INSTITUTE OF PHYSICS PUBLISHING
SMART MATERIALS AND STRUCTURES
Smart Mater. Struct. 11 (2002) 772–777
PII: S0964-1726(02)52681-1
Predicting surface properties of proteins on the Connolly molecular surface Jinan Cao, D K Pham, L Tonge and D V Nicolau Industrial Research Institute Swinburne, Swinburne University of Technology, PO Box 218, Hawthorn, Melbourne 3122, Australia E-mail:
[email protected]
Received 5 June 2002, in final form 3 July 2002 Published 20 September 2002 Online at stacks.iop.org/SMS/11/772 Abstract A novel approach has been developed to obtain surface properties of a protein to give a better interpretation of the surface related phenomena, in particular protein attachment on polymer surface. This is achieved by extending the concept of molecular surface to find out relevant surface characteristics determining the interaction behaviour. The Connolly molecular surface is useful in the modelling and computation of surface properties, which could be of fundamental importance to surface-based protein science and engineering. A methodology for obtaining electron charge, hydrophobicity as well as α-helix and β-sheet structural indices has been developed. (Some figures in this article are in colour only in the electronic version)
1. Introduction Among various research themes in protein science and engineering, protein–protein and protein–polymer interactions are a subject of extensive research due to their values to both academic interests and practical applications including the development of smart materials and the design of smart structures. This has resulted in a need to define the ‘molecular’ surface of a protein on which interactions with foreign molecules occur. The concept of a molecular surface for proteins involves the geometric properties of molecules, and is fundamental in biochemistry and biophysics. The atoms are idealized as hard spheres with van der Waals radii. As the three-dimensional structures including the co-ordinates of every atom of many proteins are now known (mostly through x-ray diffraction crystallography and the increasing use of NMR), it is now possible to define and evaluate the physical properties on the molecular surface. This facilitates the computation of a range of useful surface-related properties. For example, the technique of shape complementarity is being used as a prime consideration in docking approaches that take into account entire molecular surfaces rather than strictly active site regions; the molecular surfaces of a receptor and ligand need to match if the molecules are to bind to each other with a certain affinity [1–10]. Together with the energetic docking approach, which further studies the binding
sites on a molecular surface of proteins according to the binding energy including electrostatic, non-electrostatic and hydrogen bonding [11–15], the surface geometry-based shape complementarity docking mechanism has played a significant role in protein docking research. Initially, the solvent-accessible surface was introduced to quantify the molecular surface of a protein. Connolly improved this approach by slightly altering the definition, m, to being the inward-facing surface of a probe sphere as it rolls over a protein molecule. This simple improvement overcomes the shortcoming of the discontinuous first derivatives of the solvent accessible surface, resulting in a convenient and practicable mathematical treatment [16–23]. Depending on the geometric properties, a Connolly surface may contain: a contact convex surface, a concave surface and a saddle-shaped surface. The Connolly surface has been widely accepted as a useful molecular surface; a range of computational algorithms to calculate molecular surface of a protein have been reported in the literature [24–29]. A number of the physio-chemical properties of a protein (in terms of interaction with other protein and non-protein molecules) depend on the surface behaviour of the molecules, which in many cases may differ from those of a bulky one. The interior atoms are expected to make less contributions than the atoms on the surface. This being so, it will be desirable to develop an approach that correlates the various
0964-1726/02/050772+06$30.00 © 2002 IOP Publishing Ltd Printed in the UK
772
Predicting surface properties of proteins on the Connolly molecular surface
interactive behaviour of a protein with its surface properties. Andrade and coworkers [30–33], based on their experimental data, formulated a hypothesis that protein adsorption is mainly dominated by hydrophobic and electrostatic interactions on the surface. We consider it important that the whole set of surface properties, that can and only can be obtained conveniently by computation, are used in the interpretation and modelling of different interactive behaviours such as protein adsorption. There are a number of technical difficulties involved in using surface properties, but this does not prevent one from achieving this goal by developing a computational methodology. This has resulted in a need to further extend the concept of molecular surface. The usefulness of the concept of molecular surface in improving our understanding of the nature of proteins is also due to the fact that computational simulation has emerged as one of the most popular and powerful research techniques in molecular science and engineering. The unique capacity of computer modelling to disentangle various physio-chemical factors that cannot be otherwise disentangled is particularly important for research involving biological molecules. The incomparable speed and efficiency of a computer simulation, as compared with a real biological experiment, is another appealing advantage. It is anticipated that new concepts and algorithms in computational approaches should generate great opportunities for a range of disciplines, in particular for life science and engineering. There are a number of industrial applications in the computation of surface properties as well. Microarray patterning involves surface interactions between a substrate and bio-molecules. The pharmaceutical industry needs to have a better understanding of drug molecule–protein receptor interactions and a better tool for molecular recognition. The cosmetic industry is also a potential beneficiary. It is crucial to have a better understanding of the nature and characteristics of protein–protein, protein–polymer interactions to develop fascinating new healthy cosmetic products [34–37]. The goal of this paper is to investigate how one can obtain the surface properties of a protein using computational techniques. A methodology has been developed for the computation of electron charge, hydrophobicity and α-helical and β-pleated sheet structural characteristics on the Connolly molecular surface. These parameters have been found to be useful in a parallel study of protein adsorption, which will be presented in a consequent paper.
2. Methods Protein molecules exhibit considerable structural complexity, functional heterogeneity as well as conformational lability, which are expected to be the major causes responsible for the complex behaviour of proteins in different environments. To develop a meaningful computational tool that predicts protein–protein and protein–polymer interactions, we consider parameters that reflect the fact that the properties on the surface are most important in a computation or modelling algorithm. In the first instance, the authors computed the electron charge density, the hydrophobilicity and a parameter relating to the α-helix and β-sheet of the protein surface. This is achieved by the following procedures:
(1) extract the x–y–z coordinates of the atoms and the chemical composition of a protein from the protein data bank (PDB) maintained by The Research Collaboratory for Structural Bioinformatics (RCSB); (2) find the Connolly molecular surface of the protein; this is achieved using Connolly’s Fortran programme, freely available from http://www.biohedron.com, which has been further modified to allow computations for large proteins; (3) locate a micro surface element on the Connolly surface, obtain its x–y–z coordinates, area size and outward normal vector; (4) compute a given property of the micro surface element; (5) repeat procedures (3) and (4) until all the surface elements of the whole Connolly surface have been computed; (6) integrate the surface elements to obtain the total, average and distribution, etc; (7) interpret the interactive behaviour of the protein using the obtained parameters. The surface properties deserve further discussion. The electrostatic interaction between two charged atoms is by far the strongest of the physical forces we shall be considering; stronger even than most chemical binding forces. The electrical field, E, at a particular point on the Connolly surface that causes Coulomb interaction with foreign molecules is calculated according to qi . (1) E electrostatic = 4πε0 ε Ri i∈ protein The Coulomb force, F, acting on a charge, q, in the force field is described by Felectrostatic = q E electrostatic .
(2)
Thus the electrical field on the molecular surface can be considered to be a useful parameter describing the surface interaction. For a particle with unit charge, (1) represents the magnitude of the electrostatic force that the particle receives. Quantum mechanics computation was carried out using a commercial software HyperChem developed by the Hypercube Inc. (release 6.02) to obtain the electrical charge of atoms in amino acids. In this study, ab initio quantum mechanics computation was carried out. The electron charges for each atom from 20 amino acids were obtained through the ab initio quantum mechanics computation. The charge and the distance between an atom and a surface element are placed into (1) to calculate the field strength of a micro surface element on the Connolly surface. Hydrophobic interaction describes the strong interaction between non-polar (hydrophobic) molecules and surfaces in water; often stronger than that in free space. Although it is strong, there is no bond associated with this interaction. The hydrophobic interaction is a kind of entropic phenomenon, which arises primarily from the structural rearrangement of water molecules in the overlapping solvation zones as two hydrophobic species come together. There are two existing systems quantifying the hydrophobic propensity of amino acids, as shown in table 1; those with values above zero being hydrophobic in character [38, 39]. In this study, a Kyte– Doolittle system is used because it is a widely applied scale for 773
J Cao et al
Figure 1. Structure of lysozyme as extracted from the PDB. Table 1. Hopp–Woods and Kyte–Doolittle hydropathic characters of amino acids. Amino acid Ala Phe Lys Pro Thr Cys Gly Leu Gle Val
Hopp– Woods −0.5 −2.5 3.0 0.0 −0.4 −1.0 0.0 −1.8 0.2 −1.5
Kyte– Amino Doolittle acid −1.8 −2.8 3.9 1.6 0.7 −2.5 0.4 −3.8 3.5 −4.2
Asp His Met Arg Trp Glu Ile Asn Ser Tyr
Hopp– Woods 3.0 −0.5 −1.3 3.0 −3.4 3.0 −1.8 0.2 0.3 −2.3
Kyte– Doolittle 3.5 3.2 −1.9 4.5 0.9 3.5 −4.5 3.5 0.8 1.3
delineating the hydrophobic character of a protein. Proteins contain hydrophobic surface patches, which if exposed during a collision, become adsorbed by a hydrophobic interaction process. In this study, the hydrophobicity index for the closest amino acid is used to denote the hydrophobicity index for the micro surface element. Another important physical parameter of a protein is related to its structural characteristics, namely the α-helix or βsheet. The complex behaviour of proteins is, to a large extent, due to their conformational lability as they respond to different environments. The structural stability has been measured in this study by calculating the surface area percentage of the α-helix or β-sheet as they are considered to be more stable than other polypeptide chains without a structure [40–42]. Naturally, it is reasonable to assume that the more α-helix or β-pleated sheet polypeptide chains there are on the Connolly surface, the more stable the surface properties are expected to be for the protein.
3. Results and discussion Figure 1 shows a three-dimensional structure of hen eggwhite lysozyme, a relatively small protein described in the PDB (1lyz). Lysozyme contains 1102 atoms constituting of 129 amino acids with a molecular weight of 14 296. Further structural information tells us that 28% amino acids are in the α-helical sequence, and 3% amino acids in the β-pleated sheet structure. Lysozyme has been employed to explain our approach. All the atoms were replaced with hard balls with van 774
Figure 2. The Connolly surface of lysozyme showing a number of positively or negatively charged patches (dark, positively charged patch; shadow, negatively charged patch).
der Waals radii, (i.e. 1.25 Å for hydrogen, 1.875 Å for carbon, 1.688 Å for nitrogen, 1.558 Å for oxygen and 2.102 Å for sulfur). The Connolly molecular surface was first computed as a molecular ball rolling over the protein, obtaining x– y–z coordinates, the local areas and normal vectors for a 100 000 points constituting the Connolly molecular surface. The Connolly surface slightly changes as the radius of the rolling ball changes. This may reflect the true character of real molecular interactions; the accessibility of a protein by different foreign molecules does differ. A water molecule has a van der Waals radius of 1.4 Å. But on average, most protein and polymer molecules have a larger van der Waals radius. For this reason, a probe size of 3.0 Å was used to compute the Connolly surface. The above-mentioned properties of the surface were then calculated using our software. To visualise the result, we employed the commercial software Metlab (The Mathwork Inc., Version 4.2c.1, 1994) to display the surface points; no effort has been made to develop a better software to give quality three-dimensional graphics. Figure 2 shows a computed molecular surface of lysozyme, mapped according to whether it has a positive or negative force field. The unit of the x–y–z coordinates is the Angstrom. For simplicity, they are referred as positively or negatively charged patches hereinafter. It is clear from figure 2 that the surface charges are not uniformly sub-distributed across the protein surface; they are divided into a number of small patches, that are either positively or negatively charged. This may reflect the real circumstance of the intrinsic characteristic of complex protein surfaces. The quantitative charge intensity is stored in numerical form, permitting a range of computations to obtain the surface properties relating to electron charges. Around 34% of the Connolly surface is positively and 66% of the surface negatively charged for lysozyme. The hydrophobicity property of the lysozyme molecular surface is shown in figure 3. Like the electron charge mapping shown in figure 2, lysozyme does not have a uniform hydrophobicity across its molecular surface. However, the hydrophobic/hydrophilic patches have been observed to be larger than those for electrical charges. Around 83% of the surface exhibits hydrophobic propensity, with the remaining 17% showing a hydrophilic propensity. Furthermore, it is noted by comparing figures 2 and 3 that the electron patches do not match with these hydrophobic or hydrophilic
Predicting surface properties of proteins on the Connolly molecular surface
Table 2. Computed Connolly surface, electron charge and hydrophobicity patches.
PDB
Connolly surface area (Å2 )
Positive patch area (Å2 )
Positive charge intensity
Negative patch area (Å2 )
1a4v 1crn 4rhv 1dwr 1lyz 1rav 1ymb 1cb4 6pti 1avh 2avi 1bu5 1cob 7cat 8rat 1a75 2kin 3cyt 1znj 155c
5 117 2 175 34 851 5 810 4 871 11 547 5 745 10 643 3 265 21 620 10 707 10 251 10 894 19 020 5 002 7 851 15 783 8 191 20 670 5 199
896 370 6429 1435 1645 2895 1420 2335 673 5153 2725 2250 2413 3986 1233 1346 3034 1949 3339 791
0.345 0.088 0.196 0.367 0.168 0.198 0.332 0.229 0.228 0.243 0.203 0.182 0.256 0.211 0.237 0.38 0.284 0.365 0.133 0.426
4 222 1 805 28 423 4 375 3 226 8 651 4 325 8 307 2 592 16 469 7 981 8 000 8 480 15 035 3 768 6 475 12 748 6 241 17 332 4 408
Positive charge intensity −0.180 −0.177 −0.343 −0.196 −0.189 −0.317 −0.198 −0.190 −0.689 −0.209 −0.178 −0.179 −0.192 −0.427 −0.194 −0.182 −0.588 −0.189 −0.162 −0.182
Hydrophobic patch area (Å2 )
Average hydrophobic index
Hydrophilic patch area (Å2 )
3 815 1 402 26 031 4 844 4 040 8 909 4 899 8 811 2 392 17 736 8 200 8 620 9 088 14 213 4 274 5 058 11 490 6 693 12 214 3 972
3.06 1.87 2.46 3.01 2.95 2.71 2.96 2.56 3.11 2.84 2.65 2.84 2.59 2.91 2.63 3.03 2.93 2.86 2.50 2.99
1302 774 8823 966 831 2637 847 1832 872 3886 2506 1630 1806 4807 728 2762 4292 1498 8456 1227
Average hydrophilic index −3.55 −3.58 −3.42 −2.89 −3.09 −3.67 −2.86 −3.77 −2.98 −3.13 −3.69 −3.10 −3.78 −3.21 −2.75 −2.44 −3.35 −3.21 −3.43 −2.19
Figure 3. The Connolly surface of lysozyme mapped with either hydrophilic or hydrophobic patches (dark, hydrophobic patch; shadow, hydrophilic patch).
Figure 4. The Connolly surface of lysozyme mapped with α-helical, β-sheet and other patches (dark, α-helix patch; shadow, β-sheet patch; white, others).
patches, resulting in four types of patches with different characteristics for surface interaction, i.e., positively charged hydrophobic, positively charged hydrophilic, negatively charged hydrophobic and negatively charged hydrophilic. Thus a complex interaction between proteins is expected. To highlight the difference between the bulk and surface properties of a protein, we painted the molecular surface of lysozyme according to its structural characteristics, i.e. αhelices or β-sheets or others. A surface area is considered to be α-helical or a β-sheet area according to whether the closest atom in the area belongs to an α-helix or a β-sheet. The results show that there are 24% of α-helical and 6% of β-pleated sheet areas on the molecular surface of lysozyme, which can be compared with the 28 and 3% bulk percentage figures, explaining why the bulk descriptors are not reasonable indicators for measuring the surface interaction of proteins. Figure 4 illustrates the structural mapping. More proteins have been computed using this approach, with the results for 20 proteins being shown in tables 2 and 3. The total Connolly surface, positively or negatively charged patch areas, average charge intensity, as well as hydrophobic
or hydrophilic patches and indices are listed in table 2. Table 3 displays the percentages of α-helices and β-pleated sheets on the Connolly surface compared with the bulky percentage adopted by the PDB. Clearly, all the proteins exhibit heterogeneous characteristics in terms of positively and negatively charged surface patches, or hydrophobic and hydrophilic patches. This is consistent with the intrinsic complexity of proteins. We have found that these computed surface properties are in a far better position for interpreting the adsorption behaviour of a protein on the polymer surface, which will be reported in a subsequent paper.
4. Conclusions The concept of a protein’s molecular surface has been extended to specify various physical properties on the surface. A computation/modelling methodology has been developed for computations of electron charge, hydrophobicity and α-helical and β-pleated sheet structural characteristics on the Connolly molecular surface for large proteins with thousands of atoms. It has been found that a protein surface exhibits a complicated 775
J Cao et al
Table 3. Computed α-helical and β-sheet patches compared with the bulky values. Protein
α-helix % (PDB)
α-helix % on surface
β-sheet % (PDB)
β-sheet % on surface
PDB1a4v PDB1crn PDB4rhv PDB1dwr PDB1lyz PDB1rav PDB1ymb PDB1cb4 PDB6pti PDB1avh PDB2avi PDB1bu5 PDB1cob PDB7cat PDB8rat PDB1a75 PDB2kin PDB3cyt PDB1znj PDB155c
31 43 5 75 28 0 74 4 14 66 0 32 2 26 18 51 33 41 50 26
26 47 5 66 23 0 64 5 13 55 0 33 3 28 17 46 29 39 48 22
7 9 24 0 3 50 0 37 24 0 48 25 37 14 33 2 20 0 0 1
6 7 13 0 2 37 0 26 20 0 40 16 25 4 23 1 13 0 0 1
electron charge and hydrophobicity character, which is further compounded with its surface structural stability. A parallel conjunction study has shown that this technique is useful for modelling protein interfacial interactions, especially for protein attachment in molecularly confined spaces.
Acknowledgments This research project is sponsored by the Defence Advanced Research Projects Agency (DARPA) and the Air Force Research Laboratory, Air Force Material Command, USAF, under agreement no F30602-00-2-0614.
References [1] Norel R, Petry D, Wolfson H J and Nussinov R 1999 Examination of shape complementarity in docking of unbound proteins Proteins: Struct., Funct., Genet. 36 307 [2] Kuntz I, Blaney J, Oatley S, Langridge R and Ferrin T 1982 A geometric approach to macromolecule-ligand interactions J. Mol. Biol. 161 260 [3] Connolly M L 1986 Shape complementarity at the hemoglobin α1 β1 subunit interface Biopolymers 25 1229 [4] Norel R, Fischer D, Wolfson H J and Nussinov R 1994 Molecular surface recognition by a computer vision-based technique Protein Eng. 7 39 [5] Norel R, Wolfson H J and Nussinov R 1999 Small molecule recognition: solid angles surface representation and molecular shape complementarity Comb. Chem. High Throughput Screening 2 223 [6] Norel R, Lin S L, Wolfson H and Nussinov R 1995 Shape complementarity at protein–protein interfaces: the critical role played by surface normals at well placed, sparse points in docking J. Mol. Biol. 252 263 [7] Jiang F and Kim S 1991 Soft docking: matching of molecular surface cubes J. Mol. Biol. 219 79 [8] Shoichet B and Kuntz I 1991 Protein docking and complementarity J. Mol. Biol. 219 327 [9] Bliznyuk A A and Gready J E 1999 Simple method for locating possible ligand binding sites on protein surface J. Comput. Chem. 20 983
776
[10] Gardiner E J, Willett P and Artymiuk P J 2001 Protein docking using a genetic algorithm Proteins: Struct. Funct. Genet. 44 44 [11] Majeux N, Scarsi M, Apostolakis J, Ehrhardt C and Caflisch A 1999 Exhaustive docking of molecular fragments with electrostatic solvation Proteins: Struct. Funct. Genet. 37 88 [12] Majeux N, Scarsi M and Caflisch A 2001 Efficient electrostatic solvation model for protein-fragment docking Proteins: Struct. Funct. Genet. 42 256 [13] Mitchell A S and Spackman M A 2000 Molecular surfaces from the promolecule: a comparison with Hartree–Fock ab initio electron density surfaces J. Comput. Chem. 11 933 [14] Novotny J, Bruccoleri R E, Davis M and Sharp K A 1997 Empirical free energy calculations: a blind test and further improvements to the method J. Mol. Biol. 268 401 [15] Camacho C J, Weng Z, Vajda S and DeLisi C 1999 Free energy landscapes of encounter complexes in protein–protein association Biophys. J. 76 1166 [16] Connolly M L 1983 Solvent-accessible surfaces of proteins and nucleic acids Science 221 709 [17] Connolly M L 1983 Analytical molecular surface calculation J. Appl. Crystallogr. 16 548 [18] Connolly M L 1986 Measurement of protein surface shape by solid angles J. Mol. Graph. 4 3 [19] Via A, Ferre F, Brannetti B and Helmer-Citterich M 2000 Protein surface similarities: a survey of methods to describe and compare protein surfaces Cell. Mol. Life Sci. 57 1970 [20] Duncan B S and Olson A J 1993 Approximation and characterization of molecular surfaces Biopolymers 33 219 [21] Duncan B S and Olson A J 1993 Shape analysis of molecular surfaces Biopolymers 33 231 [22] Lin S L, Nussinov R, Fischer D and Wolfson H J 1994 Molecular surface representations by sparse critical points Proteins: Struct. Funct. Genet. 18 94 [23] Zauhar R J 1995 SMART—A solvent-accessible triangulated surface generator for molecular graphics and boundary element applications J. Comput. Aided Mol. Des. 9 149 [24] Bliznyuk A A and Gready J E 1996 Numerical calculation of molecular surface area: I. Assessment of errors J. Comput. Chem. 8 962 [25] Bliznyuk A A and Gready J E 1996 Numerical calculation of molecular surface area: II. Speed of calculation J. Comput. Chem. 8 970 [26] Bliznyuk A A and Rendell A P 1999 Faster gradients for semiempirical methods J. Comput. Chem. 6 629 [27] Totrov M 1996 The contour-buildup algorithm to calculate the analytical molecular surface J. Struct. Biol. 116 138 [28] Pearl L H and Honegger A 1983 Generation of molecular surfaces for graphic display J. Mol. Graph. 1 9 [29] Meyer A Y 1988 Molecular mechanics and molecular shape: V. On the computation of the bare surface area of molecules J. Comput. Chem. 9 18 [30] Andrade J D, Hlady V, Wei A P, Ho C H, Lea A S, Jeon S I, Lin Y S and Stroup E 1992 Proteins at interface: principles, multivariate aspects, protein resistant surfaces, and direct imaging and manipulation of adsorbed proteins Clin. Mater. 11 67 [31] Feng L and Andrade J D 1994 Protein adsorption on low-temperature isotropic carbon: 1. Protein conformational change probed by differential scanning calorimetry J. Biomed. Mater. Res. 28 735 [32] Feng L and Andrade J D 1994 Protein adsorption on low-temperature isotropic carbon: 2. Effects of surface charge of solids J. Colloid Interface Sci. 166 419 [33] Feng L and Andrade J D 1994 Protein adsorption on low temperature isotropic carbon: 3. Isotherms, competitivity, desorption and exchange of human albumin and fibrinogen Biomaterials 15 323 [34] Nicolau D V, Taguchi T, Taniguchi H and Yoshikawa S 1998 Micron-sized protein–protein patterning on diazonaphthoquinone/novolak thin polymeric films Langmuir 14 1927
Predicting surface properties of proteins on the Connolly molecular surface
[35] Nicolau D V, Taguchi T, Taniguchi H and Yoshikawa S 1999 Negative and positive tone protein patterning on e-beam/deep-UV resists Langmuir 15 3845 [36] Nicolau D V, Taguchi T, Taniguchi H and Yoshikawa S 1999 Protein patterning via radiation-assisted surface functionalization of conventional microlithographic materials Colloids Surf. A 155 51 [37] Nicolau D V and Cross R 2000 Protein profiled features patterned via confocal microscopy Biosens. Bioelectron. 15 85 [38] Kyte J and Doolittle R F 1982 A simple method for displaying the hydropathic character of a protein J. Mol. Biol. 157 105
[39] Hopp T P and Woods K R 1981 Prediction of protein antigenic determinants from amino acids sequences Proc. Natl Acad. Sci. USA 78 3824 [40] Cao J 2000 Is the α–β transition a transition of α-helices to β-pleated sheets? Part I. In situ XRD studies J. Mol. Struct. 553 101 [41] Cao J 2002 Is the α–β transition a transition of α-helices to β-pleated sheets? Part II. Synchrotron investigation for stretched single specimens J. Mol. Struct. 607 69 [42] Branden C and Tooze J 1999 Introduction to Protein Structure 2nd edn (New York: Garland)
777
