Energy-optimized structure of antifreeze protein and its binding mechanism

J. Mol. Biol. (1992) 223, 509-517

Energy-optimized Structure of Antifreeze Protein and Its Binding Mechanism Kuo-Chen

Chou

Computational Chemistry, Upjohn Research Laboratories Kalamazoo, MI 49001, U.S.A. (Received 21 June 1991; accepted 9 October 1991) A combination of Monte Carlo simulated annealing and energy minimization was utilized to determine the conformation of the antifreeze protein from the fish winter flounder. It was found from the energy-optimized structure that the hydroxyl groups of its four threonine residues, i.e. Thr2, Thr13, Thr24, Thr35, are aligned on almost the same line parallel to the helix axis and separated successively by 161, 16.0 and 16.2 A, respectively, very close to the 16.6 A repeat spacing along [Oli2] in ice. Based on such a space match, a zipper-like model is proposed to elucidate the binding mechanism of the antifreeze protein to ice crystals. According to the current model, the antifreeze protein may bind to an ice nucleation structure in a zipper-like fashion through hydrogen bonding of the hydroxyl groups of these four Thr residues to the oxygen atoms along the [Oli2] direction in ice lattice, subsequently stopping or retarding the growth of ice pyramidal planes so as to depress the freeze point. The calculated results and the binding mechanism thus derived accord with recent experimental observations. The mechanistic implications derived from such a special antifreeze molecule might be generally applied to elucidate the structure-function relationship of other antifreeze proteins with the following two common features: (1) recurrence of a Thr residue (or any other polar amino acid residue whose side-chain can form a hydrogen bond with water) in an 1l-amino-acid period along the sequenceconcerned; and (2) a high percentage of Ala residue component therein. Further experiments are suggested to test the ice binding model. Keywords:

zipper-like binding; hydrogen bonding; threonine; spacing match; [Oli2] direction

1. Introduction

1988). However, because of the limitation of resolution, no information whatsoever about the sidechain conformation of the AFP molecule was documented in their report (Yang et al., 1988). Therefore, it is not clear in their mechanism, why not all a-helices, but only those with special sequence distributions and side-chain conformations, possess the function of inhibiting the formation of large ice crystals and depressing the freezing point. To answer this question, it is crucially important to have the detailed information about the side-chain conformation of AFP molecule. Furthermore, the binding mechanism of AFP molecule to ice is critically dependent on the distance of side-chain polar groups along the helix (Knight et al., 1991). As is well known, however, even for a regular a-helix with 3.6 residues per turn, different side-chain angles can not only completely change the orientations of sidechain polar groups, but also significantly alter their distance. For example, by just rotating x1 of two Thr residues along the helix, the distance of their hydroxyl groups might change by about 5 A (1 A = 0.1 nm). Therefore, until the side-chain

Some creatures in nature, such as polar fishes and pine needles, can survive low temperatures at which their fluids would normally freeze. A common mechanism, from the molecular point of view, that enables them to do so is by their production of socalled antifreeze proteins (AFPsf ); which prevent the cell and body fluids from freezing. The properties of AFP molecules have been investigated by various techniques (DeVries, 1977; Raymond & DeVries, 1977; Berman et al., 1980; Davis et al., 1982; DeVries, 1984; Pickett et al., 1984; Hew & Fletcher, 1985; Feeney et al., 1986; Hew et al., 1986; Knight et al., 1991). The crystallographic analysis of the three-dimensional structure of an AFP molecule was reported by Yang et al. (1988). They found that the antifreeze molecule is a single a-helix. Based on such a finding, a mechanism of antifreeze binding to ice surfaces governed by dipole interactions was proposed (Yang et al., 1988) and discussed (Pain, t Abbreviation 0022%2836/92/02050949

used: AFP, antifreeze $03.00/O

protein.

509

0 1992 Academic

Press Limited

conformation of the AFP molecule as well as its backbone conformation are exactly determined, any specific model of AFP binding onto ice remains in doubt. To realize t,his, a combination of Monte Carlo simulated annealing and energy minimization was utilized. Based on the structure thus found, the mechanism of AFP molecule in inhibiting growth of ice has been further explored, and a new model, the zipper-like binding model, has been proposed. 2. Energy-optimized

Structure

All energy calculations were carried out with ECEPP/Z (Nemethy et al., 1983), which is an updated version of the original ECEPP algorithm (Empirical Conformational Energy Program for Peptides) developed by Momany et al. (1975). A general unconstrained optimization algorithm (Gay, 1983) was used whenever energy minimizations were carried out. As prescribed in ECEPP, only torsional degrees-of-freedom are allowed to vary; thus bondstretching and angle-bending terms are omitted from the potential-energy function, which expresses the potential energy as a sum of electrostatic, nonbonded (van der Waals’), hydrogen-bonding and torsional terms. All computations were carried out on an IBM 3090/4OOJ computer at Upjohn Laboratories. The st’andard conventions for nomenclature of peptide conformations have been followed (IUPACIUB Commission on Biochemical Nomenclature, 1970). All the amino acids in the structures reported here are L-type amino acids. The N and C-terminal groups are H,N- and -COOH, respectively. The a’ntifreeze molecule from winter flounder was chosen for calculations. This is an alanine-rich protein wit,h 37 amino acid residues, whose sequence is (Davies et al., 1982; Pickett et al., 1984; Feeney et al., 1986): Asp-Thr-Ala-Ser-Asp-Ala-Ala-Ala-Ala-AlaAla-Leu-Thr-Ala-Ala-Asn-Ala-Lys-Ala-AlaAla-Glu-Leu-Thr-Ala-Ala-Asn-Ala-Ala-AlaAla-Ala-Ala-Ala-Thr-Ala-Arg. As we can seefrom above, the antifreeze molecule is an alanine-rich polypeptide. Nevertheless, it is by no means easy to determine its side-chain conformation, although its backbone dihedral angles are known to be in the helix region and alanine has a relatively simple side-chain. This can be illustrated as follows. According to the sequence above, the numbers of side-chain dihedral angle variables along the AFP sequenceare 3, 3, 1; 2, 3, 1, 1, 1, 1, 1, 1, 4, 3, 1, 1, 3, 1,5, 1, 1, 1,4,4,3, 1, 1, 3, 1, 1, 1, 1, 1, 1, 1, 3, 1 and 7; that is, there would be a total of 73 side-

chain dihedral angle variables. If the starting point for each of these side chain dihedral angles is set at BO”, 180” or -6O”, respectively (Chou et al., 1983), then the starting conformations for the helix would have 373z 676 x 1O34 combinations of staggered side-chain rotamers, a number apparently too large for any existing computer to handle.

As a compromise, Carlacci et al. (1991) adopted a simplified method to build the low-energy a-helix structure. According to their method, the values of the initial backbone dihedral angles for the helix were t.hose in the computed minimum-energy conformation of an isolated regular poly(Ala) a-helix (Chou et al., 1984), namely, (qh,$* w) = ( -6&O”, - 38.0”, 180.0”). The initial side-chain dihedral angles for each of its residues, however, were selected from the data (VBsquez et al., 1983) based on energy minimization of the 20 amino acid residues (N-methylated and C-amidated) according to the following two criteria: (I) the corresponding $ and $Jangle must be in the helix region; and (2) if more than two sets of side-chain dihedral angles are compatible with an a-helical backbone, only the one with the lowest energy is selected. The helix thus generated was subjected to energy minimization: initially only the side-chain dihedral angles were allowed to vary, followed by complete optimization of all dihedral angles. In many cases the helix structures obt,ained through such a method are very stable and it was successfully applied to predict the three-dimensional structure of bovine somatotropin, a four-helix-bundle protein (Carlacci et al., 1991). If the sameminimization procedure was a,pplied to the BFP molecule, the lowest energy thus found was - 205 kcal per mol (I cal = 4-184 J). Preliminary tests indicated that, if the initial side-chain dihedral angles were selected from the most populat.ed conformations in the Brookhaven Protein Data Bank rotamer library, as done by many modehers the final energy would be -203 kcal per mol. Therefore, for the current case, only a slightly lower energy was obtained by using the method of Carlacci et a2. (1991). Actually, the energyminimized structures by these two methods are very similar. For the same molecule; however, using a combination of Monte Carlo simulated annealing and energy minimization, we generated a helix structure with much lower energy, as will be described below. The conception of simula’ted ann_ealing and its application (Kirkpatrick et al., 1983; Cerny, 1985) in helping find the global minimum is based on mimicking the annealing process in metallurgy, i.e. a process of heating and slow cooling in order to toughen and reduce brittleness. It might help under&and the principle of simulated annealing by imagining the following physical picture. If a metal is melted in a high temperature and suddenly cooled down it would become britt,le, indicating its structure is very unstable, corresponding to a local minimum energy state; if, however, the melted metal is cooled down gradually, it would become very tough and hard to break, indicating the metal, after experiencing such an annealing process, possessesthe most stable structural armngement corresponding to the global minimum energy state. The program (Chou & Carlacci, 1991b) adopted here for Monte Carlo simulated annealing was developed in an attempt to find t,he lowest energy conformation of a, polypeptide in a starting-geometry-

Structure-Function

Relationship of Antifreeze Protein

independent manner. In the simulated annealing Monte Carlo algorithm, both downhill and uphill moves are possible so that the molecule concerned is able to surmount the barrier of a local minimum and avoid getting trapped there. This is especially true at the high-temperature stages. It is obvious that an approach that avoids getting stuck in a local minimum would increase the likelihood of detecting the global minimum. During the annealing process, all the backbone angles were held fixed at (4, $, o) = ( -6&O”, -3&O”, 180.0”) (Chou et al., 1984). The starting temperature was assigned at T” = 5 x lo5 K; the cooling factor I = 1.5, the cooling steps p = 30, and the central-processing-unit time (for an IBM 3090 computer) assigned to each of the 30 temperatures for the random conformation-sampling process was @8 hour, i.e. the total central-processing-unit time for the whole annealing process was 68 x 30 = 24 hours, during which 161,304 conformations were randomly selected and examined. “Annealed” via the above process, in which only the side-chain conformation was “rearranged”, the lowest energy

511

thus found was already - 199 kcal per mol. Starting from such a structure, energy minimization was carried out with respect to all its dihedral angles, i.e. both backbone and side-chain dihedral angles. By doing that, the energy of the AFP molecule was further reduced to -217 kcal per mol, which is 12 to 14 kcal per mol lower than the aforementioned results found by using energy minimization only. This indicates that the approach of combining Monte Carlo simulated annealing and energy minimization as applied here is very powerful in preventing the molecule from being trapped into local minimum wells (or enabling it to escape from there) so as to increase significantly the likelihood that it reaches the global minimum. As expected, the final structure thus obtained is a single a-helix, whose stereoscopic drawings are given in Figure 1, in which the four threonine residues, i.e. Thr2, Thrl3, Thr24 and Thr35, are drawn in red in order to distinguish them from the other part (in blue) of the molecule because they play a special role in binding to ice lattice (discussedbelow). The dihedral angles, which uniquely define the side-

Table 1 ECEPP dihedral angles (deg.) of the energy-optimized conformation of an APP moleculefrom winter flounder Residues 1 2

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

25 26 27

28 29 30 31 32 33 34 35

36 37

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 The standard

4 -36.5 -560 -62.6 -680 -67.9 -64 -651 -666 - 65.9 -665 -65.8 -69 -66.9 -665 -653 -66.4 -652 -630 -62.8 -682 -659 -662 - 62.9 -644 -69 -66.2 -666 -6@6 -669 -66.1 -663 -643 -64.8 -65.0 -744 -692 -731

conventions

0.l

1cI -3@O -37.8 -444 -369 -397 -403 -41.1 -41.1 -405

-41.9 - 40.1 -42.5 -38.7 - 42.9 -39.7 - 39.4 -427 -456 -4@1 -40.2 -42.2 -40.6 -48 - 39.9 -40.5 -42.4 - 39.3 -40.8 -40.4 -40.8 -42.1 -41.7 -43.5 -35.4 -37.0 -37.7 - 30.9

-174.1

-

1799 177.6 17%6 179.3 1799 1786 1787 178.5 1789 179.5 1783 179.9 - 1794 -1791 -178.7

179.1 -177.1 -177.3

- 1789 - 1793 179.2 - 1783 - 1796 - 1784 - 178.4 - 1789 - 178.9 - 1788 - 179.1 - 1786 - 1797 - 1768 - 178.3 - 179.4 179.4 179.0

for the nomenclature

x values 68.9 -49.8 60.3 520 -558 60.8 - 58.9 -597 -59-4 -54 -179.3 177.1 - 649 62.2 - 59.3 -177.6 -591 -752 -52 60.6 -1791 -729 175.0 -643 -50 61.0 -665 -596 -593 607 -1795 - 59.5 606 - 59.2 -66 - 59.0 -75

of polypeptide

115.1

- 179.8

-55.3 -67% -241

626 9@2

846 176.6

168.3 61.5 93.6

-10.3

81.4 163.2

conformations

55.3

-178.2

172.6 1753

-60.9

-1794 176.7

-95 53.2 175.6

650

62.8

- 179.7 1790

179.0

1752

- 175.9

173.8

are followed

(IUPAC-IUB,

-177.6 1970).

0.1

1.0

512

K.-C'. Chou

-___

Figure 1. Stereo drawings of the energy-optimized AFP molecule. A unique feature of such a single a-helix is that its-4 threonine residues, i.e. Thr2, Thrl3, Thr24 and Thr35, are all located at the same flank of the helix and are successively separated by the same spacing, with their side-chain hydroxyl oxygen atoms aligned in almost the same line parallel t80 the axis of the helix. These 4 threonine residues play a special role in binding to ice lattice (see text) and hence are drawn in red to distinguish them from the other part of the helix in blue. The M terminus of the polypeptide structure is at the bottom. Only heavy atoms are shown.

Structure-Function

Relationship

chain conformation as well as backbone conformation, are listed in Table 1. The helix length, defined by the distance between the projections of the first and last C” atoms on the helix axis (Chou et al., 1984) is 52.7 A. A conspicuous feature of Figure 1 is that the hydroxyl groups of the four threonine residues are tidily aligned and evenly separated. However, the structures obtained by the methods without combining the simulated annealing procedure do not share this feature, although they also belong to a-helices. In those structures, the four Thr hydroxyl groups apparently deviate from a straight line. The alignment of these four hydroxyl groups is important to the function of the AFP molecule, as will be discussed in the next section. 3. Binding

Mechanism

By means of sequence studies, DeVries & Lin (1977) deduced that in the AFP molecule residues Thr2 and Asp5 are separated by 45 A, a repeat distance that also separates the oxygen atoms in the ice lattice along the a-axis (Fletcher, 1970). They further pointed out that such a match suggested that the binding of AFP to ice might occur through hydrogen bonding of Thr2 and Asp5 residues to the water molecules in the ice lattice, so as to stop the growth of ice. However, it is well documented that, in the absence of AFP molecules, ice nuclei grow perpendicular to their c-axes, thus enlarging the basal plane (Fletcher, 1970). Therefore, it is not clear how binding of AFP to ice along its a-axis, as suggested by DeVries & Lin (1977), can suppress ice growth along a direction not blocked by the AFP molecule. Also, according to our computed results, both the distance of hydroxyl oxygen atoms between Thr2 and Asp5 side-chains and that between Asp1 and of Asp5 side-chains are 4.5 A, suggesting that hydrogen bonding of AFP to the ice lattice via Asp1 and Asp5 could also happen. Furthermore, Aspl, Thr2 and Asp5 are located in an area whose dimension is less than one-seventh of the AFP helix structure. If the binding to ice is due to these two or three amino acid residues, what role does the other major part of the molecule play? Can we find a model by which the AFP molecule inhibits the growth of ice lattice in a more unique and efficient way! Raymond & DeVries (1977) postulated that the AFP molecule might bind to the ice nuclei.t.hat are parallel to the c-axis (prism faces) and subsequently inhibit growth of the basal plane. The RaymondDeVries model can account for the suppression of ice growth, but it does not explain why and how AFP molecules are bound to ice along its c-axis. Yang et al. (1988) proposed that the interaction of the AFP dipole with those of the water molecules in the bound ice nucleus may provide a better explanation for the preferential orientation of the AFP helix with respect to the ice lattice. It is not yet clear, however, why not all a-helix structures, but only those with sequences similar to those of AFPs,

of Antifreeze

Protein

513

would function as antifreeze agents. Therefore, besides helix structure and dipole-dipole inter-action, some other factors associated with the sequence and side-chain conformation that are unique to the AFP molecule must be taken into account in order to reveal the mechanism of its binding. Recently, Knight et al. (1991) reported that the adsorption plane for winter flounder antifreeze was determined experimentally and found to be a pyramidal plane extending off the primary prism plane { 2021) and the molecular alignment was deduced to be along the [Oli2] direction. This is undoubtedly an important finding. However, as claimed in their paper (Knight et al., 1991), the mechanism of binding is not yet clear, although a conjecture was made by them that the flounder/plaice AFP might be adsorbed via hydrogen bonding between the polar side-chains of threonine residues and water molecules in the ice lattice along the [Oli2] direction. Without knowing the side-chain conformation of AFP, however, neither the orientations of the polar side-chains of threonine residues, nor the accurate distances between these polar groups can be uniquely determined, and hence it is hard to give a convincing binding mechanism. In this section, we see what can be derived based on the complete AFP conformation obtained here. Examining the computed three-dimensional structure of the AFP molecule (Fig. 1), we find the following two features: (1) all its four threonine residues, i.e. Thr2, Thrl3, Thr24 and Thr35, are located at the same flank of the helix, with their side-chain hydroxyl oxygen atoms aligned almost in the same line parallel to the axis of the helix; (2) the distance between the side-chain hydroxyl oxygen atom of Thr3 and that of Thrl3 is 16.2 A, the distance between the side-chain hydroxyl oxygen atom of Thrl3 and that of Thr24 is 16.0 A, and the corresponding distance between Thr24 and Thr35 is 16.1 A; i.e. the distances between side-chain hydroxyl oxygen atoms of two neighboring threonine residues are almost identical and are, on average, equal to 16.1 A. It should be pointed out that the above two characters, i.e. the alignment and equal spacing of the side-chain hydroxyl oxygen atoms of the four successive Thr residues, are very sensitively dependent on their side-chain dihedral angles. For example, the distance between the side-chain hydroxyl oxygen atom of Thr3 and that of Thr13 can be anywhere between 13.4 and 18.4 A, depending on their side-chain dihedral angles x1. Therefore, by rotating side-chain dihedral angles of Thr residues, the alignment of their hydroxyl oyxgen atoms and the approximate equality of their spacings can be completely destroyed, even when keeping AFP as a regular 3.6 per turn a-helix, indicating the necessity of using energy optimization to uniquely determine its side-chain conformation. O_n the other hand, the repeat spacing along [0112] in ice (Fig. 2(a)) is given by: dioliq = 2(ai sin’ 60” + ci)l”.

(1)

c-axis

ff3 [ii2

Figure 2. (a) The ice cell and its vector expression: a, = 4.50 A and co = 7.32 (Fletcher, 1970) are the ice iatt,ice parameters from which a repeat spacing d~ol~21= 166.4 (b) Binding of the .4FP along the [Olf2] pl ane in ice is derived. helix to an ice-nucleation structure along it,s [OlT2] plane through hydrogen bonding. Only heavy atoms are shown. The shaded atoms are the hydroxyl oxygen atoms of Thr residues and the oxygen atoms along the [Oli2] plane in ice, respectively. They are associated with each other by an arrow, indicating that a hydrogen bond is to be formed for each of such associated pairs. Because of the limitation of space, the associated pairs for Thr24 and Thr35 and the reievant atoms are not shown.

Structure-Function

Relationship

Substitution of the ice-cell parameters a, = 450 A and c,, = 7.32 (Fletcher, 1970) into equation (1) yields dioiizl= 16.6 A, which is very close to 161 A, the distances between side-chain hydroxyl oxygen atoms of two neighboring Thr residues of the AFP molecule. The presence of almost identical spacings along the AFP helix and along the [Oli2] direction of ice crystal indicates a lattice match. Such a match suggests that the binding of AFP molecule to ice would occur through hydrogen bonding of Thr2, Thr13, Thr24 and Thr35 to the oxygen atoms of water molecules along the [Oli2] direction of the ice lattice, so as to stop or retard the growth of ice (Fig. 2(b)). This finding is consistent with the recent observation by Knight et aZ. (1991). The trivial difference of 05 A in the spacing interval can be adjusted by the following two factors: (1) a 0.5 A stretching fluctuation is allowed for a hydrogen bond (Chou et al., 1990; Chou & Carlacci, 1991a); (2) at 0°C (i.e. 273 K) the accordion-like motion of the AFP molecule along its helix axis would generate a low-frequency wave number of 9 = 17 cm-l with the amplitude of 084 A (Chou, 1983, 1984, 1988). A combination of these two flexible factors would be more than enough movement for such a minor adjustment and make them have a perfect matching and binding, as can be illustrated by a zipper-like binding model (Fig. 3), in which, however, it is the hydrogen bonds that play the role of the nylon teeth for fastening the two “counterparts” to each other. Examining the sequence of AFP molecule as given above, we found that out of its 37 amino acid residues there are 23 Ala residues, i.e. the component of Ala is extremely high, occupying more than 62% of the total constituent amino acid residues of the AFP molecule. As is well known, apart from glycine, alanine has the smallest sidechain. Therefore, a large percentage of Ala residues in the AFP molecule would significantly reduce the likelihood of spatial hindrance for such a zipper-like binding manner. Besides Ala and Thr residues, the other residues in AFP are Aspl, Ser4, Asp5, Leul2, Lysl8, Glu22, Leu23, Asn27 and Arg37. However, as shown in the energy-optimized AFP structure (Fig. l), most of their side-chains point towards the direction far away from the zipper-like binding line and hence would not cause any spatial hindrance problem either.

4. Work Ahead Based on the above calculation and discussion, the following experimental work and theoretical calculations are suggested that might further stimulate the studies of this field. It would be intriguing to generate some mutant AFP molecules by single amino-acid replacement and observe the change of their properties. According to the recent observation by Knight et al. (1991) and the binding mechanism derived from the above calculated results, the single amino-acid substitution for Thr2, Thr13, Thr24 and Thr35 of

of Antifreeze

Protein

515

@2i

Figure 3. Schematic illustration to show the zipper-like binding model between the AFP molecule and ice lattice. The 4 zigzag lines between the hydroxyl oxygen atoms of Thr2, Thr13, Thr24 and Thr35 and the oxygen atoms along the [Oli2] plane of the ice lattice represent 4 hydrogen bonds from which the hydrogen atoms are omitted.

an AFP molecule should be particularly sensitive to the ice-binding property, and hence the function of depressing ice growth. Undoubtedly, a lot of interesting information can be obtained through these experiments. The knowledge and insights thus acquired will be useful in the de nouo design of new antifreeze proteins.

K.-C.

516

The present AFP structure was caldated in terms of ECEPP potential. As reported in a recent paper by Roterman et aE. (1989), a comparison of the CHARMM (Brooks et al., 1983), AMBER {Weiner & Kollman, 1981) and ECEPP potentials for peptides has indicated that the ECEPP potential is very good, and its restriction on bond lengths and bond angles is also an appropriate approximation for polypeptides. Nevertheless, it would be interesting to know what structure will be obtained if the calculation is carried out in the CHARMM or AMBER system. Furthermore, it should be pointed out that all calculations here for the AFP molecule were carried out in the absence of explicit solvent molecules. In vivo, the protein is surrounded by water. The validity of ECEPP in dealing with polypeptides without explicitly including solvent molecules has been discussed by Roterman et al. (1989). As a matter of fact, in the ECEPP algorithm an effective dielectric constant has been used to take approximate account of the solvent effect. However, it would provide much more detailed information about the ice-binding mechanism by AFP if calculations can be made in the presence of explicit solvent molecules, although at present it is quite difficult to do so. In this sense,the present calculations should be regarded as a prelude to calculations that include some of the solvation shells.

Chou The author expresses his gratitude to Dr Boryeu Maa and .Dr B. Vernon Cheney for valuable discussions. Illuminating discussions with Professor Guo-Zhi Zhou are also gratefully a,cknowledged. The at&or is particularly indebted to the editor and referees whose careful review and valuable comments have greatly improved this paper.

Berman, E., $llerhand, A. & DeVries, A. L. (1980). Natural abundance carban 13 nuclear magnetic resonance spectroscopy of antifreeze glycoproteins. J. Biol. Cliem. 255, 4407-4410. Brooks, B. R., Bruccoleri, R. E., Olafson, B. ID., States, D. J., Swaminathan, 8. & Karplus, M. J. (1983). CHARMM: a program for macromolecular energy, minimization, and dynamics calculations. J. Camp. Chem. 4, 187-217. Carla&, L., Chou, K. C. C Maggiora, 6. M. (1991). A heuristic approach to predicting t,he tertiary structure of bovine somatotropin. Biochemistry, 30, 43894398. cerny, V. (1985). Thermodynamics approach to the traveling salesman problem-an efficient simulation algorithm. J. 0ptiml;zation Theory Appl. 45, 41-51. Chou, K.-C. (1983). Low-frequency vibrations of helical structures in protein molecules. Biochem. J. 215, 465-469. Chou, K.-C. (1984). Biological functions of low-frequency vibrations (phonons). III. Helical structures and microenvironment. Biophys. J. 45, 881-890. Chou, K.-C. (1988). Low-frequency collective motion in biomacromolecules and its biological functions. Biophys.

5. Conclusion Based on the computed AFP helix molecule, it is found that the hydroxyl groups of its four threonine residues, i.e. Thr2, Thrl3, Thr24 and Thr35, are almost aligned on the same line parallel to the helix axis and separated successively by - 16.1 8, a distance very close to the 16.6 A repeat spacing along the [Oli2] direction in ice. Such a spacing match suggeststhat the AFP helix molecule might bind to the ice lattice in a zipper-like fashion through hydrogen bonding of the side-chains of these four Thr residues to the oxygen atoms along the [Oli2] direction in the ice lattice, subsequently inhibiting the growth of ice pyramidal planes so as to depress the freeze point. The calculated results and the binding mechanism thus derived are fully consistent with the experimental observations by Knight et al. (1991). A large proportion of Ala residues, as occurs in the AFP molecule, is consistent with the suggested model in the sensethat the zipper-like binding requires no spatial hindrance around the binding line, indicating once again that structure and function in nature are highly harmonized. The antifreezing mechanism derived from such a special AFP molecule and the relationship of its structure and function seemto have a rather general implication; it is noteworthy that the recurrence of a Thr residue for every 11 amino acid residuesalong the sequencehas been observed in a number of other alanine-rich AFP molecules as well.

Them.

30, 3-48.

Chou, K.-C. & Carla&, L. (1991a). Energetic approach to the folding of affl barrels. Proteins: Struct. Funct. Tenet. 9, 280-295. Thou, K.-C. & Carla&, L. (1991b). Simulated annealing approach to the study of protein structures. Protein Eng. 4, 661-661. Chou, K.-C., Nemethy, G. & Scheraga, H. A. (1983). Role of interchain interactions in the stabilization of the right-handed twist of b-sheets. J. Mol. Bid. 168, 389-407. Chou, K.-C., Nemethy, G. & Scheraga, H. A. (1984). Energetic approach to the packing of a-helices. 2. General treatment of nonequivalent and nonregular helices. J. Amer. Chem. Xoc. 106, 3161-3170. Chou, K.-C., Carlacci, L. & Maggiora, 6. M. (1990j. Conformational and geometrical properties of idedized P-barrels in proteins. J. Mol. Biol. 213, 315-326. Davis, P. L., Roach, A. H. & Hew, C. L. (1982). DNA sequence coding for an antifreeze protein precursor from winter flounder. Proc. Nat. Acad. Sci., U.S.A. 79, 335-339. DeVries, A. L. (1977). Glycoproteins as biological a.ntifreeze agents in AntarcOic fishes. Science, 172, 115% 1155. DeVries, A. L. (1984). Role of glycopeptides and peptides in inhibition of crystallization of water in polar fishes. Phil. Trans. Roy. Sot. ser. B, 304, 575-588. De Vries, A. L. & Lin, Y. (1977). Structure of a peptide antifreeze and mechanism of adsorption to ice. Biochim. Biophys. Acta, 495, 388-392. Feeney, R. E., Bureham, T. S. & Yeh, Y. (1986). Antifreeze glycoproteins from polar fish blood. Annu. Rev. Biophys. Biophys. Chem. 15, 59-78. Fletcher, N. H. (1970). In The Chemical Physics of Ice, chapt. 3, Cambridge University Press, Cambridge,

Structure-Function

Relationship

Gay, D. M. (1983). Subroutines for unconstrained minimization using a model/trust-region approach. Assoc. Comput. Mach. Trans. Math. Software, 9, 503-524. Hew, C. L. & Fletcher, G. L. (1985). In Circulation, Respiration Metabolism (Gilles, R., ed.), pp. 553-563, Springer, Berlin. Hew, C. L., Scott, G. K. & Davis, P. L. (1986). In Living in the Cold, Physiological and Biochemical Adaptatkm (Heller, H. C. & Musacchia, X. J., eds), pp. 117-123, Elsevier, New York. IUPAC-IUB Commission on Biochemical Nomenclature. (1970). Abbreviation and symbols for the description of the conformation of polypeptide chains. Biochemistry, 9, 3471-3479. Kirkpatrick, S., Gelatt, C. D., Jr & Vecchi, M. P. (1983). Optimization by simulated annealing. Science, 220, 671-680. Knight, C. A., Cheng, C. C. & DeVries, A. L. (1991). Adsorption of a-helical antifreeze peptides on specific ice crystal surface planes. Biophys. J. 59, 409-418. Momany, F. A., McGuire, R. F., Burgess, A. W. & Scheraga, H. A. (1975). Energy parameters in polypeptides. VII. Geometric parameters, partial atomic charges, nonbonded interactions, hydrogen bond interactions, and intrinsic torsional potentials for the naturally occurring amino acids. J. Phys. Chem. 79, 2361-2381. NBmethy, G., Pottle, M. S. & Scheraga, H. A. (1983). Energy parameters in polypeptides. 9. Updating of geometrical parameters, nonbonded interactions, and hydrogen bond interactions for the naturally occurEdited

of Antifreeze

Protein

517

ring amino acids. J. Phys. Chem. 87, 1883-1887. R. H. (1988). Helices of antifreeze. Nature (London), 333, 207-208. Pickett, M. H., Scott, G., Davies, P., Wang, N., Joshi, S. & Hew, C. (1984). Sequence of an antifreeze protein precursor. Eur. J. Biochem. 143, 35-38. Raymond, J. A. & DeVries, A. L. (1977). Adsorption inhibition as a mechanism of freezing resistance in polar fishes. Proc. Nat. Acad. Sk., U.S.A. 74, 25892593. Roterman, I. K., Lambert, M. H., Gibson, K. D. & Scheraga, H. A. (1989). A comparison of the CHARMM, AMBER and ECEPP potentials for peptides. II. &JI maps for N-acetyl alanine r-methyl amide: comparison, contrasts and simple experimental tests. J. Biomol. Struct. Dynam. 7, 421-453. VBsquez, M., NQmethy, G. & Scheraga, H. A. (1983). Computed conformational states of the 20 naturally occurring amino acid residues and of the prototype residue a-aminobutyric acid. Macromolecules, 16, 1043-1049. Weiner, P. K. & Kollman, P. A. (1981). AMBER: assisted model building with energy refinement. A general program for modeling molecules and their interactions. J. Comp. Chem. 2, 287-303. Yang, D. S. C., Sax, M., Chakrabartty, A. & Hew, C. L. (1988). Crystal structure of an antifreeze polypeptide and its mechanistic implications. Nature (London), 333, 232-237. Pain,

by F. Cohen

Note added in proof. The co-ordinates of all atoms, including hydrogens, for the 3-dimensional antifreeze protein structure have been deposited with the Protein Data Bank, Chemistry Department, Brookhaven National Laboratories, Upton, Long Island, NY 11973, from which copies are available. The entry name has been assigned IATF.