Sergio E.
Ruiz Hernandez
*ab,
Ian
Streeter
c and
Nora H.
de Leeuw
ab
aDepartment of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, UK
bSchool of Chemistry, Cardiff University, Main Building, Park Place, Cardiff, CF10 3AT, UK. E-mail: RuizHernandezS@cardiff.ac.uk
cEMBL-European Bioinformatics Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SD, UK
First published on 31st July 2015
Classical molecular dynamics (MD) simulations have been employed to study the interaction of the saccharides glucuronic acid (GlcA) and N-acetylgalactosamine (GalNAc) with the (0001) and (010) surfaces of the mineral hydroxyapatite (HAP). GlcA and GalNAc are the two constituent monosaccharides of the glycosaminoglycan chondroitin sulfate, which is commonly found in bone and cartilage and has been implicated in the modulation of the hydroxyapatite biomineralization process. MD simulations of the mineral surfaces and the saccharides in the presence of solvent water allowed the calculation of the adsorption energies of the saccharides on the HAP surfaces. The calculations show that GalNAc interacts with HAP principally through the sulfate and the carbonyl of acetyl amine groups, whereas the GlcA interacts primarily through the carboxylate functional groups. The mode and strength of the interaction depends on the orientation of the saccharide with respect to the surface and the level of disruption of the layer of water competing with the saccharide for adsorption sites on the HAP surface, suggesting that chondroitin 4-sulfate binds to the layer of solvent water rather than to HAP.
Glycosaminoglycans (GAGs) are polysaccharide molecules that are abundant in mineralised tissues, and are known to interact strongly with apatite crystals.3–5 A GAG molecule is an unbranched polymer chain of two different types of monosaccharides linked together in alternation.6 The most prevalent GAG found in mineralised tissues is chondroitin sulfate, shown in Fig. 1(a), in which the alternating monosaccharides are glucuronic acid (GlcA) and N-acetylgalactosamine (GalNAc) shown in Fig. 1(b) and (c) respectively. This GAG can be sulfated to some extent on any of its hydroxy groups, and the most commonly sulfated position is the 4-hydroxy group of the GalNAc monosaccharide. GAGs, including chondroitin sulfate, are usually found as part of a larger macromolecular complex known as a proteoglycan, in which multiple GAG molecules are covalently attached to a central protein, and each GAG chain typically comprises 40–120 monosaccharide units.7
There are a number of ways in which the biomolecules secreted by cells can influence the mechanism of tissue mineralisation: either by providing a template for crystal nucleation, by adsorbing onto crystal surfaces and thus inhibiting crystal growth, or by sequestering dissolved mineral ions.8 A plethora of biomolecules have been implicated as either promoters or inhibitors of tissue mineralisation, including GAG-containing proteoglycans9 and the collagen,10 alkaline phosphatase11 and SIBLING proteins.12 It is likely, however, that mineralisation is controlled by many different organic molecules simultaneously, with exquisite control of the process effected by the cell's use of system feedback loops.
It has been found that under in vitro conditions of limited calcium availability, GAGs, including chondroitin sulfate, and their proteoglycans inhibit the deposition of hydroxyapatite, perhaps because the adsorption of GAG molecules on the mineral surfaces is an obstacle to the mechanism of crystal growth.3–5 The role of proteoglycans in the mechanism of apatite formation in vivo is more difficult to interpret, because proteoglycans and GAGs have multiple functions in physiological tissues, and interact with many other biomolecules and with cells.8,9,13–15 Studies in vivo and in cell culture have often produced conflicting results, with proteoglycans observed either to promote or to inhibit apatite formation.16 Evidence for the importance of the GAG–apatite interface in vivo has come from solid-state NMR experiments of bones,17,18 teeth,19 plaque20 and kidney stones:21 in all cases 31P–13C coupling showed that the apatite mineral was predominantly in contact with GAG molecules rather than with any other organic component.
In this paper we have used computational methods to investigate the interfacial interactions between GAGs and HAP. We present molecular dynamics (MD) simulations of chondroitin sulfate saccharides adsorbed onto a slab of HAP in the presence of interfacial water. Adsorption structures have been calculated for the (0001) surface of HAP, which is the thermodynamically most stable surface, and the (010) surface, which is the dominant plane of apatite nanocrystals in a biological environment.22 We have simulated the adsorption of the monosaccharides GlcA and GalNAc 4-sulfate (GalNAc-4S) independently, rather than longer chondroitin sulfate chains, because this reduces the number of possible HAP–saccharide interactions in each simulation, and simplifies the interpretation of the results. The simulated monosaccharides were methylated in two different positions, as shown in Fig. 1, in order to the reproduce the effect of their position within a long polymer GAG chain.
Classical MD simulations have been used previously to investigate the adsorption of small organic molecules at an apatite–water interface.23,24 For these molecules it was reported that the principal adsorption interactions were the electrostatic attractions between surface calcium ions and the adsorbate oxygen and nitrogen atoms, and that an important secondary interaction was hydrogen-bonding from adsorbate polar hydrogen atoms to the surface phosphate oxygen atoms. Recently we have used DFT calculations to predict the adsorption structures of GlcA and GalNAc (not sulfated) on HAP in vacuo.25 It was found that GalNAc interacts with HAP principally through its hydroxy and acetyl amine functional groups, and that GlcA interacts principally through its hydroxy and carboxylate functional groups. However, the inclusion of interfacial water in the present study will give us a better understanding of adsorption processes under more realistic physiological conditions, rather than in vacuo.
Following conventional MD simulation procedures, the system was described by a simulation cell with three-dimensional periodic boundary conditions, containing a slab of HAP with a surface on either side. The thickness of the slab was greater than 20 Å for each surface studied, and the slab was separated by a gap of approximately 53 Å from its images in the next cell. Under these conditions the system was run for 1200 ps with 400 ps of equilibration. Then the gap between the HAP slabs was filled with water molecules, leading to a simulation cell containing approximately 3000 atoms in total. Now the simulation times were increased to 1600 ps with 400 ps of equilibration. The Shake algorithm was also employed to constrain intramolecular hydrogen-bonds.
The mineral slab was described by the interatomic potential model developed by de Leeuw for modelling apatite crystals.31 In this forcefield, phosphate and hydroxy group bonds are parameterised as the sum of a Morse potential and a coulombic potential, phosphate bond angles by a harmonic potential, and non-bonded interactions by Buckingham potentials (see Table S1, ESI†). This forcefield makes use of a shell model,32 in which each oxygen anion in the phosphate and hydroxy groups consists of both a core and a massless shell connected by a spring, in order to model the atom's electronic polarisability. The forcefield has been validated by comparison with DFT calculations,31 and produces consistent defect and dissolution energies.33–35
HAP–water interactions and HAP–GAG interactions were parameterised using non-bonded potentials derived in previous MD studies,23,24 which are compatible with the current HAP shell model but do not consider pH effects (see Section 2.2). Specifically, the interfacial interactions were described by Buckingham potentials and Lennard-Jones potentials, and the repulsive portion of the potential was scaled for consistency with the charges on the adsorbate atoms, using a method originally described by Schröder et al.,36 and subsequently used in a number of surface adsorption studies.23,24,31,37,38
![]() | ||
Fig. 2 (a) Hydroxyapatite unit cell. (b) View onto (0001) surface, showing the hexagonal channels: Ca green, P purple, O red, H white. CaI atoms are columnar calcium and CaII atoms are axial calcium. |
The crystallographic structure of HAP contains columns of hydroxy groups in the hexagonal channels that are oriented parallel to the c axis. The OH groups are oriented either hydrogen "up" or hydrogen "down" in the channels, where all "up" and "down" OH− positions have occupancy of 0.5.1 In an MD simulation the occupancy of a site is necessarily restricted to either unity or zero, so in our simulated slabs of HAP the hydroxy groups were placed in only half of the possible positions, such that within each column they all pointed in the same direction. The unit cell was then doubled in the b direction, so that neighbouring columns alternated in the orientation of the hydroxy groups. Previous DFT calculations have identified this arrangement of hydroxy groups as the lowest energy configuration of bulk HAP,39 and it is similar to the synthetic monoclinic HAP material, where all hydroxy positions are fully ordered in a similar configuration as modelled here.
In calcium deficient environments (i.e. bones, teeth), the apatite obtained is nonstoichiometric, whereas the surface phosphate ions are usually protonated,40–45 with the protonation rate of the ions typically ranging from PO3–OH at pH ∼10.5 to PO2–(OH)2 for pH ∼5.5.46
We have used three different HAP slabs for the MD simulations of the adsorption of GAGs, which are shown in Fig. 3. We have considered the low index (0001) and (010) crystal surfaces, which are both present in the experimental morphology of HAP, where for the (01
0) surface we have considered two different possible terminating planes.
The (010) termination shown in Fig. 3b contains columnar Ca atoms in the surface, with the hydroxy channels remaining intact below the surface, whereas the terminating plane in Fig. 3c cuts through the hydroxy channels and leaves the columnar Ca atoms below the surface. The surface in Fig. 3b has been identified previously as the most stable (01
0) termination, using computational methods,47 but the surface in Fig. 3c has been identified as the principal (01
0) surface of fluorapatite using high-resolution specular X-ray reflectivity data.48 We note that both of these possible terminations are likely to be present transiently as a crystal grows in the (01
0) direction.
These three stoichiometric surface structures allow us to compare our results with previous theoretical results23,24,38,49 while also keeping the Ca/PO4 ratio, which is important to investigate experimental findings regarding the high Ca2+ binding capacity of chondroitin 4-sulfate.50
For each monosaccharide and for each HAP slab, we also simulated the system where the saccharide was fully solvated, i.e. the saccharide was situated in the solution midway between the upper and lower surfaces of two slabs from neighbouring repeat cells. The adsorption energy of a saccharide on a particular surface, Eads, is given by:
Eads = Esys,ads − Esys,solv | (1) |
The hydration energy of the surfaces is given by:
Ehyd = (Esurf,solv − Esurf + nEH2O)/A | (2) |
On the (010) surface it can form bridges between two calcium ions, all at distances similar to those shown in Fig. 5b. The oxygen of the carbonyl group is also often oriented towards surface calcium (Fig. 5a and b).
In the case of the crenellated (010) plane, the sulfate group is initially located in the “trench” where the hydroxy group is also interacting with the surface. In all configurations the molecule lies flat on the surface. On the (0001) surface, the hydroxy group is closest to the surface (Fig. 5c) and the sulfate is only coordinating more than one calcium ion in two configurations, whereas the molecule lies flat on the surface in four of the configurations.
GlcA remains in the low-energy chair conformation (Fig. S4b, ESI†), in the presence of solvent. The carboxylate hydroxy and methoxy groups are all in equatorial positions. Intramolecular non-bonded interactions are absent: the polar functional groups generally interact with the surrounding water molecules, rather than with each other. The carboxylate group is negatively charged, and it therefore attracts the water molecule's hydrogen atoms. There are always typically 1–2 water molecules within 2.5 Å of the carboxylate oxygen and all other polar functional groups are also observed to form hydrogen bonds to water.
The (0001) surface is the thermodynamically most stable surface in both fluorapatite38 and HAP,23,25 with columns of calcium atoms and hydroxy groups running perpendicular to the surface. The water molecules interact with surface calcium ions via their oxygen atoms at distances from 2.28 Å to 2.61 Å, in line with the 2.28 Å reported by the experimental work by Pareek et al.54 and the 2.29 Å and 2.44 Å obtained by DFT calculations by Corno et al.55 The water molecules (red stars in Fig. 5a) also interact with surface hydroxy and phosphate oxygen atoms via their hydrogen atoms, with almost all the phosphate ions involved in the interactions at distances of less than 2 Å, in a good agreement with the 1.8 Å found by Pareek et al.54 and the 1.5 Å by Corno et al.55 The water molecules closest to the surface do not form a regular adsorption pattern on the surface (Fig. 6b), and they are barely exchanging with the bulk solution (∼1.7 s−9). This surface has a smaller surface area than the (010) and the hydration energy is −1.36 J m−2, indicating that the water is more strongly attracted to this surface than to either termination of the (01
0) surface, which agrees with the much lower frequency of water exchange. The gap (∼1.5 Å in Fig. 6a) between the first and second layers of water molecules on the (0001) surface is due to the stronger interactions with surface ions and also the existence of an extensive network of hydrogen-bonded interactions between the water molecules within the layer nearest the surface, at the expense of interactions with the water molecules in the second layer. Although a more stable surface may be expected to be less reactive towards water, when the geometry and composition of the surface allows strong interactions between water and surface, in addition to extensive hydrogen-bonding between the molecules within the water layer, the formation of a regular water layer is often found to be an especially favourable process. A similar situation was observed in previous work on the α-quartz (0001) surface,56 but in that case the gap was even more pronounced (∼2.5 Å) due the highly ordered nature of the first monolayer of water.
GlcA | |||
---|---|---|---|
Configuration | E ads/kJ mol−1 | ||
(0001) | Flat (01![]() |
Crenellated (01![]() |
|
1 | 37.6 | 23.5 | 22.7 |
2 | −34.7 | 12.2 | 24.9 |
3 | −55.7 | 30.6 | 13.9 |
4 | 52.2 | 30.2 | 28.7 |
5 | 17.7 | 64.9 | 61.0 |
Table 1 shows that GlcA adsorbs favourably onto the (0001) surface for two configurations (2 and 3). The adsorption energies on the (010) surfaces are positive for all configurations, suggesting unfavourable adsorption of the saccharide on to these surfaces. Although the values of adsorption energy are less positive in the case of the flat terminated (01
0) surface than for the crenellated termination and for the (0001) surface. In all cases the GAG and the solvent compete for the adsorption sites on the surfaces and generally the saccharides interact with the mineral surfaces via a monolayer of interfacial water, but the few exceptions to this general behaviour, corresponding to the configurations with the least positive or negative adsorption energies in Table 1 are discussed in the sections below.
The GAG has undergone a conformational change and the carboxylate group is now in the axial position, not in the most energetically favoured equatorial position. The GlcA forms hydrogen-bonds with the surface in all five configurations studied (Fig. S8, ESI†), but these interactions by themselves are not enough to stabilize the adsorption of the saccharide.
A combination of a strong and long-lasting interaction through the carboxylate group, the formation of hydrogen-bonds and the relative position of the molecule with respect to the surface plane (discussed further in Section 3.9) are the main factors contributing to the relatively small positive energy of configuration (2).
It is noted that configuration 3 does not form hydrogen-bonds to the surface as found for the most strongly bound configuration on the flat (010) surface. GlcA in configuration 3 is adsorbed on the crenellated surface, with the molecule pinned to the surface in a vertical position in which the hydroxy groups are more exposed to the solution, hence interacting more with the water molecules, which also stabilises this configuration.
When the organic molecule is lying along the trench in the surface, all functional groups are able to interact strongly with the surface atoms. The interactions between the oxygen atoms of the saccharide functional groups and the calcium ions help to lower some of the positive adsorption energies compared to the flat (010) surface.
The intensity of the RDF peak (Fig. S11, ESI†) is less than for the (010) surfaces, indicating a weaker interaction, because the oxalate group is not able to form bridges with the calcium ions. For all the configurations in which GlcA remains in the first layer of adsorption, it forms hydrogen-bonds with the surface.
As occurs on the (010) surfaces, the combination of the various interactions lowers the positive adsorption energies, reaching negative values with a change in conformation of the organic molecule. However, when the GAG remains in its most stable conformation in solution the adsorption is even weaker than on the (01
0) surfaces, for two reasons: The interactions between surface and carboxylate are not particularly strong, because it cannot interact with more than one calcium ion at a time, and the relative position of the organic molecule with respect to the surface in certain cases removes the solvent from the monolayer, as is discussed in more detail in Section 3.9.
GalNAc-4S | |||
---|---|---|---|
Configuration | E ads/kJ mol−1 | ||
(0001) | Flat (01![]() |
Crenellated (01![]() |
|
1 | 113.4 | — | 59.2 |
2 | — | 4.8 | 85.0 |
3 | 41.8 | 154.9 | 142.7 |
4 | 50.7 | 27.1 | 75.8 |
5 | — | 60.0 | 84.2 |
The values in Table 2 indicate the adsorption of the saccharide on to the three surfaces covered in this study is thermodynamically unfavourable. For some of the configurations analysed, the molecule experiments complete desorption from the material. The adsorption energies are generally higher than those obtained for GlcA. The GalNac-4S also prefers to interact with the material through the monolayer of water and as it occurs the case of GlcA the least positive values of adsorption energies are for the (010) surface with the flat termination.
In the sections below we will discuss the geometries leading to the least unfavourable adsorptions on each surface.
In configuration 1 (Table 2), the organic molecule desorbs from the surface. The other configurations exhibit interactions between the sulfate and the surface, where in all cases the sulfate, and in one case the hydroxy group, are the only groups of the GAG within the first adsorption layer. The configuration with the least positive adsorption energy (Fig. 10) is the only configuration forming hydrogen-bonds with the surface (Fig. S12, ESI†), which as we described in Section 3.7.1 for GlcA is a determinant factor to increase the strength of binding. Configuration 2 in Fig. 10 is showing a very similar adsorption to the one obtained for GlcA (Fig. 7). As no other functional group of GalNAC-4S is interacting with the surface, the stabilization of the adsorption of GalNAC-4S compared with the GlcA is due to the nature of the sulfate, the adsorption position of the molecule and the more frequent occurrence of hydrogen-bonds, between the hydroxy group of the organic molecule and the phosphate ions, shown in the RDF plot in Fig. S12 (ESI†). The intensity of the first peak is higher than the one obtained for a similar configuration of adsorbed GlcA (Fig. S8, ESI†), even though GalNAc-4S only has one hydroxy group while there are two in GlcA. However, we suggest that hydrogen-bonding is more transient, breaking and re-forming frequently during the simulation.
Configuration 1 (Fig. 11) exhibits the strongest interaction between surface and adsorbate, with the sulfate group as the only functional group inside the trench, and an additional interaction between the carbonyl oxygen and the calcium ion of the surface. With the main part of the molecule outside the trench, no formation of hydrogen-bonds with the surface is observed (Fig. S13, ESI†).
As the GalNAc-4S molecule has a single hydroxy group, the presence of peaks in Fig. S13 (ESI†), also supports the fact that in the other configurations the organic molecule resides further into the trench of the surface, to allow the hydroxy group to interact with surface phosphate oxygens. However, here the several interactions occurring at the same time do not produce the same effects observed for GlcA on this surface.
The presence of the sulfate and the acetylamine groups in the molecule makes it a bigger system compared to GlcA, and the diffusion of the GalNAc-4S in the trench causes more rotations of the phosphate groups, and hence less coordination between Ca2+–Ophosphate.
The distance between calcium ions is decreased in comparison with the pure surface, causing more strain in the surface and further instabilities. A combination of these effects due to the presence of the molecule, and the loss of interactions with the solvent (discussed in Section 3.9), causes the relative instabilities of these systems.
Configuration 3 is another example of the importance of the position of the adsorbate on the surface and the formation of hydrogen-bonds between the organic and the surface, but the interactions between the surface and the sulfate are transient during the entire simulation time.
A common issue affecting the stabilization of both saccharides at the three hydroxyapatite surfaces is the interaction of the organic molecule with the solvent molecules near the surface, which are identified in Section 3.6 and represented in dark blue in Fig. 6b–d.
A similar situation was reported in previous studies of the behavior of water at quartz surfaces, when a desorbed Si(OH)4 molecule remained near the quartz/water interface56 in a process calculated to be endothermic. During adsorption on the flat (010) and (0001) surfaces both saccharides generally displace ∼10% of the water molecules closest to the surface, which value rises (up to 20% for the (0001) surface), if the molecule lies flat on the surface. In the case of the crenellated (01
0) surface, both GAGs affect significantly the order of the water adsorbed in the bottom of the trench. The displacement of the water molecules in the first hydration layer prevents the formation of hydrogen-bonds between the water and the phosphate, and there are also fewer interactions between Ca2+ and oxygen of the water, hence the system becomes less stable than the hydrated surface in the absence of GAGs. The strongest binding of GAGs is obtained for the systems in which the first hydration layer is less affected by the adsorbate, confirming that the interactions between calcium ions and water are very important, as was also shown in a previous study.24
To overcome the effect of the loss of water at the surface, the saccharide has to mimic the water interaction in the same area, in a similar way as described in Section 3.7.3 for configuration 3 of GlcA on the (0001) surface (Fig. 8), where less than the typical amount of water is displaced by the adsorbate and where interactions between the oxygen of the hydroxy group and the surface calcium ions replace the Ca2+–Ow interaction of the lost water (see Fig. S14 and S15 in the ESI†). The hydroxy groups of the saccharide also coordinate to the remaining water molecules closest to the surface. As the (0001) surface is also the surface with the largest value for its hydration energy, and is the one where the water monolayer is most pronounced (Fig. 5a), the competition between GAG and water for the adsorption sites on the surface will have the highest impact on GAG adsorption energies. The crenellated (010) surface, where the water molecules are highly ordered (Fig. 6d), which ordering is broken after adsorption, will be the second surface most affected by the competition mentioned above.
Recent simulations of the adsorption of peptides considering pH effects46 report considerable changes of adsorption sites and geometries as a result of change in the pH of the solvent, leading to a change in adsorption energies with more negative values at lower pH. In view of those findings and suggestions by Posner and Park40,48 regarding the influence of phosphate protonation on the structure and binding strength of the water monolayer, we can speculate that at the typical pH of ∼6–7.4 in the kidneys,57 the changes in the structure of the water monolayer could lessen the competition between the organic and the water in certain adsorption sites, thereby increasing the binding probabilities for the GAGs. Some proteoglycans for example, have shown an increase in affinity at low pH.50
Previous experimental work established that in general, there is a direct relationship between the ability of a macromolecule to bind to HAP and its ability to block crystal growth60 where the interaction occurs mainly through surface calcium ions,61 demonstrating that the calcium binding capacity of GAGs is a significant factor in their ability to inhibit HAP growth.50 In the present study we have shown that only those monosaccharide functional groups able to bind to Ca2+ and causing less perturbation of the surface hydration layer are viable for adsorption. Sulfate and carboxylate groups are the desirable functional groups for adsorption to take place, with the sulfate group showing the greater affinity for calcium, also in agreement with experimental findings demonstrating the importance of the sulfate group for the adsorption capacities of chondroitin 4-sulfate.62
The values of the adsorption energies obtained in this work support experimental reports, which considered that proteoglycans and GAGs with more sulfate groups are better adsorbates to HAP than chondroitin 4-sulfate.3,5 The present work suggests that further stabilization of the surface–saccharide complex compared to the purely hydrated surface is due to hydroxy groups in the organic molecule replacing the missing water–water and water−surface hydrogen-bonds, while other stronger interactions also take place, e.g. between the carboxylate and sulfate functional groups and the surface.
The GlcA displaces fewer surface water molecules and maintains multiple surface−organic interactions for longer periods of time on the (0001) than on the flat (010) surface, leading to stronger binding on the (0001), but it is also possible that the kinetic barrier to desorption is too large for this to be observed on the timescale of the simulations. The interactions of the saccharide with the surfaces are within the energy penalty of a conformational change in the adsorbate, while the GalNAc-4S is not able to sustain long lasting interactions with (0001), but its adsorptions onto the flat (01
0) surface show very small positive values.
However if the saccharides are forming chondroitin 4-sulfate, the conformational changes observed in the monomers, specifically those stabilizing the interactions with (0001), are energetically forbidden by the presence of the glycosidic linkage and the length of the chain. Our simulations suggest that amino acids and GAGs behave differently when adsorbing onto HAP surfaces in the presence of solvent. The amino acids or peptides manage to displace the water and bind favourably to the surfaces,24,53 whereas the trajectories and values of the adsorption energies obtained here for the GAGs support previous experimental findings which show that disaccharides present in proteoglycans do not bind to HAP for long enough to affect crystal growth.63 If water is present, chondroitin sulfate binds to the layer of water rather than to HAP.58 Our results also support suggestions from experiments at pH > 10.5, that GAGs act as templates for HAP formation, instead of moderators of crystal growth by binding to the surfaces.58 The same experimental work showed that when HAP was formed in situ at pH ∼7−8 in the presence of the GAG, the polysaccharide was more closely attached to the surface, confirming the importance of the pH and therefore the inclusion of both solvent and surface protonation in future simulations to obtain more accurate adsorption energies.
Footnote |
† Electronic supplementary information (ESI) available: Potential parameters used in this work. Adsorption positions of GlcA and GalNAc-4s on to the dry surfaces. RDF graphs. See DOI: 10.1039/c5cp02630j |
This journal is © the Owner Societies 2015 |