Inhibition of ice crystal growth by preferential peptide adsorption: a molecular modelling study

摘要

Faraday Discuss., 1993,95,299-306 Inhibition of Ice Crystal Growth by Preferential Peptide Adsorption: A Molecular Modelling Study M. La1 Unilever Research, Port Sunlight Laboratory, Bebington, Wirral, UK A. H. Clark and A. Lips Unilever Research, Colworth Laboratory, Sharnbrook, Bedfordshire, UK J. N. Ruddock and D. N. J. White Department of Chemistry, The University, Glasgow, UK A molecular modelling approach has been used to study the relative binding of a winter flounder antifreeze peptice (AFP) to the faces and the internal planes, in particular the c-face and the 2021 plane, of the ice (Ih) crystal. The in vacm binding energies of the peptide molecule with the two surfaces differ by a factor of more than three, which could account for the suggested higher affinity of the molecule for the 2021 plane than for the c-face or the prism face.The propensities of the molecule to hydrogen bond with these surfaces are not so different. The key difference in the mode of attachment of the molecule to the various surfaces appears to stem from their geometrical features: the c-face and the prism face are flat, precluding intimate physical contact to be made with the peptide, but the 2021 plane contains lsquo;ridgesrsquo; and lsquo;valleysrsquo; which produce a near-perfect steric match for the structure of the peptide molecule, so accommodating it in the lsquo;lock and keyrsquo; fashion. A notable example of the role of crystal growth inhibition in biological processes is furnished by the action of certain peptides and glycopeptides in controlling the size of ice crystallites in the body fluids of polar fishes.rsquo; It is believed that this action (commonly known as the anti freeze effect) is largely responsible for the survival of the fishes at sub-zero temperatures.* Also, many other proteins are known to exhibit this effect, controlling the growth of cholesterol crystals, enamel, kidney stones and bone minerals among other important biocrystalline materials.Indeed, the wide variety of biomineralisation processes occurring in nature bears testimony to the subtle and profound effects which biomolecules can produce in the control and engineering of crystal struct~res.~3~ A good deal of work has been reported in the literature directed towards gaining a fundamental understanding of the function of both biological and non-biological molecules in bringing about the foregoing effect^.^ The antifreeze effect of peptides and glycopeptides has been studied experimentally by DeVries and co-workers amongst others.rsquo;j-12 They pointed out that in molar terms the concentrations of the active peptides in the fish blood are much too low to produce the usual colligative freezing-point depression effect to a significant degree.Thus it appears that the antifreeze effect is not a result of the lowering of the melting point of ice by the peptide acting as solutes, but arises from the specific adsorption of the active molecules onto the ice crystal faces. Harrison et al. showed that between the normal freezing point of ice and the threshold supercooling temperature (ca.-2.2ldquo;C)the active peptides stop the crystal growth, the antifreeze effect diminishing rapidly below the threshold temperature. l3 The antifreeze activity, measured in terms of the 299 Inhibition of the Ice Crystal Growth temperature range (A T)of the supercooled state, has been found to depend appreciably on the peptide concentration. Conclusions drawn from the various experimental studies on how the peptides stop the crystal growth are at best plausible. So far no concrete evidence has been presented which establishes the preference of the peptide to be adsorbed onto certain ice faces. Clearly, a fresh approach is called for to address the mechanistic questions of the antifreeze effect at molecular level.We have initiated a molecular modelling study of the interaction of the active and inactive peptide molecules with ice in an attempt to identify such features of the molecular structure and interactions as are responsible for the effect in question. In order to achieve our objective we must first investigate the differences in the propensities of the active and inactive molecules to be adsorbed onto some familiar ice crystal faces and internal planes. In this way it will be possible to assess, in terms of molecule considerations, the validity of the inference derived by Knight et a1.14 that when it is being adsorbed onto an ice crystal the active molecule shows a strong preference for the 2021 plane of ice Ih. Molecular Modelling Approach The effectiveness of the computer simulation approach in the treatment of realistic models for interacting molecular systems was first recognised in the fifties.I5 In subsequent years the simulation methods, both Monte Carlo and molecular dynamics, have led to rapid progress in the development of a comprehensive understanding of the role of intermolecular interactions in determining the structural and dynamic behaviour of materials.l6l7 The spectacular achievements made in this field owe much to the advent of supercomputers, allowing the simulation of systems involving complex interactions. More recent times have witnessed widespread accessibility of computer power occasioned by the development of high-performance workstations which, in terms of both cost and convenience, offer at least an order of magnitude increase in computational efficiency over the older main-frame machines.A significant outcome of these new developments in hardware technology is that simulation/modelling methods have generated extensive interest among researchers from a wide range of areas including drug design, molecular biophysics and materials science. Pursuit of these new applications has been greatly facilitated by the emergence of user-friendly molecular modelling packages developed by academics as well as by commercial organisations. Among the key issues which have attracted most attention in current molecular modelling studies, the relationship of the functional specificity of a molecule to its detailed molecular structure is paramount. The problem under our present consideration falls within the domain of this issue.Software A most important feature of good molecular modelling software would reside in the facility it would offer for calculating accurately both intramolecular and intermolecular interaction energies arising from interatomic forces, bond torsions and from variations in the bond lengths and bond angles. It is, therefore, essential that the software uses reliable force field parameters validated against known properties of various calibration systems. Other parameters characterising the molecule, which are of no less importance, are the mean values of bond lengths and bond angles. A significant contribution to the interactions derives from the partial charges assumed to be located on the atoms within the molecule.Approximate values of these charges can be readily calculated using suitable quantum- mechanical approaches; the corresponding software is incorporated in most packages. The software, COMMET, employed in the present work was developed by White and co-workers at Glasgow University. This software has proved highly successful in the conformational analysis of complex biological and other organic m01ecules.~*J~ Using M. La1 et al. 30 1 known geometrical and interaction parameters corresponding to bond lengths, bond angles, interatomic van der Waals (dispersion and ionic) interactions and bond rotational energy, the software will build desired molecular structures through a menu-driven operation. The dispersion interactions are expressed in terms of the familiar Lennard-Jones (L-J) potential, bond-length and bond-angle vibrational energies in terms of the simple harmonic potential and bond rotational energies in terms of Fourier series.Values of the various parameters involved have been arrived at through extensive optimisation procedures carried out over a period spanning some 15 years. Starting from molecular structures thus built, a range of computations including Monte Carlo, molecular dynamics, sequential iterative torsional angle refinement (SITAR), energy minimisation and docking can be performed conveniently. Molecular graphics facilities incorporated in the software permits lsquo;liversquo; visualisation of the configurational changes as MD computations progress in time.Static configurations can be examined in detail by activating the rotation routines which allow the molecule to be rotated about the x-, y-and z-axes. Hardware Parallel computers offer a highly effective means of performing complex modelling calculations on realistic timescales.20 Parallelisation of the COMMET code makes possible the running of this software on such computers. The present calculations were carried out on a 32-node parallel machine with a central processing unit of four interconnected arrays of eight Transputers (T800) each. The Present Systems Antifreeze Active Peptides Many antifreeze peptides contained in the body fluids of various polar fishes have been isolated and attempts have been made to determine their molecular structure.It has been shown from X-ray and circular dichroism studies that the molecules of most of the active peptides are helical and have an overall rod-like shape.21+22 For our present study we have selected a commonly known AFP found in winter flounder. Consisting of 37 residues, this molecule is predominantly hydrophobic (alanine-rich) with the following primary structure:21 Asp-Thr-Ala-Ser-Asp-Ala-Ala-Ala-Ala-Ala-Ala--Leu-Thr- Ala- Ala- Asn- Ala-Lys- Ala- Ala- Ala-Glu--Leu-Thr- Ala- Ala- Asn- Ala- Ala- Ala-Ala-Ala-Ala--Ala-Thr- Ala- Arg It exists as a single ~t-helix.~~ A conspicuous feature of the structure is a salt bridge between the lysine and the glutamic acid residues which are separated from each other by a sequence of three alanine residues.It is believed that the salt bridge contributes to the stability of the helical structure.24 The helix is ca. 50 A long and has a diameter of ca. 10 A. A molecular model of this peptide depicting gross features of its structure is shown in Plate 1. Ice Crystal Of the several polymorphic forms in which ice is known to exist, ice Ih is of most common occurrence under ordinary condition^.^^ The structure of ice has been investigated extensively using a variety of experimental techniques including X-ray and neutron diffraction, NMR and other spectroscopic techniques. The crystal structure of ice Ih is hexagonal in which each oxygen is hydrogen-bonded to four neighbouring oxygen atoms.The lattice unit is characterised in terms of four axes, a,, a2,a3and c. The lattice consists of puckered layers normal to the c-axis, which contain hexagonal rings of the H20molecules Inhibition of the Ice Crystal Growth prism face (1070)1201 12101 12 1101 Fig. 1 Schematic representation of the hexagonal unit of the ice crystal structure. The shaded area is a c-face; the hatched area is a prism face. in the lsquo;chairrsquo; conformation. The hexagonal rings formed by three molecules of one layer and three of its adjacent layer exist in a lsquo;boatrsquo; conformation. The surface of the hexagonal unit, shown schematically in Fig. 1, comprises eight faces, two basal faces normal to the c-axis and six prism faces.Since it is normal to the c-axis, the basal face is known as a c-face (0001). In addition to the external faces, many internal planes in the structure are also of interest in the adsorption studies on ice. These will be discussed in the following section. Molecular Modelling Computations Generation of the Ice Crystal A unit cell was constructed using the data on fractional coordinates obtained from a neutron diffraction study reported by Peterson and Levy.26 The unit cell thus generated was replicated along all three directions to build the required size of the crystal, typically 20 x 20 x 10 unit cells. To expose a given face or an internal plane, the excess of water molecules was removed from appropriate positions in the crystal.Calculations of the energy of interaction between the adsorbate molecule and a crystal plane included the interaction with all water molecules located in the crystal up to a depth of five layers below the plane. The Peptide The model of the AFP molecule was generated using the COMMET lsquo;peptide buildrsquo; option. The sequence was folded into an a-helix by employing a COMMET routine. The carboxyl group of the glutamic acid and the amino group of the lycine residue, separated by a sequence of three alanine residues, fell into place to form the salt bridge. The salt bridge in the helix is clearly visible in the pictures of the model in Plate 1. Docking The docking routine allows the molecule to move towards the target (crystal face or plane in the present case) where it can dock on an active site according to energy and/or geometric M.Lal et al. 303 requirements. In the present study, the optimum docking position was searched using an energy-minimisation procedure based on the block-diagonal Newton-Raphson method. Short minimisations of 100 iterations were performed periodically as the molecule moved over the ice surface. Progress towards identification of the minimum-energy position of the molecule at the surface was monitored by examining the positional variations of the energy. The lsquo;finalrsquo; minimum-energy position was subjected to a further 500 minimisation iterations to ensure that the molecule assumed a position at the crystal face which truly corresponded to a minimum (or very close to it) in the configurational phase space of the system.Positions of the atoms both in the crystal and in the peptide molecule were allowed to relax to expedite the energy-minimisation process. Results and Discussion n view of a substantial number of residues in the active peptide with propensity to hydrogen bond with water molecules, it has been postulated in the literature that the affinity of the peptide for the ice crystal derives essentially from the spatial compatibility of the H-bonding residues in the chain with the repeat distance of the oxygen atoms on the crystal ~urface.~Thus the maximum binding strength would correspond with a situation where all the H-bonding hydrogens in the residues are located above the oxygen atoms in the crystal surface (and vice versa)at a distance optimum for the hydrogen-bond formation.However, it is possible that other contributory factors to the binding strength, such as dispersion and charge-charge interactions, would play an important part in determining the affinity of the adsorbate molecule for the crystal. In our calculations we have included all such interactions and, therefore, should be able to establish how significant non-H-bonding considerations are in the peptide/crystal binding mechanism. Calculation of the interaction of the molecule with the external faces of the ice crystal is a useful starting point in our study. The c-Face and the Prism Face The c-face is characterized by a hexagonal arrangement of the oxygen atoms on it with a repeat distance of 4.5 A.A similar repeat distance is also found on the prism face. Another common feature of the two faces is that they are flat. Thus the mode of interaction of the peptide with these two faces would be quite similar. The minimum-energy orientation of the molecule on the face, obtained through our docking procedure, is shown in Plate 2 in three perspectives, front, top and side. The mode of contact of the molecule with the face is clearly revealed in the front view, which shows that intimate contact is formed at only four points, the bulk of the molecule residing some distance away from the face. The points of contact are identified with the hydrogen-bond-forming sites. The calculated value of the energy of interaction is -26 kcal mol-l which is due largely to the four hydrogen bonds between the molecule and the face.Our preliminary work on the docking and interaction energy calculation of the AFP on the prism face gives a similar picture. Furthermore, the symmetry of the two faces is such that the molecule is able to achieve minimum interaction energy in several orientational states on these faces, which would entail lack of intermolecular cooperativity in the adsorption process. The 2021 Plane Among the numerous internal planes in the crystal, the a-plane (normal to an a-axis) and the 2021 plane have attracted much attention due to their possible involvement in the antifreeze mechanism.14 The location of the two planes in the crystal are shown in Fig.2. It is believed that the antifreeze activity of certain peptides found in sculpin is due mainly to their adsorption on the a-plane exposed in the growing crystal. As already mentioned, experiments of Knight et al. point to the significance of the 2021 plane in respect Inhibition of the Ice Crystal Growth Fig. 2 2021 plane and an a-plane in the hexagonal unit cell of ice Ih of the antifreeze activity of the winter flounder peptides.14 It is clear from Plate 3 that unlike the c-face and the prism face, the 2021 plane is not flat but is characterised by lsquo;ridgesrsquo; and lsquo;valleysrsquo; occurring alternately in parallel orientation. The minimum-energy state of the peptide molecule on this plane, achieved through the docking procedure, is shown in Plate 4.The contrast in the modes of contact of the AFP molecule with the c-face and the 2021 plane is apparent in the front views of the docked molecule shown in Plates 2 and 4. On the 2021 plane, sections of the molecule are accommodated neatly into the lsquo;valleysrsquo;, leading to a close fit between the structure of the molecule and the geometrical texture of the plane. We find that the minimum-energy state does not correspond to the molecule confined to a single lsquo;valleyrsquo; in the plane, but it spans several lsquo;ridgesrsquo; and lsquo;valleysrsquo; at an angle of ca 60rdquo;. The calculated energy of interaction is equal to -84 kcal mol- I, more than three times the value for the adsorption on the c-face. This clearly establishes the preference of the molecule for this plane over the c-face and prism faces.The 60rdquo;orientation of the molecule on the plane is unique in the minimum-energy state, implying that the adsorbed molecules would lie parallel or anti-parallel to each other, in this way maximising the intermolecular interaction and so producing the cooperative effect in the adsorption process. One may anticipate that such an effect would further enhance the preference of the peptide for this particular plane. The large difference between the 2071-AFP and the c-face-AFP attractive energies is due essentially to the very close attachment of the molecule to the 2021 plane, the result of a near-perfect fit of the molecule into the contours of this plane. Such a close proximity of the molecule to the adsorbent plane produces a several-fold increase in the van der Waals component of the interaction energy, thereby leading to an energy of adsorption that is much greater than on the c-face where, as noted, the molecule is unable to attain steric intimacy due to the flatness of the face.It has been pointed out by Knight et al. that there exists a close correspondence between the oxygen-oxygen periodicity (16.7 A) on the 2021 plane and the distance between the successive threonine residues on the helix (16.5 A).14The near-collinearity of the threonines will facilitate their interaction with the corresponding oxygens on the plane. Our molecular modelling work confirms that the steric compatibility of the molecule with the plane stems from this correspondence. The essential picture emerging from our study is that the threonine residues are positioned over the oxygen atoms at the top of the lsquo;ridgesrsquo; with the sequences of the residues between threonines slotting neatly into the lsquo;valleysrsquo;. The adsorption-induced inhibition of crystal growth would be most effective if the adsorbed molecules impede the growth of the fastest-growing face of the crystal.This originally led DeVries and co-workers to propose that an AF would function most efficiently if it was adsorbed on the prism face which constitutes a step between two successive c-faces.rsquo; This mode of binding was suggested to lead to residual growth in the c- M. La1 et al. 305 direction with reduced area of this face (i.e. development of specules, as is observed experimentally).Some experimental studies offer tentative support for this mechanism.rsquo; Chakrabartty et al. have measured the rates of growth of ice crystals both in the c and a directions at various AFP concentrations.Il They found that the peptide retards the growth in both directions, although the effect is greater in the a direction, suggesting that adsorption occurs at the prism face as well as at the c-face. However, the more recent experimental results of Knight et al.I4 indicate that a particularly strong uptake of the AFP is actually by the 2021 internal planes, a finding which may provide an alternative explanation for the experimental observations. The relative concentrations of the peptide at various faces and internal planes is determined by the differences in the Gibbs free energy of adsorption at respective planes and faces, the adsorption occurring at the aqueous/ice interface. Our present molecular modelling study excludes the effects arising from the presence of water.Therefore, the calculated values of the interaction energies do not correspond to the real situation. However, we hope that the solvent effects would cancel out to some extent when taking the difference in the interaction energy. Thus it may be argued that, notwithstanding the neglect of the complications associated with water as a solvent, the large difference found in the interaction energies of the AFP with the 2021 face and the c-face provides a good indication of the overwhelming preference of the molecule for the former, lending some credence to the revised mechanism for the AF effect put forward by Knight et al.I4 Concluding Remarks The work presented here marks the start of a systematic, comprehensive study that we have undertaken on the application of the molecular modelling approach to the problem of peptide adsorption on ice.We have been able to establish the efficacy of molecular modelling in determining the relative affinities of the AFP molecules for the faces and the internal planes of the ice crystal. The high affinity of the winter flounder AFP for the 2021 plane derives mainly from the steric compatibility between this plane and the peptide molecule, which gives rise to a many-fold increase in the van der Waals component of the surface/molecule binding energy.Thus the fundamental difference in the mode of attachment of the molecule to the 2021 plane and to the c-face does not reside in the propensity of the molecule to hydrogen bond with these surfaces but in the way the molecule is geometrically accommodated on the two surfaces. So, it is the lsquo;lock and keyrsquo; mechanism which seems to operate in this particular case. This work will be extended to the binding of other AFP molecules to the various faces and higher-index internal planes in order to elucidate further the structural and interaction characteristics of the surfaces and the molecules which govern the strength and nature of binding. Furthermore it will be of interest to explore the importance of the cooperative effect in the adsorption process, arising from the interaction between the adjacent adsorbate molecules bound to the surface.Note added in Proof Since the submission of this manuscript, the authors came across a paper by D. Wen and R. A. Laursen (Biophys. J., 1992, 1659), which also presents a molecular modelling of AFP adsorption on ice. The findings of the two studies on the mode of attachment of the winter flounder AFP with the 2021 plane are in excellent qualitative agreement. However, there is a significant quantitative difference in the calculated values of the energy of attachment. This may be due to different forcefields used in the two studies. Inhibition of the Ice Crystal Growth References 1 A.L. DeVries, Philos. Trans. R. Soc., B, 1984, 304, 575. 2 J. T. Eastman and A. L. DeVries, Sci. Am., 1980, 55, 90. 3 S. Mann, New Scientist, 1990, 1707, 42. 4 L. Adadi and S. Weiner, Angew. Chem., 1992,31, 153. 5 C. L. Hew and D. S. C. Yang, Eur. J. Biochem., 1992,203,33. 6 A. L. DeVries, S. K. Komatsu and R. L. Feeney, J. Biol. Chem., 1992, 245, 2901. 7 P. F. Scholander and J. E. Maggert, Cryobiology, 1971,8, 371. 8 J. A. Raymond and A. L. DeVries, Proc. Natl. Acad. Sci. USA, 1977,74,2589. 9 A. L. DeVries, Comp. Biochem. Physiol., 1982,73A, 627. 10 A. L. DeVries, Annu. Rev. Physiol., 1983, 45, 245. 11 A. Chakrabartty, D. S. C. Yang and C. L. Hew, J. Biol. Chem., 1989,264, 11313. 12 C. A. Knight and A. L. DeVries, Science, 1989, 245, 505.13 K. Harrison, J. Hallett, S. Burcham, R. E. Feeney, W. L. Kerr and Y. Yeh, Nature (London), 1987,328, 241. 14 C. A. Knight, C.C. Cheng and A. L. DeVries, Biophys. J., 1991,59,409. 15 N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller and E. Teller, J. Chem. Phys., 1953,21, 1087. 16 M. Lal, R. I. C. Rev., 1971, 4, 97. 17 M. P. Allen and D. J. Tildesley, Computer Simulation of Liquids, Clarenden Press, Oxford, 1987. 18 D. N. J. White, Philos. Trans. R. SOC. A, 1986, 317, 359. 19 D. N. J. White and C. Morrow, Comput. Chem., 1979,3, 33. 20 D. N. J. White, J. N. Ruddock and P. R. Eddington, Mol. Simul., 1989, 3, 71. 21 D. S. C. Yang, M. Sax, A. Chakrabartty and C. L. Hew, Nature (London), 1988,333, 232. 22 V. S. Ananthanarayanan and C. L. Hew, Biochem. Biophys. Res. Commun., 1977,74,685. 23 A. Chakrabartty, V. S. Ananthanarayanan and C. L. Hew, J. Biol. Chem., 1989, 264, 11307. 24 S. Marqusee and R. L. Baldwin, Proc. Natl. Acad. Sci. USA, 1987,84, 8898. 25 F. Franks, Water-A Comprehensive Treatise, Plenum Press, New York, 1972, vol. I, ch. 4. 26 S. W. Peterson and H. A. Levy, Acta Crystallogr., 1957, 10, 70. Paper 3/00232B; Received 13th January, 1993