorable by approximately 111.00 kJ mol-1. This proton transfer leads to the formation of a molecule of water that competes with the anionic form of glycine for adsorption on the surface of hydroxyapatite. The GLY-4 model is similar to GLY-3, although the positions of the amine group and carboxylic acid group are different. Glycine is the only adsorbate in a study where this proton Langmuir 2009, 25(9), 5018–5025

Almora-Barrios et al.

Article

Figure 5. Geometry-optimized structures of proline adsorbed at the hydroxyapatite surfaces, showing interatomic distances: (a) overall and (b) detailed view of the (0001) surface and (c) overall and (d) detailed view of the (0110) surface (Ca = green, O = red, P = purple, H = white, C = gray, and N = blue). transfers to a surface hydroxy group. With the other adsorbates, the carboxyl proton migrates to the oxygen atom of a surface phosphate group (see the Supporting Information). Figure 5 shows the proline molecule adsorbed at the two surfaces. On the (0001) surface, the hydrogen atoms of the amino group interact with the surface phosphate oxygen atoms (NH 3 3 3 OPO3 = 1.48-1.78 A˚), whereas the carboxylic acid group interacts directly with two calcium ions and a hydroxy group (CO 3 3 3 Ca = 2.41-2.54 A˚ and CO 3 3 3 HO = 2.61 A˚) (Figure 5a, b). These strong interactions between the surface and proline molecule again cause the carboxyl proton to migrate to the amine group, where the three hydrogen atoms interact with the oxygens of a surface phosphate group. On the (0110) surface, however, the most favorable adsorption mode for the proline is through coordination of both oxygen atoms of the carboxylic acid group to a surface calcium atom, at distances of 2.35, 2.43, and 2.49 A˚, and a surface hydroxy group (CO 3 3 3 HO = 2.88 A˚) (Figure 5c, d), together with electrostatic interactions between nitrogen atoms to surface calcium (-NH 3 3 3 Ca = 2.59 A˚). Again, the carboxyl proton migrates, this time first to the amine group, followed by a hop to the oxygen atom of a surface phosphate group. Hydroxyproline adsorbs in a similar way to proline, which is not surprising as the structures of the two molecules are similar, where only one ring hydrogen of proline is replaced by a hydroxy group to form hydroxyproline. The substitution adds to the number and strength of the interactions between the adsorbate and the surface species. For example, in hydroxyproline, both carboxylic acid and hydroxy groups can interact with surface calcium ions, Langmuir 2009, 25(9), 5018–5025

whereas their hydrogens can form numerous hydrogenbonded interactions to phosphate and hydroxy oxygen atoms. Figure 6a, b (HYP-1) shows adsorption on the (0001) surface, where the interaction with the surface occurs via the hydrogen and oxygen atoms of the hydroxy group and hydrogens of the amine group (OH 3 3 3 OPO3 = 1.592.71 A˚, OH 3 3 3 HO = 2.14 A˚, and NH 3 3 3 OPO3 = 1.46-2.96 A˚). The carboxylic acid group, again, interacts with surface calcium ions (COO 3 3 3 Ca = 2.29-2.79 A˚) as well as oxygen atoms of the hydroxy group (HO 3 3 3 Ca = 2.91 A˚), and the carboxyl proton migrates to the amine group. On the (0110) surface, hydroxyproline adsorbs almost flat onto the surface, where a number of close interactions are formed between the molecule’s nitrogen atom and oxygen atoms of the carboxylic acid group and three surface calcium ions (COO 3 3 3 Ca = 2.34-2.38 A˚, N 3 3 3 Ca = 2.61 A˚). In addition, hydrogen bonding takes place between oxygen atoms of a surface phosphate group and the hydrogen atoms of the molecule’s -OH and >CH2 groups (-OH 3 3 3 OPO3 = 1.78-2.99 A˚, CH 3 3 3 OPO3 = 2.04-2.19 A˚). As in proline, the carboxylic acid proton migrates to an oxygen atom of the surface phosphate group, and further hydrogen bonding occurs between the oxygen atoms of the carboxylate group and the hydrogen atoms of the two topmost hydroxy groups (COO 3 3 3 HO = 2.83-2.85 A˚). 3.3. Adsorption Energies. The adsorption energies of the amino acids at the surfaces, calculated according to eq 2, are listed in Table 1. Experimental studies16,17 have shown that amino acids interact strongly with HA, which is borne out by our calculations. However, all adsorbates bind much more strongly to the (0110) surface than to the (0001) DOI: 10.1021/la803842g

5023

Article

Almora-Barrios et al.

Figure 6. Geometry-optimized structures of hydroxyproline adsorbed at the hydroxyapatite surfaces, showing interatomic distances: (a) overall and (b) detailed view of the (0001) surface and (c) overall and (d) detailed view of the (0110) surface (Ca = green, O = red, P = purple, H = white, C = gray, and N = blue). Table 1. Calculated Adsorption Energies of Amino Acids at Hydroxyapatite Surfaces adsorption energies (kJ mol-1) structure

(0001)

(0110)

GLY

1 2 3 4

-291.00 -211.81 -173.18 -150.76

-496.56 -489.14 -385.56 -359.92

PRO

1 2 3 4

-322.30 -132.29 -115.32 -70.58

-554.50 -509.77 -414.04 -318.23

HYP

1 2 3 4

-507.65 -383.02 -270.00 -106.73

-609.79 -527.76 -250.62 -226.62

surface. The (0001) surface is a stable surface with only a few dangling bonds before adsorption, and as a result the adsorbates have little effect on the surface structure. However, the many interactions between adsorbate and surface lead to the large energies released upon adsorption. As discussed, the (0110) surface is less stable than the (0001) surface, as is shown by its higher surface energy (see the Supporting Information), which is due to a larger number of undercoordinated surface species with dangling bonds, making the (0110) surface more reactive than the (0001) surface. Saturation of the dangling bonds through the 5024

DOI: 10.1021/la803842g

adsorption of the amino acids, as well as the proton transfer occurring upon adsorption from the adsorbate to the surface, leads to the large chemisorption energies released upon adsorption at the (0110) surface. The process of adsorption at the (0110) surface always includes chemical reactions in the way of proton transfer to the surface, which creates charge separation between the surface and adsorbate, leading to strong ionic bonding between the carboxylate group and surface calcium ions, explaining the high adsorption energies for this surface. When we compare the adsorption energies for the different amino acids (Table 1), we see that, in general, the energies released by adsorption of proline and hydroxyproline are larger than those of glycine, even though the glycine molecule is more flexible. As the carboxylic acid and amine functional groups are present in all those adsorbates, the differences in binding strengths and modes of adsorption between the glycine and the proline or hydroxyproline must be due to the presence of the extra carbon and hydrogen atoms (and hydroxy group), which increase the interactions with the surface. This strong interaction is important, as the rings of the proline and hydroxyproline in the collagen strand are positioned on the outside of the triple helix and hence are most accessible to calcium ions and phosphate groups nucleating at the template.43 (43) Prockop, D. J.; Kivirikko, K. I. Annu. Rev. Biochem. 1995, 64, 403–434.

Langmuir 2009, 25(9), 5018–5025

Almora-Barrios et al.

Article

Sorption of hydroxyproline is energetically the most favorable. The enhanced interaction of this molecule with the HA surfaces compared to that of proline is due to the extra hydroxy functional group. The hydrogen atom of the hydroxy group is flexible and thus can easily form interactions with surface oxygen atoms. In addition, the electrostatic interaction between this polar group and the surface increases the binding strength.

4.

Conclusions

In this computer modeling study, we have investigated the adsorption of the three major amino acids glycine, proline, and hydroxyproline, which are constituents of the collagen I protein, at two major hydroxyapatite surfaces, (0001) and (0110). Our simulations show that the strength of interaction of the amino acid molecules with the surfaces depends on both the stability of the surface and the capability of the amino acid molecules to form multiple interactions with the surface species, particularly if they can bridge between two or more surface calcium ions. The primary association between amino acids and the HA surfaces is via interactions between -COOand surface-bound Ca2+, but the additional side group interactions lead to significant variations in their affinities, consistent with experimental results.16,17,34 Our work shows a distinct preference by the three amino acids for binding to the (0110) surface, which is required if the bone mineral platelets are to become elongated in the c-direction and express the (0110) surface, rather than the (0001) surface, as is observed in the morphology of the natural bone platelets. The (0001) surface is the thermodynamically

Langmuir 2009, 25(9), 5018–5025

preferred surface morphology, but this surface is not expressed significantly in the natural bone tissue, which grows from a collagen matrix. All three amino acids adsorb strongly to the hydroxyapatite (0110) surface, which indicates that their functional groups should be good sites for the nucleation and growth of this surface at the collagen fibrils. Future work will include the derivation and optimization of interatomic potential parameters from the electronic structure calculations in this work, to model the interface between the collagen molecules and the apatite mineral. As the nucleation/ growth process obviously occurs in body fluid, we will also include water and salt ions in future calculations, which may form additional links between the peptide strands in the collagen helices and hence need to be included. However, such large-scale simulations are well beyond the scope of the ab initio techniques employed in this work, which have, however, provided the detailed chemical interactions between the surfaces and adsorbates. Acknowledgment. We acknowledge computational resources on the U.K. HPCx service, which were provided via our membership of the U.K.’s HPC Materials Chemistry Consortium and EPSRC Grant No. EP/D504872. N.A.-B. is grateful for a UCL Dorothy Hodgkin Overseas Research Studentship. Supporting Information Available: Models of hydroxyapatite surfaces, and geometry-optimized structures of the configurations of amino acids adsorbed at hydroxyapatite surfaces. This material is available free of charge via the Internet at http://pubs.acs.org.

DOI: 10.1021/la803842g

5025