- Email: [email protected]
The effects of steric mutations on the structure of type III antifreeze protein and its interaction with ice1
J. Mol. Biol. (1998) 275, 515±525
The Effects of Steric Mutations on the Structure of Type III Antifreeze Protein and its Interaction with Ice Carl I. DeLuca, Peter L. Davies, Qilu Ye and Zongchao Jia* Department of Biochemistry Queen's University, Kingston Ontario, K7L 3N6, Canada
The interaction of proteins with ice is poorly understood and dif®cult to study, partly because ice is transitory and can present many binding surfaces, and partly because structures have been determined for only two ice-binding proteins. This paper focuses on one of these, a 66-residue antifreeze protein (AFP) from eel pout. The high resolution X-ray structure of this ®sh AFP demonstrated that the proposed ice-binding surface is remarkably ¯at for such a small protein. The residues on the planar surface thought to be involved in ice binding are restrained by hydrogen bonds or by tight packing of their side-chains. To probe the requirement for a ¯at binding surface, a conserved alanine in the center of the AFP planar surface was substituted with larger residues. Six alanine replacement mutants (Ala16 > Cys, Thr, Met, Arg, His and Tyr), designed to disrupt the planarity of the surface and sterically block binding to ice, were characterized by X-ray crystallography and compared with the wild-type AFP. In each case, the detail provided by these crystal structures has helped explain the effects of the mutation on antifreeze activity. The substitutions, Ala16 > His and Ala16 > Tyr, were large enough to shield Gln44, one of the putative ice-binding residues, contributing to their very low thermal hysteresis activity. In addition to sterically hindering the putative ice-binding site, the bulkier residues also caused shifts in the putative ice-binding residues owing to the tight packing of side-chains on the planar surface. This unexpected consequence of the mutations helps account for the severely reduced antifreeze activity. One explanation for residual antifreeze activity in some of the mutants lies in the possibility that AFPs have a role in shaping the site on the ice to which they bind. Thus, side-chain dislocations might be partially accommodated by ice that can freeze around them. It is evident that the disruption of the planarity, by introducing larger residues at the center of the proposed ice-binding site, is not the only factor responsible for the loss of antifreeze activity. There are multiple causes including positional change and steric blockage of some putative ice-binding residues. # 1998 Academic Press Limited
*Corresponding author
Keywords: antifreeze; mutagenesis; ice lattice; thermal hysteresis; crystal structure
Introduction Antifreeze proteins Antifreeze proteins (AFPs) are structurally diverse proteins that allow organisms to survive at ambient temperatures below the colligative freezPresent address: C. I. DeLuca, Institute for Molecular Biology and Biophysics, ETH-Hoenggerberg, CH-8093 ZuÈrich, Switzerland. Abbreviations used: AFP, antifreeze protein. 0022±2836/98/030515±11 $25.00/0/mb971482
ing point of their body ¯uids (Davies & Hew, 1990). They function by binding to the ice surface, thereby stopping further ice crystal growth. Coverage of the ice need not be complete because the exposed ice surface will grow with a curvature between the bound AFPs (Wilson et al., 1993; Wen & Laursen, 1993). This curvature yields a larger ratio of surface area to volume, which in turn causes a local depression of the freezing point of that surface relative to the bulk solvent freezing point (Knight et al., 1991; Cheng & DeVries, 1991). # 1998 Academic Press Limited
516 Bound AFPs only lower the non-equilibrium freezing point and the solution is in fact supercooled. The binding of an AFP to ice is a speci®c process, and different AFPs interact with different ice surfaces (Knight et al., 1991; Cheng & DeVries, 1991). It is generally accepted that binding is mediated by a speci®c match between hydrophilic groups on the protein and oxygen atoms in the ice lattice. It is unlikely, however, that the binding energy contributed by these hydrogen bonds alone is suf®cient to provide nearly irreversible binding to the ice surface. Various mechanisms have been proposed to account for such strong af®nity. For type I AFP, which possesses a linear a-helical structure, a model has been proposed whereby AFPs associate through peptide-peptide interaction on the ice surface to form patches which give rise to suf®cient hydrogen bonding energy to bind the ice irreversibly (Wen & Laursen, 1992a, 1993). When this model was tested with type III AFP there was no evidence that cooperativity was required for tight binding (DeLuca, 1997); hence all the binding energy would need to be contributed by an individual protein. SoÈnnichsen et al. (1996) have proposed that the binding is strengthened by entropic gains resulting from the release of bound water from both the ice and the protein. Also, the planarity of the protein surface and its potential close ®t to the ice surface have led to the hypothesis that van der Waals interactions may contribute to the energetics of binding (Jia et al., 1996; SoÈnnichsen et al., 1996). Putative ice-binding site of type III AFP The low resolution NMR structure of type III AFP (SoÈnnichsen et al., 1993), together with the conservation of speci®c hydrophilic residues between type III AFP isoforms (Davies & Hew, 1990), implicated one region of the protein as the putative ice-binding site. When four residues (Gln9, Asn14, Thr18 and Gln44) from this region were replaced by hydrophilic amino acid residues, either individually or in combination, the mutants displayed altered thermal hysteresis activity and ice-crystal morphology (Chao et al., 1994). Most of these mutations involved substitutions with amino acid residues with smaller side-chains. An exception was Thr18 > Asn, which had a drastic effect on thermal hysteresis activity (10% of wild-type). It is likely that in addition to the alteration in the position of hydrogen bonding groups this substitution also introduced a steric effect on the ice surface which prevented the docking of a section of the AFP to the ice. Further, a conserved surface-accessible alanine residue (Ala16) on the putative ice-binding surface was mutated to amino acid residues with larger side-chains in order to examine the effect of steric interference on ice binding (DeLuca et al., 1996). It was found that bulkier residues at position 16 had more deleterious
Steric Mutants of Type III AFP
effects on thermal hysteresis activity and ice-crystal morphology. Type III AFP structure and proposed icebinding model The high resolution structure of type III AFP from eel pout has now been determined, both by X-ray crystallography (Jia et al., 1996) and heteronuclear NMR (SoÈnnichsen et al., 1996), resulting in the ®rst detailed picture of a globular antifreeze protein. The X-ray structure reveals a match between the residues interacting with ice and the lattice oxygen atoms on the deduced prism binding plane for type III AFP {1010} (Jia et al., 1996). The structure suggested an additional putative icebinding residue, Thr15, which on mutation resulted in a decrease in thermal hysteresis activity. In addition, a potential hydrogen bonding match to the carbonyl oxygen of Ala16 (designated as A16O0 ) is described. The proposed ice-binding atoms in the protein form a pseudo-parallelogram with Thr18, Ala16O0 and Gln44 on one side and Gln9, Thr15 and Asn14 on the other (Figure 1). The Ê separating these sets of parallel distance of 4.5 A groups matches the spacing of oxygen atoms in the prism plane. All of the proposed ice-binding groups interact with the prism plane {1010} except for Asn14, which binds to an advanced layer of ice (Figure 2). For this reason the binding plane should be de®ned as a higher index plane that is very close to the prism plane. The ice-docking model shown in Figure 2 was derived from the ¯atness and spatial geometry of the putative ice-binding
Figure 1. A space ®lling representation of type III AFP showing atoms thought to hydrogen bond to ice (red), carbon atoms (gold) and nitrogen and oxygen atoms (blue). Located at the front surface is the proposed icebinding site. The ice-binding atoms (red) form a pseudoparallelogram which matches the prism ice plane. All diagrams were generated using SETOR (Evans, 1993) unless otherwise stated.
517
Steric Mutants of Type III AFP
Figure 2. A surface representation of type III AFP binding to the stepped region between the prism plane {1010} and the basal {0001} plane. Hydrogen bonds between the protein and the ice are shown as thin lines. Secondary structures are colored green for a-helix and pink for b-strands. The ice oxygen atoms involved in hydrogen bonds are colored red. Residue Asn14, located at bottom left of the protein, interacts with an advanced layer of ice. The Figure was generated using GRASP (Nicholls et al., 1991), MOLSCRIPT (Kraulis, 1991) and RASTER3D (Merritt & Murphy, 1994).
site observed in the crystal structure of type III AFP and suggests a plausible mode by which the AFP interacts with the speci®c ice lattice. Here we have based most of our interpretations upon this model and in the absence of a structure of AFP bound to ice, ``ice-binding'' atoms should be considered as ``putative ice-binding'' atoms. In the ice docking model proposed by Jia et al. (1996), the Asn14 side-chain forms a hydrogen Ê ) to an advanced ice water molecule in bond (2.7 A the conjunction or hinge region between the prism and basal plane (Figure 2). It appears that Asn14 not only contributes to the overall ice-binding af®nity but also plays a pivotal role in initiating step growth inhibition. The crystal structure of a double mutant (Asn14 > Ser/Gln44 > Thr, 10% activity of wild-type) clearly shows that the shortened Ser side-chain would not be able to make an appropriate hydrogen bond to ice (Jia et al., 1996). The mutant Asn14 > Gln (67% activity of wild-type) provides an example of lengthened side-chain at residue 14. Because residue 14 is recessed, it does not have the opportunity to interfere with ice-binding residues the way A16 mutants do on the planar surface (DeLuca, 1997). The ice site to which the protein binds may be entirely or partially preformed. Although a fully formed surface site may occur by chance, it is more likely that an incomplete binding surface will be present. The reversible binding of type III AFP to an incomplete ice surface may in fact help to shape the ®nal binding site. In an extreme case of this graduated binding site model, the AFP binds very weakly (through Asn14) at the leading edge of a new hexagonal ice
layer on the basal plane, thereby slowing its lateral expansion. If an additional layer of ice catches up with the lower layer it would contact a second icebinding residue (Gln44). At these intermediate stages AFP binding to ice might still be reversible, but as additional contacts to subsequent ice layers are formed, the dissociation of the protein is slowed until the protein-ice contact site becomes fully formed. At that point the AFP binding would be virtually irreversible (Jia et al., 1996). The graduated binding site model helps account for the incongruities between type I and type III AFP mutants with reduced antifreeze activity. Type I AFP binds to {2021} and other equivalent planes of ice. This is an inclined plane and so the binding of type I to the ice crystal will naturally produce a tapered crystal with a c:a axis ratio of 3.3. Mutants of type I AFP with reduced af®nity for the ice also grow with the same 3.3 : 1 ratio (Wen & Laursen; 1992b; H. Chao, personal communication). Since type III AFP sits over a portion of the basal plane while binding to a high index plane very close to {1010}, it is possible for this AFP to taper the ice crystal to a hexagonal bipyramid. Mutations that weaken type III AFP binding to ice produce hexagonal bipyramids with steeper bipyramidal surfaces and larger c : a axis ratios than those obtained with the wild-type AFP. The severity of the effects of mutation on the thermal hysteresis activity directly relates to the slope of the ice crystal formed with an in®nite number of possible slopes (DeLuca et al., 1996). At the time of previous study of steric mutations of type III AFP, the high resolution structure was unavailable. Thus the A16 mutants were examined only by one and two-dimensional NMR to ensure that the mutations did not grossly affect the protein fold (DeLuca et al., 1996). Although low resolution NMR structure had shown that the methyl group (Cb) of Ala16 is directed towards the ice surface (SoÈnnichsen et al., 1996), no protein-ice surface interaction model was available for examining Ala16 further. With the availability of the high resolution X-ray structure and the resultant model for the interaction of the protein with the ice lattice (Jia et al., 1996), many aspects of the ice-binding features of type III AFP can be examined in detail. Here we have determined the crystal structures of six Ala16 mutants with varying effects on thermal hysteresis activity and ice-crystal morphology. The structural consequences of the Ala16 mutations are analyzed, and the implications for ice binding are discussed. New insights into the mechanism of action of type III AFP and the requirement for a ¯at binding surface for ice binding are examined.
Results and Discussion Model quality The crystal structure of native type III AFP was Ê resolution (Jia et al., 1996). determined to 1.25 A Figure 3 shows the electron densities (2Fo ÿ Fc) for
518
Figure 3. The electron density map (2Fo ÿ Fc) of type III AFP around (a) two main ice-binding residues Asn14 and Gln44 which are within hydrogen-bonding distance of Lys61, and (b) the region between two ice-binding residues Thr18 and Thr15 which contains Ala16 the site of the substitutions. The maps are contoured at 1s level.
two sections of ice-binding residues. For the planar amide groups of Asn14, Gln44 and Gln9, the Ê electron density map approaches suf®cient 1.25 A quality to allow us to assign the placement of the oxygen and nitrogen atoms. From the structure the atoms that are hydrogen bonded to ice appear to be nitrogen for Asn14 and Gln44, and oxygen for Gln9. This is consistent with the high resolution NMR structure which assigns the nitrogen atoms of Asn14 and Gln44 binding to ice, but places both the oxygen and nitrogen of Gln9 within hydrogenbonding distance of the surface (SoÈnnichsen et al., 1996). The side-chain of residue 16 in the mutant structures (Ala16 > Cys, His, Met, Arg, Thr, Tyr) are all well de®ned (Figure 4), and all of the nonglycine residues are in the allowed regions of the Ramachandran plot (Figure 5). Other side-chains in all six mutant structures are well de®ned except for the N-terminal residues Met1 and Asn2. The C-terminal alanine residues can be seen in all six structures although they are poorly de®ned. Structure description The high resolution crystal structure of type III AFP is composed of two triple strands of antiparal-
Steric Mutants of Type III AFP
lel b-sheets aligned orthogonally (Figure 2). The sheets originally classi®ed as a b-sandwich in the low resolution NMR structure (SoÈnnichsen et al., 1993) can now be seen to be signi®cantly splayed. The b-structures seen in the protein deviate signi®cantly from the classical de®nition but still retain characteristics of this structural element. This causes some dif®culty in classi®cation, depending upon the algorithm used. At the two extremes, the protein is either almost entirely b-sheet (phi/psi angle de®nition) or contains only one b-strand (DSSP program; Kabsch & Sander, 1983). We have chosen STRIDE as the classi®cation method because it uses the combination of hydrogen bond, phi/psi angles and Ca position in its classi®cation (Frischman & Argos, 1995). In this particular case, STRIDE provides a satisfying classi®cation, since no strands are left unpaired. The secondary structure assignment shows similarity to the low resolution NMR structure (SoÈnnichsen et al., 1993) but with some exceptions: the single turn of a-helix (Figure 2) was not previously de®ned, and an area originally assigned as a two-stranded b-sheet is now shown to be too far separated for this classi®cation. The lack of classical b-sheet pattern may be a consequence of forming a ¯at ice-binding site (Jia et al., 1996). The inherent righthanded twist of a b-sheet would not result in such a planar surface. Although the protein lacks classical secondary structural elements there is an extensive hydrogen-bonding network with 37 main-chain hydrogen bonds giving the protein a very tight fold (Figure 6). Aside from the main-chain hydrogen bonds, two side-chains, Ser4 and Tyr63, contribute to the structural stabilization of the protein. The hydroxyl of Tyr63 forms a hydrogen bond to the Oe of Asp36 and the hydroxyl group of Ser4 forms a hydrogen bond to the amide group of Gly31. The latter linkage is consistent with the experimental observation that mutant AFPs containing Ser4 substitutions did not accumulate in the bacterial expression system (Chao, 1994). Another factor contributing to the stabilization of the protein is the large amount of buried hydrophobic surface, which, as noted by SoÈnnichsen et al. (1996), may serve to protect the protein from cold denaturation. Type III AFP is active over a very broad pH range, indicating that the protein structure is not sensitive to pH denaturation (Chao, 1994). This is in agreement with the observation that no electrostatic interactions are present and no charged residues are buried in the protein core. In the region of the ice-binding site all the sidechains seem to be constrained by neighboring residues. This creates a well-ordered ¯at plane for interacting with the ice surface (Figure 2). The sidechains of Asn14 and Gln44 are restrained by hydrogen bonds with the conserved Lys61 (Figure 3(a)). Thr15, Thr18, and Gln9 are restrained by van der Waals interactions with residues Met21, Val20 and Ala16 (Figure 1). This conformational
Steric Mutants of Type III AFP
519
Figure 4. The electron density maps (2Fo ÿ Fc) of the six Ala16 mutants of type III AFP around residues 16. The maps are contoured at 1s level.
restriction may be necessary to give the precise ®t of groups on the protein with the ice surface that includes a hydrogen bonding match to oxygen atoms in the ice lattice. Water structure and protein crystal packing During re®nement of the native protein crystal structure, 90 water molecules were included in the model. It was thought that the ordered water molecules surrounding the protein might be oriented in positions such that they would resemble oxygen positions of the ice surface interacting with AFP. There was, however, no obvious ``ice-like'' character to the water molecules. This may not be surprising if protein crystal packing dictates the water structure. Overall, the crystal has a relatively low solvent content of 30%, which may explain the dif®culty in obtaining heavy-atom derivatives that did not disrupt crystal packing (Jia et al., 1996). Intermolecular contacts do occur in the crystal, most notably a hydrogen bond between Asn8 and Asn14. Asn14 is one of the main ice-binding residues (Figure 2). If its position is in¯uenced by the crystal packing, it may affect the model for the
Figure 5. A composite Ramachandran plot of backbone torsion angles of the six mutants of type III AFP which contain substitutions at position 16. The dark gray area corresponds to the most favored region and the other shaded areas are also allowed as de®ned by PROCHECK (Laskowski et al., 1993). Glycine residues are marked as triangles.
520
Steric Mutants of Type III AFP
Figure 6. Stereo view of the main-chain hydrogen bonds of type III AFP illustrating the lack of regular b-sheet hydrogen-bonding pattern. However, many ``b-structure-like'' interstrand main-chain hydrogen bonds give the protein a compact and rigid fold. The amino and carboxyl termini are designated N and C, respectively. The proposed icebinding site is located at the left surface plan of the protein (orientation is similar to that of Figure 2).
docking of the protein to ice, since it is sterically possible to rotate the side-chain of Asn14 into a position where it contacts the same binding plane as the other ice-binding residues. Bringing the sidechain into the same binding plane also restores the near-perfect planarity of the binding surface of the AFP. However, the side-chain position determined from the crystal structure is attractive, since the side-chain is at the transition area between the prism and basal planes. From the high resolution NMR structure some of the models had the Asn14 side-chain in the recessed position present in the crystal structure with the average position being Ê back from the main binding plane (SoÈnnichsen 2A et al., 1996). Therefore, the same side-chain position does occur in solution and is not simply a crystallization artifact.
phology (Jia et al., 1996). The presence of the sulfur atom did not noticeably interfere with the ability of the protein to bind to ice. From the superimposition of the six mutant and the wild-type structures (Figure 7), it can be seen that the added bulk of the Cys residue caused little perturbation of the protein backbone (Figure 8) and no displacement of the ice-binding atoms (Table 1). The preferred rotamer found in the crystal structure would result in little steric clash with the ice surface at its present Ê (Table 2). distance of 3.2 A Ala16 > Met The mutation of Ala16 to methionine resulted in a subtle change in thermal hysteresis activity (85%
Effects of mutation on ice-binding Now that detailed structures are available for the Ala16 mutants a more educated analysis of their loss of activity can be undertaken. One of the initial surprises about the Ala16 mutations was the more drastic effect on activity of branched side-chain amino acid residues compared to longer straight side-chain ones. When the mutant structures were placed in the icedocked model of type III AFP all the long chain amino acid residues (even unbranched) protruded into positions occupied by ice and would be expected to have a drastic effect on antifreeze activity, particularly if the AFP-binding surface of ice was preformed. However, if the protein binds to a site which is not fully formed the addition of a bulky group would have quite different consequences. Both possibilities must be considered and they may have different consequences for each substitution. Ala16 > Cys The substitution of alanine for cysteine was the most innocuous of the mutations with no change in thermal hysteresis activity or ice-crystal mor-
Figure 7. Superimposition of the six mutant structures and the native type III AFP structure. Only the segment around residue 16 is shown. The positions of Gln44 and Thr15 (amino acid label in dark green) are also indicated. The amino acid label color of residue 16 corresponds to the color of each mutant structure.
521
Steric Mutants of Type III AFP
Table 1. Shift in position of the ice-binding atoms due to mutation Position 16 mutant
Thr15Og
Asn14Nd
Gln44Ne
Cysteine Methionine Threonine Arginine Histidine Tyrosine
± ± ± 1.0 ± 1.0 1.2 0.7 1.1 1.0 0.8 0.4 1.1 0.8 1.1 1.1 0.6 2.3 Ê ) of the ice-binding atoms affected The change in position (in A Ê by residue 16 are shown. Distance changes of less than 0.3 A are indicated by a dash.
a branched amino acid residue. There is a severe Ê ) between the ice surface and the Ce clash (1.3 A of the side-chain conformation found in the crystal structure (Table 2). However, a surface-accessible methionine would have considerable ¯exibility and should ®nd a position that can be accommodated on docking to the ice. If the sidechain is manually rotated, a position can be found where the side-chain has no clash with the ice surface (Table 2). Although this may seem arti®cial, it may be a reasonable interpretation of the molecular events of ice binding. The model described by Jia et al. (1996) proposes that the advancing water layers grow into the protein. This process may literally sweep the interfering side-chain out of the way.
Table 2. Steric interactions between side-chains and ice Figure 8. The r.m.s deviation between the main-chain atoms of the native structure and that of each of the position 16 substitutions is shown. The residue substituted into position 16 is indicated in the top right corner of each plot. Lines indicate the position of the ice-binding residues.
of wild-type) and the ice crystal formed in its presence grew slightly over ten minutes at ÿ0.2 C (DeLuca et al., 1996). However the mutant caused substantial backbone shifts in two regions, namely segments 12 to 15 and 35 to 42 (Figures 7 and 8). The shifts are as drastic, or more so, than those caused by the mutants with signi®cantly reduced activity (Ala16 > His, Ala16 > Tyr); yet this mutant has reasonable thermal hysteresis activity. One explanation for the relatively benign effect of Ala16 > Met is that the position of Asn14 is not altered. Asn14 is arguably the key residue (Asn14 > Ser, 25% activity of wild-type) (Chao et al., 1994) in the ice-binding model because it is potentially the ®rst to interact with the ice surface (Jia et al., 1996). Other mutants with much reduced activity resulted in larger shifts at this crucial residue (Table 1). The steric consequences of introducing a methionine would not be as severe as those caused by
Mutant A16Y
A16H
A16R
A16M A16T A16C
Atom
Manual placementa
Crystalb
Rot. 1c
Rot. 2c
Cg Cd2 Ce2 Cz Ce1 Cd1 OZ Cg Nd Ce Ne Cd Cg Cd Ne Cz NZ1 NZ2 Cg Sd Ce Og Cg Sg
3.8 3.5 3.2 3.6 3.2 3.5 3.2 3.6 2.9 3.1 3.1 2.9 3.6 3.1 3.2 3.3 2.4 2.4 3.7 3.3 3.8 2.9 2.9 3.3
3.6 >5 3.8 3.1 2.2 2.9 3.1 3.3 >5 3.7 2.5 2.9 3.5 3.0 2.5 2.9 3.0 4.0 3.2 2.8 1.3 2.9 2.9 3.2
2.2 2.4 2.2 2.2 1.7 2.8 2.0 3.1 2.5 2.0 3.0 3.8 3.3 2.7 2.3 2.0 1.7 1.0 3.3 2.6 1.9 1.9 2.9 2.9
3.4 >5 3.5 2.7 1.9 2.7 2.8 2.0 0.6 1.4 1.5 2.6 3.3 1.9 1.9 2.4 1.5 3.4 3.2 1.6 1.3 n/a n/a 3.3
Ê ) between the substituted side-chain atoms The distance (in A Ê are in and ice is shown . Those with distances less than 3 A bold. The side-chain positions were either optimized manually (a), unmodi®ed from the position present in the crystal (b), or the main rotamers as de®ned by O (c; Jones et al., 1990). Any rotamers that clashed with the protein backbone were excluded automatically. n/a, not applicable; Rot., rotamer.
522 Ala16 > Thr The mutation to Thr caused a signi®cant drop in thermal hysteresis activity (75% of wild-type) and the ice crystal formed in the presence of the protein showed noticeable growth over ten minutes at ÿ0.2 C (DeLuca et al., 1996). Unlike the Met mutation, the change to Thr caused a larger change in the position of Asn14 (Table 1). In addition, both rotamers of the b-branched Thr have con¯icts with the ice surface and no manual positioning can alleviate this con¯ict (Table 2). Ala16 > Arg The mutation to arginine had a substantial effect on thermal hysteresis activity (65% of wild-type: DeLuca et al., 1996) and the ice crystal formed in its presence grew signi®cantly over the ten minute time course. The shifts in carbon backbone and icebinding residues caused by this mutation are comparable to those seen with the Thr mutation (Figures 7 and 8). The Arg16 side-chain in the crystal structure projects down towards Asn14 and Gln44. The greater loss of thermal hysteresis activity of Ala16 > Arg, compared with Ala16 > Met, may result from a more severe steric clash Ê ) and a shielding or with the ice surface (2.4 A blocking of residue Gln44 by the long side-chain (Figure 7). Mutations of Gln44 showed considerable loss in thermal hysteresis activity (50% of wild-type), and it is very likely that steric shielding of this important residue will have a deleterious effect. Even if the side-chain is pushed aside during AFP adsorption to ice, the terminal branching of an Arg residue will not be easily accommodated by the ice lattice. Ala16 > His and Ala16 > Tyr The Ala16 > His and Ala16 > Tyr mutants are presented together to illustrate the contrast between two mutants that had similar drastic effects on thermal hysteresis activity (Ala16 > His 25% of wild-type; Ala16 > Tyr 33% of wild-type: DeLuca et al., 1996). In the presence of the wildtype AFP there is no detectable ice crystal growth prior to reaching the non-equilibrium freezing point of the solution, at which point rapid uncontrolled freezing of the solution occurs. This point is referred to as the ``burst'' point, since the seed ice crystal almost instantaneously expands to encompass the entire sample well. The Ala16 > His and Ala16 > Tyr mutants caused a similar decrease in the ``burst'' point temperature, and yet their growth phenotypes are quite distinct. The Ala16 > Tyr mutant will hold an ice crystal without growth at constant undercooling. In contrast, Ala16 > His forms ice crystals that grow continuously and evenly. The two mutants both effectively shield Gln44 from the ice surface (Figure 7). They caused comparable shifts in most ice-binding atoms on the protein, although the movement of Gln44 is greater
Steric Mutants of Type III AFP
in Ala16 > Tyr (Table 1). This difference may be a contributing factor but it is unlikely to account for the large phenotypic differences. Although rotation may alleviate some of the steric obstruction of the larger side-chain, a mutation could still reduce antifreeze activity in two ways: (1) the effective ``binding'' concentration of the AFP in solution is lowered if a percentage of the side-chains are in positions unfavorable for binding; and (2) once binding occurs the strength of binding may be lower, thus allowing dissociation of the bound protein from the ice surface. The difference in phenotype of the Tyr and His substitutions may lie in the conformation obtained on binding. If only one of the Tyr rotamers allows binding to the ice lattice, then only a percentage of the AFP in solution will bind. This would give a ``burst'' point at a low thermal hysteresis activity value because the effective AFP concentration would be low. However, if once binding occurs the mutant is as ef®cient an antifreeze as the wild-type AFP it could contain the crystal with little or no growth. This is consistent with the observation that the three main rotamers of Tyr do not allow clashfree binding to the surface but a clash-free conformation can exist. In contrast, if all possible sidechain conformations of a mutant bind with a lower af®nity than the wild-type conformations, dissociation of the protein could occur and would permit slow growth of the ice crystal prior to the ``burst'' point. This observed phenotype of the His mutant is consistent with the fact that no clash-free orientation can be found for the His side-chain. These two mutant phenotypes would then be a result of altered on- and off-rates. We hypothesize that the Tyr mutant would have a reduced on-rate because many collisions with the ice surface would be non-productive, but once the crystal formed it would not grow until the non-equilibrium freezing point was exceeded. In contrast, the His mutant would have a signi®cant off-rate such that even at constant undercooling the ice crystal grows. This is not to say that both effects are not present for each mutant, only that one effect is more pronounced in one case than another. Accommodation of the larger residues The method by which the protein-ice interface compensates for a steric mutation is not known. Docking of the mutated protein to the formed ice surface may not be an accurate representation of the molecular events that are occurring, since the ice surface to which the protein binds will not always be preformed. The advancing ice surface may accommodate the larger residue if one or more water molecule(s) are simply not added to the advancing layer, in effect forming a pocket in the advancing surface that would ®t the larger residue. However, if the surface match of the protein for the ice is important, as proposed by Jia et al. (1996) and SoÈnnichsen et al. (1996), then this accommodation would not provide a perfect ®t
523
Steric Mutants of Type III AFP
and the binding energy would de®nitely be reduced. The exact magnitude of this decrease would depend on the ®t of the residue into that pocket. Alternatively, if the side-chain cannot be accommodated by the growing ice surface, its presence will limit the ability of the ice-binding residues to match that surface. This too would create an overall lower binding af®nity. This study serves as a caveat to interpretation of mutagenic studies in the absence of structural detail. The initial study by DeLuca et al. (1996) attempted to sterically preclude the ice-binding site. The availability of the high resolution structures has provided new insight into why the steric mutants retain some antifreeze activity. An important observation is that the methyl group of Ala16 does not protrude directly from the surface as suggested by the initial NMR study (SoÈnnichsen et al., 1993), but projects with an angle of 30 off the perpendicular direction from the surface. This angle may relieve some of the steric pressure induced by mutation, since the side-chains can be de¯ected more easily. The NMR data (DeLuca et al., 1996) also indicated that there was no global alteration in the protein but did not give the detail necessary to indicate that a perturbation of the ice-binding residues had occurred. With the detailed crystal structures of the Ala16 mutants it is apparent that substitution of larger residues into position 16 did not simply result in an increase in the side-chain length of the residue. These substitutions resulted in the shifting of some of the ice-binding atoms from their wild-type position either due to a slight backbone shift around the Thr15 region or due to steric interactions between the substituted side-chain and the surrounding residues.
Conclusion The crystal structures of the Ala16 substitution mutants have afforded a new level of understanding of the effects of these mutations on thermal hysteresis activity and ice crystal morphology. From this study it may be concluded that there are multiple causes for the loss of thermal hysteresis activity, including the steric in¯uence of the side-chain on binding to the ice surface as well as the effect of perturbation of the positions of the ice-binding atoms by the new residue. This is consistent with the apparent rigidity of the ice-binding surface due to tight side-chain packing. If the residues are arranged with little ¯exibility then the Ala16 substitution by bulkier residues will necessarily affect adjacent residues. This study also supports the belief that the presence of hydrogen-bonding residues matching the ice lattice is not the sole requirement for binding AFP. The surface design incorporates numerous features essential for full antifreeze activity. The complete conservation of Ala16 between type III AFP isoforms (Davies & Hew, 1990) implies a role in maintaining the complementarity of the protein for ice. Like hydrogen bonding groups, non-hydrogen bonding groups on the ice-binding surface may also play an important role, since they may help impart the complementarity and rigidity of the ice-binding surface. The complementarity enables the match of the protein surface to the ice lattice, allowing for van der Waals contacts between the two surfaces; and the rigidity stops untethered side-chains from blocking the interaction of the hydrophilic residues with the ice surface. Any side-chain substitution of residues in this region would affect this match and reduce thermal hysteresis activity.
Table 3. Structure determination and re®nement A. Data statistics Data set A16M mutant A16H mutant A16T mutant A16C mutant A16R mutant A16Y mutant
Total reflections
Unique reflections
Completeness (%)
Rsym
1.9 1.43 1.6 1.6 1.65 1.7
19,230 40,197 28,480 19,644 26,882 30,404
4933 9650 8075 7177 7041 6956
99.1 83.9 95.2 84.5 92.5 97.8
0.077 0.065 0.059 0.062 0.067 0.075
A16M
A16H
A16T
A16C
A16R
A16Y
33.27 39.89 44.54 8.0 - 1.9 0.193 (0.261) 24
33.30 39.93 44.66 8.0 - 1.45 0.191 (0.250) 70
33.27 40.11 44.86 8.0 - 1.6 0.199 (0.26) 72
32.65 39.10 46.26 8.0 - 1.6 0.194 (0.268) 70
333.16 30.04 44.81 8.0 - 1.65 0.215 (0.280) 90
33.11 39.99 45.36 8.0 - 1.7 0.191 (0.207) 32
Ê) Resolution (A
B. Refinement Statistics Data set Ê) Unit cell a (A Ê) b (A Ê) c (A Ê) Resolution range (A Rfactor (Rfree 5%) No. of water atoms
524
Materials and Methods Sample preparation and crystallization Ala16 mutants of type III AFP were produced and puri®ed as described (DeLuca et al., 1996). Protein crystals of mutants were grown under conditions similar to those established for the wild-type protein (Jia et al., 1995) with slight variations in the ammonium sulfate concentration and pH. Table 3 summarizes the data collection statistics for the six mutant AFPs. Determination of mutant structures Ê native structure (Jia et al., 1996) was used The 1.25 A as a starting model for the re®nement of the six Ala16 mutants. Data were collected using a Mar Research imaging plate equipped with a Rigaku rotating anode generator. Data were processed using DENZO (Otwinowski, 1993) and CCP4 (1994). The mutant structures were re®ned using X-PLOR (BruÈnger, 1992). The mutated side-chain was not substituted into the structure until after the ®rst round of re®nement to prevent any model bias. The ®nal re®nement statistics for all of the mutant structures are listed in Table 3. All coordinates have been deposited in the Protein Data Bank and the accession codes are 2MSI for Ala16 > Met mutant, 3MSI for Ala16 > His, 4MSI for Ala16 > Thr, 5MSI for Ala16 > Cys, 6MSI for Ala16 > Arg and 7MSI for Ala16 > Tyr, respectively. Overlap of mutant structures The overlap of the mutant protein structures with the native structure was performed using LSQKAB in the CCP4 (1994) package. The overlapping calculation was carried out over the entire model on main-chain atoms. Since the backbone alignment is averaged over the entire protein, distances of deviation of the ice-binding atoms should not be taken as precise. They do, however, give an indication of the degree of alteration of the ice-binding site. The alternative rotamers for the substituted side-chains were obtained using the program O (Jones et al., 1990); see Figure 1 legend.
Acknowledgements We thank Sherry Gauthier for technical assistance and Steffen Graether for help in producing diagrams. This work was supported by research grants from the Medical Research Council (MRC) of Canada. Z.J. is an MRC scholar and C.I.D. was a recipient of an MRC studentship award.
References BruÈnger, A. T. (1992). X-PLOR (version 3.1) manual. A System for X-ray Crystallography and NMR, Yale University Press, New Haven. Chao, H. (1994). Structure-function relationships in type III antifreeze protein. PhD thesis, Queen's University, Kingston, Canada.
Steric Mutants of Type III AFP Chao, H., SoÈnnichsen, F. D., DeLuca, C. I., Sykes, B. D. & Davies, P. L. (1994). Structure-function relationship in the globular type III antifreeze protein: identi®cation of a cluster of surface residues required for binding to ice. Protein Sci. 3, 1760 ±1769. Cheng, C. C. & DeVries, A. L. (1991). The role of antifreeze glycopeptides and peptides in the freezing avoidance of cold-water ®sh. In Life Under Extreme Conditions (di Prisco, G., ed.), pp. 1 ± 14, SpringerVerlag, Berlin. Collaborative Computational Project No. 4. (1994). The CCP4 suite: Programs for protein crystallography. Acta Crystallog. sect. D, 50, 760± 763. Davies, P. L. & Hew, C. L. (1990). Biochemistry of ®sh antifreeze proteins. FASEB J. 4, 2460 ±2468. DeLuca, C. I. (1997). The high resolution structure of type III antifreeze protein and its implication for ice binding. PhD thesis, Queen's University, Kingston, Canada. DeLuca, C. I., Chao, H., SoÈnnichsen, F. D., Sykes, B. D. & Davies, P. L. (1996). Effect of type III antifreeze protein dilution and mutation on the growth inhibition of ice. Biophys. J. 71, 2346± 2355. Evans, S. V. (1993). SETOR: hardware lighted threedimensional solid model representations of macromolecules. J. Mol. Graph. 11, 134±138. Frishman, D. & Argos, P. (1995). Knowledge based secondary structure assignment. Proteins: Struct. Funct. Genet. 23, 566± 579. Jia, Z., DeLuca, C. I. & Davies, P. L. (1995). Crystallization and preliminary X-ray crystallographic studies on type III antifreeze protein. Protein Sci. 4, 1236± 1238. Jia, Z., DeLuca, C. I., Chao, H. & Davies, P. L. (1996). Structural basis for a globular antifreeze protein binding to ice. Nature, 284, 285± 288. Jones, T. A., Bergdol, M. & Kjaldgaard, M. (1990). O: a macromolecular modeling environment. In Crystallographic and Modeling Methods in Molecular Design (Bugg, C. E. & Ealick, S. E., eds), pp. 189± 195, Springer-Verlag, New York. Kabsch, W. & Sander, C. (1983). Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers, 22, 2577± 2637. Knight, C. A., Cheng, C. C. & DeVries, A. L. (1991). Adsorption of a-helical antifreeze peptides on speci®c ice crystal surface planes. Biophys. J. 59, 409± 418. Kraulis, J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallorg. 24, 946± 950. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283± 291. Merritt, E. A. & Murphy, M. E. P. (1994). RASTER3D (version 2.0): a program for photorealistic molecular graphics. Acta. Crystallog. sect. D, 50, 869± 873. Nicholls, A., Bharadwaj, R. & Honig, B. (1993). GRASP: graphic representation and analysis of surface properties. Biophys. J. 64, 116± 170. Otwinowski, Z. (1993). Oscillation data reduction program. Proceedings of the CCP4 Study Weekend (Sawyer, L., Issacs, N. & Bailey, S., eds), pp. 80 ± 86, SERC Daresbury Laboratory, Warrington. SoÈnnichsen, F. D., Sykes, B. D., Chao, H. & Davies, P. L. (1993). The nonhelical structure of antifreeze protein type III. Science, 259, 1154± 1157.
Steric Mutants of Type III AFP SoÈnnichsen, F. D., DeLuca, C. I., Davies, P. L. & Sykes, B. D. (1996). Re®ned solution structure of type III antifreeze protein: a possible involvement of hydrophobic groups in the energetics of the protein-ice interaction. Structure, 4, 1325± 1337. Wen, D. & Laursen, R. A. (1992a). A model for binding of an antifreeze polypeptide to ice. Biophys. J. 63, 1659± 1662. Wen, D. & Laursen, R. A. (1992b). Structure-function relationships in an antifreeze polypeptide: the role
525 of neutral, polar amino acids. J. Biol. Chem. 267, 14102± 14108. Wen, D. & Laursen, R. A. (1993). Structure-function relationships in an antifreeze polypeptide: the effect of added bulky groups on activity. J. Biol. Chem. 268, 16401± 16406. Wilson, P. W., Beaglehole, D. & DeVries, A. L. (1993). Antifreeze glycopeptide adsorption on single crystal ice surfaces using ellipsometry. Biophys. J. 64, 1878± 1884.
Edited by I. A. Wilson (Received 21 January 1997; received in revised form 17 June 1997; accepted 9 October 1997)
The Effects of Steric Mutations on the Structure of Type III Antifreeze Protein and its Interaction with Ice Carl I. DeLuca, Peter L. Davies, Qilu Ye and Zongchao Jia* Department of Biochemistry Queen's University, Kingston Ontario, K7L 3N6, Canada
The interaction of proteins with ice is poorly understood and dif®cult to study, partly because ice is transitory and can present many binding surfaces, and partly because structures have been determined for only two ice-binding proteins. This paper focuses on one of these, a 66-residue antifreeze protein (AFP) from eel pout. The high resolution X-ray structure of this ®sh AFP demonstrated that the proposed ice-binding surface is remarkably ¯at for such a small protein. The residues on the planar surface thought to be involved in ice binding are restrained by hydrogen bonds or by tight packing of their side-chains. To probe the requirement for a ¯at binding surface, a conserved alanine in the center of the AFP planar surface was substituted with larger residues. Six alanine replacement mutants (Ala16 > Cys, Thr, Met, Arg, His and Tyr), designed to disrupt the planarity of the surface and sterically block binding to ice, were characterized by X-ray crystallography and compared with the wild-type AFP. In each case, the detail provided by these crystal structures has helped explain the effects of the mutation on antifreeze activity. The substitutions, Ala16 > His and Ala16 > Tyr, were large enough to shield Gln44, one of the putative ice-binding residues, contributing to their very low thermal hysteresis activity. In addition to sterically hindering the putative ice-binding site, the bulkier residues also caused shifts in the putative ice-binding residues owing to the tight packing of side-chains on the planar surface. This unexpected consequence of the mutations helps account for the severely reduced antifreeze activity. One explanation for residual antifreeze activity in some of the mutants lies in the possibility that AFPs have a role in shaping the site on the ice to which they bind. Thus, side-chain dislocations might be partially accommodated by ice that can freeze around them. It is evident that the disruption of the planarity, by introducing larger residues at the center of the proposed ice-binding site, is not the only factor responsible for the loss of antifreeze activity. There are multiple causes including positional change and steric blockage of some putative ice-binding residues. # 1998 Academic Press Limited
*Corresponding author
Keywords: antifreeze; mutagenesis; ice lattice; thermal hysteresis; crystal structure
Introduction Antifreeze proteins Antifreeze proteins (AFPs) are structurally diverse proteins that allow organisms to survive at ambient temperatures below the colligative freezPresent address: C. I. DeLuca, Institute for Molecular Biology and Biophysics, ETH-Hoenggerberg, CH-8093 ZuÈrich, Switzerland. Abbreviations used: AFP, antifreeze protein. 0022±2836/98/030515±11 $25.00/0/mb971482
ing point of their body ¯uids (Davies & Hew, 1990). They function by binding to the ice surface, thereby stopping further ice crystal growth. Coverage of the ice need not be complete because the exposed ice surface will grow with a curvature between the bound AFPs (Wilson et al., 1993; Wen & Laursen, 1993). This curvature yields a larger ratio of surface area to volume, which in turn causes a local depression of the freezing point of that surface relative to the bulk solvent freezing point (Knight et al., 1991; Cheng & DeVries, 1991). # 1998 Academic Press Limited
516 Bound AFPs only lower the non-equilibrium freezing point and the solution is in fact supercooled. The binding of an AFP to ice is a speci®c process, and different AFPs interact with different ice surfaces (Knight et al., 1991; Cheng & DeVries, 1991). It is generally accepted that binding is mediated by a speci®c match between hydrophilic groups on the protein and oxygen atoms in the ice lattice. It is unlikely, however, that the binding energy contributed by these hydrogen bonds alone is suf®cient to provide nearly irreversible binding to the ice surface. Various mechanisms have been proposed to account for such strong af®nity. For type I AFP, which possesses a linear a-helical structure, a model has been proposed whereby AFPs associate through peptide-peptide interaction on the ice surface to form patches which give rise to suf®cient hydrogen bonding energy to bind the ice irreversibly (Wen & Laursen, 1992a, 1993). When this model was tested with type III AFP there was no evidence that cooperativity was required for tight binding (DeLuca, 1997); hence all the binding energy would need to be contributed by an individual protein. SoÈnnichsen et al. (1996) have proposed that the binding is strengthened by entropic gains resulting from the release of bound water from both the ice and the protein. Also, the planarity of the protein surface and its potential close ®t to the ice surface have led to the hypothesis that van der Waals interactions may contribute to the energetics of binding (Jia et al., 1996; SoÈnnichsen et al., 1996). Putative ice-binding site of type III AFP The low resolution NMR structure of type III AFP (SoÈnnichsen et al., 1993), together with the conservation of speci®c hydrophilic residues between type III AFP isoforms (Davies & Hew, 1990), implicated one region of the protein as the putative ice-binding site. When four residues (Gln9, Asn14, Thr18 and Gln44) from this region were replaced by hydrophilic amino acid residues, either individually or in combination, the mutants displayed altered thermal hysteresis activity and ice-crystal morphology (Chao et al., 1994). Most of these mutations involved substitutions with amino acid residues with smaller side-chains. An exception was Thr18 > Asn, which had a drastic effect on thermal hysteresis activity (10% of wild-type). It is likely that in addition to the alteration in the position of hydrogen bonding groups this substitution also introduced a steric effect on the ice surface which prevented the docking of a section of the AFP to the ice. Further, a conserved surface-accessible alanine residue (Ala16) on the putative ice-binding surface was mutated to amino acid residues with larger side-chains in order to examine the effect of steric interference on ice binding (DeLuca et al., 1996). It was found that bulkier residues at position 16 had more deleterious
Steric Mutants of Type III AFP
effects on thermal hysteresis activity and ice-crystal morphology. Type III AFP structure and proposed icebinding model The high resolution structure of type III AFP from eel pout has now been determined, both by X-ray crystallography (Jia et al., 1996) and heteronuclear NMR (SoÈnnichsen et al., 1996), resulting in the ®rst detailed picture of a globular antifreeze protein. The X-ray structure reveals a match between the residues interacting with ice and the lattice oxygen atoms on the deduced prism binding plane for type III AFP {1010} (Jia et al., 1996). The structure suggested an additional putative icebinding residue, Thr15, which on mutation resulted in a decrease in thermal hysteresis activity. In addition, a potential hydrogen bonding match to the carbonyl oxygen of Ala16 (designated as A16O0 ) is described. The proposed ice-binding atoms in the protein form a pseudo-parallelogram with Thr18, Ala16O0 and Gln44 on one side and Gln9, Thr15 and Asn14 on the other (Figure 1). The Ê separating these sets of parallel distance of 4.5 A groups matches the spacing of oxygen atoms in the prism plane. All of the proposed ice-binding groups interact with the prism plane {1010} except for Asn14, which binds to an advanced layer of ice (Figure 2). For this reason the binding plane should be de®ned as a higher index plane that is very close to the prism plane. The ice-docking model shown in Figure 2 was derived from the ¯atness and spatial geometry of the putative ice-binding
Figure 1. A space ®lling representation of type III AFP showing atoms thought to hydrogen bond to ice (red), carbon atoms (gold) and nitrogen and oxygen atoms (blue). Located at the front surface is the proposed icebinding site. The ice-binding atoms (red) form a pseudoparallelogram which matches the prism ice plane. All diagrams were generated using SETOR (Evans, 1993) unless otherwise stated.
517
Steric Mutants of Type III AFP
Figure 2. A surface representation of type III AFP binding to the stepped region between the prism plane {1010} and the basal {0001} plane. Hydrogen bonds between the protein and the ice are shown as thin lines. Secondary structures are colored green for a-helix and pink for b-strands. The ice oxygen atoms involved in hydrogen bonds are colored red. Residue Asn14, located at bottom left of the protein, interacts with an advanced layer of ice. The Figure was generated using GRASP (Nicholls et al., 1991), MOLSCRIPT (Kraulis, 1991) and RASTER3D (Merritt & Murphy, 1994).
site observed in the crystal structure of type III AFP and suggests a plausible mode by which the AFP interacts with the speci®c ice lattice. Here we have based most of our interpretations upon this model and in the absence of a structure of AFP bound to ice, ``ice-binding'' atoms should be considered as ``putative ice-binding'' atoms. In the ice docking model proposed by Jia et al. (1996), the Asn14 side-chain forms a hydrogen Ê ) to an advanced ice water molecule in bond (2.7 A the conjunction or hinge region between the prism and basal plane (Figure 2). It appears that Asn14 not only contributes to the overall ice-binding af®nity but also plays a pivotal role in initiating step growth inhibition. The crystal structure of a double mutant (Asn14 > Ser/Gln44 > Thr, 10% activity of wild-type) clearly shows that the shortened Ser side-chain would not be able to make an appropriate hydrogen bond to ice (Jia et al., 1996). The mutant Asn14 > Gln (67% activity of wild-type) provides an example of lengthened side-chain at residue 14. Because residue 14 is recessed, it does not have the opportunity to interfere with ice-binding residues the way A16 mutants do on the planar surface (DeLuca, 1997). The ice site to which the protein binds may be entirely or partially preformed. Although a fully formed surface site may occur by chance, it is more likely that an incomplete binding surface will be present. The reversible binding of type III AFP to an incomplete ice surface may in fact help to shape the ®nal binding site. In an extreme case of this graduated binding site model, the AFP binds very weakly (through Asn14) at the leading edge of a new hexagonal ice
layer on the basal plane, thereby slowing its lateral expansion. If an additional layer of ice catches up with the lower layer it would contact a second icebinding residue (Gln44). At these intermediate stages AFP binding to ice might still be reversible, but as additional contacts to subsequent ice layers are formed, the dissociation of the protein is slowed until the protein-ice contact site becomes fully formed. At that point the AFP binding would be virtually irreversible (Jia et al., 1996). The graduated binding site model helps account for the incongruities between type I and type III AFP mutants with reduced antifreeze activity. Type I AFP binds to {2021} and other equivalent planes of ice. This is an inclined plane and so the binding of type I to the ice crystal will naturally produce a tapered crystal with a c:a axis ratio of 3.3. Mutants of type I AFP with reduced af®nity for the ice also grow with the same 3.3 : 1 ratio (Wen & Laursen; 1992b; H. Chao, personal communication). Since type III AFP sits over a portion of the basal plane while binding to a high index plane very close to {1010}, it is possible for this AFP to taper the ice crystal to a hexagonal bipyramid. Mutations that weaken type III AFP binding to ice produce hexagonal bipyramids with steeper bipyramidal surfaces and larger c : a axis ratios than those obtained with the wild-type AFP. The severity of the effects of mutation on the thermal hysteresis activity directly relates to the slope of the ice crystal formed with an in®nite number of possible slopes (DeLuca et al., 1996). At the time of previous study of steric mutations of type III AFP, the high resolution structure was unavailable. Thus the A16 mutants were examined only by one and two-dimensional NMR to ensure that the mutations did not grossly affect the protein fold (DeLuca et al., 1996). Although low resolution NMR structure had shown that the methyl group (Cb) of Ala16 is directed towards the ice surface (SoÈnnichsen et al., 1996), no protein-ice surface interaction model was available for examining Ala16 further. With the availability of the high resolution X-ray structure and the resultant model for the interaction of the protein with the ice lattice (Jia et al., 1996), many aspects of the ice-binding features of type III AFP can be examined in detail. Here we have determined the crystal structures of six Ala16 mutants with varying effects on thermal hysteresis activity and ice-crystal morphology. The structural consequences of the Ala16 mutations are analyzed, and the implications for ice binding are discussed. New insights into the mechanism of action of type III AFP and the requirement for a ¯at binding surface for ice binding are examined.
Results and Discussion Model quality The crystal structure of native type III AFP was Ê resolution (Jia et al., 1996). determined to 1.25 A Figure 3 shows the electron densities (2Fo ÿ Fc) for
518
Figure 3. The electron density map (2Fo ÿ Fc) of type III AFP around (a) two main ice-binding residues Asn14 and Gln44 which are within hydrogen-bonding distance of Lys61, and (b) the region between two ice-binding residues Thr18 and Thr15 which contains Ala16 the site of the substitutions. The maps are contoured at 1s level.
two sections of ice-binding residues. For the planar amide groups of Asn14, Gln44 and Gln9, the Ê electron density map approaches suf®cient 1.25 A quality to allow us to assign the placement of the oxygen and nitrogen atoms. From the structure the atoms that are hydrogen bonded to ice appear to be nitrogen for Asn14 and Gln44, and oxygen for Gln9. This is consistent with the high resolution NMR structure which assigns the nitrogen atoms of Asn14 and Gln44 binding to ice, but places both the oxygen and nitrogen of Gln9 within hydrogenbonding distance of the surface (SoÈnnichsen et al., 1996). The side-chain of residue 16 in the mutant structures (Ala16 > Cys, His, Met, Arg, Thr, Tyr) are all well de®ned (Figure 4), and all of the nonglycine residues are in the allowed regions of the Ramachandran plot (Figure 5). Other side-chains in all six mutant structures are well de®ned except for the N-terminal residues Met1 and Asn2. The C-terminal alanine residues can be seen in all six structures although they are poorly de®ned. Structure description The high resolution crystal structure of type III AFP is composed of two triple strands of antiparal-
Steric Mutants of Type III AFP
lel b-sheets aligned orthogonally (Figure 2). The sheets originally classi®ed as a b-sandwich in the low resolution NMR structure (SoÈnnichsen et al., 1993) can now be seen to be signi®cantly splayed. The b-structures seen in the protein deviate signi®cantly from the classical de®nition but still retain characteristics of this structural element. This causes some dif®culty in classi®cation, depending upon the algorithm used. At the two extremes, the protein is either almost entirely b-sheet (phi/psi angle de®nition) or contains only one b-strand (DSSP program; Kabsch & Sander, 1983). We have chosen STRIDE as the classi®cation method because it uses the combination of hydrogen bond, phi/psi angles and Ca position in its classi®cation (Frischman & Argos, 1995). In this particular case, STRIDE provides a satisfying classi®cation, since no strands are left unpaired. The secondary structure assignment shows similarity to the low resolution NMR structure (SoÈnnichsen et al., 1993) but with some exceptions: the single turn of a-helix (Figure 2) was not previously de®ned, and an area originally assigned as a two-stranded b-sheet is now shown to be too far separated for this classi®cation. The lack of classical b-sheet pattern may be a consequence of forming a ¯at ice-binding site (Jia et al., 1996). The inherent righthanded twist of a b-sheet would not result in such a planar surface. Although the protein lacks classical secondary structural elements there is an extensive hydrogen-bonding network with 37 main-chain hydrogen bonds giving the protein a very tight fold (Figure 6). Aside from the main-chain hydrogen bonds, two side-chains, Ser4 and Tyr63, contribute to the structural stabilization of the protein. The hydroxyl of Tyr63 forms a hydrogen bond to the Oe of Asp36 and the hydroxyl group of Ser4 forms a hydrogen bond to the amide group of Gly31. The latter linkage is consistent with the experimental observation that mutant AFPs containing Ser4 substitutions did not accumulate in the bacterial expression system (Chao, 1994). Another factor contributing to the stabilization of the protein is the large amount of buried hydrophobic surface, which, as noted by SoÈnnichsen et al. (1996), may serve to protect the protein from cold denaturation. Type III AFP is active over a very broad pH range, indicating that the protein structure is not sensitive to pH denaturation (Chao, 1994). This is in agreement with the observation that no electrostatic interactions are present and no charged residues are buried in the protein core. In the region of the ice-binding site all the sidechains seem to be constrained by neighboring residues. This creates a well-ordered ¯at plane for interacting with the ice surface (Figure 2). The sidechains of Asn14 and Gln44 are restrained by hydrogen bonds with the conserved Lys61 (Figure 3(a)). Thr15, Thr18, and Gln9 are restrained by van der Waals interactions with residues Met21, Val20 and Ala16 (Figure 1). This conformational
Steric Mutants of Type III AFP
519
Figure 4. The electron density maps (2Fo ÿ Fc) of the six Ala16 mutants of type III AFP around residues 16. The maps are contoured at 1s level.
restriction may be necessary to give the precise ®t of groups on the protein with the ice surface that includes a hydrogen bonding match to oxygen atoms in the ice lattice. Water structure and protein crystal packing During re®nement of the native protein crystal structure, 90 water molecules were included in the model. It was thought that the ordered water molecules surrounding the protein might be oriented in positions such that they would resemble oxygen positions of the ice surface interacting with AFP. There was, however, no obvious ``ice-like'' character to the water molecules. This may not be surprising if protein crystal packing dictates the water structure. Overall, the crystal has a relatively low solvent content of 30%, which may explain the dif®culty in obtaining heavy-atom derivatives that did not disrupt crystal packing (Jia et al., 1996). Intermolecular contacts do occur in the crystal, most notably a hydrogen bond between Asn8 and Asn14. Asn14 is one of the main ice-binding residues (Figure 2). If its position is in¯uenced by the crystal packing, it may affect the model for the
Figure 5. A composite Ramachandran plot of backbone torsion angles of the six mutants of type III AFP which contain substitutions at position 16. The dark gray area corresponds to the most favored region and the other shaded areas are also allowed as de®ned by PROCHECK (Laskowski et al., 1993). Glycine residues are marked as triangles.
520
Steric Mutants of Type III AFP
Figure 6. Stereo view of the main-chain hydrogen bonds of type III AFP illustrating the lack of regular b-sheet hydrogen-bonding pattern. However, many ``b-structure-like'' interstrand main-chain hydrogen bonds give the protein a compact and rigid fold. The amino and carboxyl termini are designated N and C, respectively. The proposed icebinding site is located at the left surface plan of the protein (orientation is similar to that of Figure 2).
docking of the protein to ice, since it is sterically possible to rotate the side-chain of Asn14 into a position where it contacts the same binding plane as the other ice-binding residues. Bringing the sidechain into the same binding plane also restores the near-perfect planarity of the binding surface of the AFP. However, the side-chain position determined from the crystal structure is attractive, since the side-chain is at the transition area between the prism and basal planes. From the high resolution NMR structure some of the models had the Asn14 side-chain in the recessed position present in the crystal structure with the average position being Ê back from the main binding plane (SoÈnnichsen 2A et al., 1996). Therefore, the same side-chain position does occur in solution and is not simply a crystallization artifact.
phology (Jia et al., 1996). The presence of the sulfur atom did not noticeably interfere with the ability of the protein to bind to ice. From the superimposition of the six mutant and the wild-type structures (Figure 7), it can be seen that the added bulk of the Cys residue caused little perturbation of the protein backbone (Figure 8) and no displacement of the ice-binding atoms (Table 1). The preferred rotamer found in the crystal structure would result in little steric clash with the ice surface at its present Ê (Table 2). distance of 3.2 A Ala16 > Met The mutation of Ala16 to methionine resulted in a subtle change in thermal hysteresis activity (85%
Effects of mutation on ice-binding Now that detailed structures are available for the Ala16 mutants a more educated analysis of their loss of activity can be undertaken. One of the initial surprises about the Ala16 mutations was the more drastic effect on activity of branched side-chain amino acid residues compared to longer straight side-chain ones. When the mutant structures were placed in the icedocked model of type III AFP all the long chain amino acid residues (even unbranched) protruded into positions occupied by ice and would be expected to have a drastic effect on antifreeze activity, particularly if the AFP-binding surface of ice was preformed. However, if the protein binds to a site which is not fully formed the addition of a bulky group would have quite different consequences. Both possibilities must be considered and they may have different consequences for each substitution. Ala16 > Cys The substitution of alanine for cysteine was the most innocuous of the mutations with no change in thermal hysteresis activity or ice-crystal mor-
Figure 7. Superimposition of the six mutant structures and the native type III AFP structure. Only the segment around residue 16 is shown. The positions of Gln44 and Thr15 (amino acid label in dark green) are also indicated. The amino acid label color of residue 16 corresponds to the color of each mutant structure.
521
Steric Mutants of Type III AFP
Table 1. Shift in position of the ice-binding atoms due to mutation Position 16 mutant
Thr15Og
Asn14Nd
Gln44Ne
Cysteine Methionine Threonine Arginine Histidine Tyrosine
± ± ± 1.0 ± 1.0 1.2 0.7 1.1 1.0 0.8 0.4 1.1 0.8 1.1 1.1 0.6 2.3 Ê ) of the ice-binding atoms affected The change in position (in A Ê by residue 16 are shown. Distance changes of less than 0.3 A are indicated by a dash.
a branched amino acid residue. There is a severe Ê ) between the ice surface and the Ce clash (1.3 A of the side-chain conformation found in the crystal structure (Table 2). However, a surface-accessible methionine would have considerable ¯exibility and should ®nd a position that can be accommodated on docking to the ice. If the sidechain is manually rotated, a position can be found where the side-chain has no clash with the ice surface (Table 2). Although this may seem arti®cial, it may be a reasonable interpretation of the molecular events of ice binding. The model described by Jia et al. (1996) proposes that the advancing water layers grow into the protein. This process may literally sweep the interfering side-chain out of the way.
Table 2. Steric interactions between side-chains and ice Figure 8. The r.m.s deviation between the main-chain atoms of the native structure and that of each of the position 16 substitutions is shown. The residue substituted into position 16 is indicated in the top right corner of each plot. Lines indicate the position of the ice-binding residues.
of wild-type) and the ice crystal formed in its presence grew slightly over ten minutes at ÿ0.2 C (DeLuca et al., 1996). However the mutant caused substantial backbone shifts in two regions, namely segments 12 to 15 and 35 to 42 (Figures 7 and 8). The shifts are as drastic, or more so, than those caused by the mutants with signi®cantly reduced activity (Ala16 > His, Ala16 > Tyr); yet this mutant has reasonable thermal hysteresis activity. One explanation for the relatively benign effect of Ala16 > Met is that the position of Asn14 is not altered. Asn14 is arguably the key residue (Asn14 > Ser, 25% activity of wild-type) (Chao et al., 1994) in the ice-binding model because it is potentially the ®rst to interact with the ice surface (Jia et al., 1996). Other mutants with much reduced activity resulted in larger shifts at this crucial residue (Table 1). The steric consequences of introducing a methionine would not be as severe as those caused by
Mutant A16Y
A16H
A16R
A16M A16T A16C
Atom
Manual placementa
Crystalb
Rot. 1c
Rot. 2c
Cg Cd2 Ce2 Cz Ce1 Cd1 OZ Cg Nd Ce Ne Cd Cg Cd Ne Cz NZ1 NZ2 Cg Sd Ce Og Cg Sg
3.8 3.5 3.2 3.6 3.2 3.5 3.2 3.6 2.9 3.1 3.1 2.9 3.6 3.1 3.2 3.3 2.4 2.4 3.7 3.3 3.8 2.9 2.9 3.3
3.6 >5 3.8 3.1 2.2 2.9 3.1 3.3 >5 3.7 2.5 2.9 3.5 3.0 2.5 2.9 3.0 4.0 3.2 2.8 1.3 2.9 2.9 3.2
2.2 2.4 2.2 2.2 1.7 2.8 2.0 3.1 2.5 2.0 3.0 3.8 3.3 2.7 2.3 2.0 1.7 1.0 3.3 2.6 1.9 1.9 2.9 2.9
3.4 >5 3.5 2.7 1.9 2.7 2.8 2.0 0.6 1.4 1.5 2.6 3.3 1.9 1.9 2.4 1.5 3.4 3.2 1.6 1.3 n/a n/a 3.3
Ê ) between the substituted side-chain atoms The distance (in A Ê are in and ice is shown . Those with distances less than 3 A bold. The side-chain positions were either optimized manually (a), unmodi®ed from the position present in the crystal (b), or the main rotamers as de®ned by O (c; Jones et al., 1990). Any rotamers that clashed with the protein backbone were excluded automatically. n/a, not applicable; Rot., rotamer.
522 Ala16 > Thr The mutation to Thr caused a signi®cant drop in thermal hysteresis activity (75% of wild-type) and the ice crystal formed in the presence of the protein showed noticeable growth over ten minutes at ÿ0.2 C (DeLuca et al., 1996). Unlike the Met mutation, the change to Thr caused a larger change in the position of Asn14 (Table 1). In addition, both rotamers of the b-branched Thr have con¯icts with the ice surface and no manual positioning can alleviate this con¯ict (Table 2). Ala16 > Arg The mutation to arginine had a substantial effect on thermal hysteresis activity (65% of wild-type: DeLuca et al., 1996) and the ice crystal formed in its presence grew signi®cantly over the ten minute time course. The shifts in carbon backbone and icebinding residues caused by this mutation are comparable to those seen with the Thr mutation (Figures 7 and 8). The Arg16 side-chain in the crystal structure projects down towards Asn14 and Gln44. The greater loss of thermal hysteresis activity of Ala16 > Arg, compared with Ala16 > Met, may result from a more severe steric clash Ê ) and a shielding or with the ice surface (2.4 A blocking of residue Gln44 by the long side-chain (Figure 7). Mutations of Gln44 showed considerable loss in thermal hysteresis activity (50% of wild-type), and it is very likely that steric shielding of this important residue will have a deleterious effect. Even if the side-chain is pushed aside during AFP adsorption to ice, the terminal branching of an Arg residue will not be easily accommodated by the ice lattice. Ala16 > His and Ala16 > Tyr The Ala16 > His and Ala16 > Tyr mutants are presented together to illustrate the contrast between two mutants that had similar drastic effects on thermal hysteresis activity (Ala16 > His 25% of wild-type; Ala16 > Tyr 33% of wild-type: DeLuca et al., 1996). In the presence of the wildtype AFP there is no detectable ice crystal growth prior to reaching the non-equilibrium freezing point of the solution, at which point rapid uncontrolled freezing of the solution occurs. This point is referred to as the ``burst'' point, since the seed ice crystal almost instantaneously expands to encompass the entire sample well. The Ala16 > His and Ala16 > Tyr mutants caused a similar decrease in the ``burst'' point temperature, and yet their growth phenotypes are quite distinct. The Ala16 > Tyr mutant will hold an ice crystal without growth at constant undercooling. In contrast, Ala16 > His forms ice crystals that grow continuously and evenly. The two mutants both effectively shield Gln44 from the ice surface (Figure 7). They caused comparable shifts in most ice-binding atoms on the protein, although the movement of Gln44 is greater
Steric Mutants of Type III AFP
in Ala16 > Tyr (Table 1). This difference may be a contributing factor but it is unlikely to account for the large phenotypic differences. Although rotation may alleviate some of the steric obstruction of the larger side-chain, a mutation could still reduce antifreeze activity in two ways: (1) the effective ``binding'' concentration of the AFP in solution is lowered if a percentage of the side-chains are in positions unfavorable for binding; and (2) once binding occurs the strength of binding may be lower, thus allowing dissociation of the bound protein from the ice surface. The difference in phenotype of the Tyr and His substitutions may lie in the conformation obtained on binding. If only one of the Tyr rotamers allows binding to the ice lattice, then only a percentage of the AFP in solution will bind. This would give a ``burst'' point at a low thermal hysteresis activity value because the effective AFP concentration would be low. However, if once binding occurs the mutant is as ef®cient an antifreeze as the wild-type AFP it could contain the crystal with little or no growth. This is consistent with the observation that the three main rotamers of Tyr do not allow clashfree binding to the surface but a clash-free conformation can exist. In contrast, if all possible sidechain conformations of a mutant bind with a lower af®nity than the wild-type conformations, dissociation of the protein could occur and would permit slow growth of the ice crystal prior to the ``burst'' point. This observed phenotype of the His mutant is consistent with the fact that no clash-free orientation can be found for the His side-chain. These two mutant phenotypes would then be a result of altered on- and off-rates. We hypothesize that the Tyr mutant would have a reduced on-rate because many collisions with the ice surface would be non-productive, but once the crystal formed it would not grow until the non-equilibrium freezing point was exceeded. In contrast, the His mutant would have a signi®cant off-rate such that even at constant undercooling the ice crystal grows. This is not to say that both effects are not present for each mutant, only that one effect is more pronounced in one case than another. Accommodation of the larger residues The method by which the protein-ice interface compensates for a steric mutation is not known. Docking of the mutated protein to the formed ice surface may not be an accurate representation of the molecular events that are occurring, since the ice surface to which the protein binds will not always be preformed. The advancing ice surface may accommodate the larger residue if one or more water molecule(s) are simply not added to the advancing layer, in effect forming a pocket in the advancing surface that would ®t the larger residue. However, if the surface match of the protein for the ice is important, as proposed by Jia et al. (1996) and SoÈnnichsen et al. (1996), then this accommodation would not provide a perfect ®t
523
Steric Mutants of Type III AFP
and the binding energy would de®nitely be reduced. The exact magnitude of this decrease would depend on the ®t of the residue into that pocket. Alternatively, if the side-chain cannot be accommodated by the growing ice surface, its presence will limit the ability of the ice-binding residues to match that surface. This too would create an overall lower binding af®nity. This study serves as a caveat to interpretation of mutagenic studies in the absence of structural detail. The initial study by DeLuca et al. (1996) attempted to sterically preclude the ice-binding site. The availability of the high resolution structures has provided new insight into why the steric mutants retain some antifreeze activity. An important observation is that the methyl group of Ala16 does not protrude directly from the surface as suggested by the initial NMR study (SoÈnnichsen et al., 1993), but projects with an angle of 30 off the perpendicular direction from the surface. This angle may relieve some of the steric pressure induced by mutation, since the side-chains can be de¯ected more easily. The NMR data (DeLuca et al., 1996) also indicated that there was no global alteration in the protein but did not give the detail necessary to indicate that a perturbation of the ice-binding residues had occurred. With the detailed crystal structures of the Ala16 mutants it is apparent that substitution of larger residues into position 16 did not simply result in an increase in the side-chain length of the residue. These substitutions resulted in the shifting of some of the ice-binding atoms from their wild-type position either due to a slight backbone shift around the Thr15 region or due to steric interactions between the substituted side-chain and the surrounding residues.
Conclusion The crystal structures of the Ala16 substitution mutants have afforded a new level of understanding of the effects of these mutations on thermal hysteresis activity and ice crystal morphology. From this study it may be concluded that there are multiple causes for the loss of thermal hysteresis activity, including the steric in¯uence of the side-chain on binding to the ice surface as well as the effect of perturbation of the positions of the ice-binding atoms by the new residue. This is consistent with the apparent rigidity of the ice-binding surface due to tight side-chain packing. If the residues are arranged with little ¯exibility then the Ala16 substitution by bulkier residues will necessarily affect adjacent residues. This study also supports the belief that the presence of hydrogen-bonding residues matching the ice lattice is not the sole requirement for binding AFP. The surface design incorporates numerous features essential for full antifreeze activity. The complete conservation of Ala16 between type III AFP isoforms (Davies & Hew, 1990) implies a role in maintaining the complementarity of the protein for ice. Like hydrogen bonding groups, non-hydrogen bonding groups on the ice-binding surface may also play an important role, since they may help impart the complementarity and rigidity of the ice-binding surface. The complementarity enables the match of the protein surface to the ice lattice, allowing for van der Waals contacts between the two surfaces; and the rigidity stops untethered side-chains from blocking the interaction of the hydrophilic residues with the ice surface. Any side-chain substitution of residues in this region would affect this match and reduce thermal hysteresis activity.
Table 3. Structure determination and re®nement A. Data statistics Data set A16M mutant A16H mutant A16T mutant A16C mutant A16R mutant A16Y mutant
Total reflections
Unique reflections
Completeness (%)
Rsym
1.9 1.43 1.6 1.6 1.65 1.7
19,230 40,197 28,480 19,644 26,882 30,404
4933 9650 8075 7177 7041 6956
99.1 83.9 95.2 84.5 92.5 97.8
0.077 0.065 0.059 0.062 0.067 0.075
A16M
A16H
A16T
A16C
A16R
A16Y
33.27 39.89 44.54 8.0 - 1.9 0.193 (0.261) 24
33.30 39.93 44.66 8.0 - 1.45 0.191 (0.250) 70
33.27 40.11 44.86 8.0 - 1.6 0.199 (0.26) 72
32.65 39.10 46.26 8.0 - 1.6 0.194 (0.268) 70
333.16 30.04 44.81 8.0 - 1.65 0.215 (0.280) 90
33.11 39.99 45.36 8.0 - 1.7 0.191 (0.207) 32
Ê) Resolution (A
B. Refinement Statistics Data set Ê) Unit cell a (A Ê) b (A Ê) c (A Ê) Resolution range (A Rfactor (Rfree 5%) No. of water atoms
524
Materials and Methods Sample preparation and crystallization Ala16 mutants of type III AFP were produced and puri®ed as described (DeLuca et al., 1996). Protein crystals of mutants were grown under conditions similar to those established for the wild-type protein (Jia et al., 1995) with slight variations in the ammonium sulfate concentration and pH. Table 3 summarizes the data collection statistics for the six mutant AFPs. Determination of mutant structures Ê native structure (Jia et al., 1996) was used The 1.25 A as a starting model for the re®nement of the six Ala16 mutants. Data were collected using a Mar Research imaging plate equipped with a Rigaku rotating anode generator. Data were processed using DENZO (Otwinowski, 1993) and CCP4 (1994). The mutant structures were re®ned using X-PLOR (BruÈnger, 1992). The mutated side-chain was not substituted into the structure until after the ®rst round of re®nement to prevent any model bias. The ®nal re®nement statistics for all of the mutant structures are listed in Table 3. All coordinates have been deposited in the Protein Data Bank and the accession codes are 2MSI for Ala16 > Met mutant, 3MSI for Ala16 > His, 4MSI for Ala16 > Thr, 5MSI for Ala16 > Cys, 6MSI for Ala16 > Arg and 7MSI for Ala16 > Tyr, respectively. Overlap of mutant structures The overlap of the mutant protein structures with the native structure was performed using LSQKAB in the CCP4 (1994) package. The overlapping calculation was carried out over the entire model on main-chain atoms. Since the backbone alignment is averaged over the entire protein, distances of deviation of the ice-binding atoms should not be taken as precise. They do, however, give an indication of the degree of alteration of the ice-binding site. The alternative rotamers for the substituted side-chains were obtained using the program O (Jones et al., 1990); see Figure 1 legend.
Acknowledgements We thank Sherry Gauthier for technical assistance and Steffen Graether for help in producing diagrams. This work was supported by research grants from the Medical Research Council (MRC) of Canada. Z.J. is an MRC scholar and C.I.D. was a recipient of an MRC studentship award.
References BruÈnger, A. T. (1992). X-PLOR (version 3.1) manual. A System for X-ray Crystallography and NMR, Yale University Press, New Haven. Chao, H. (1994). Structure-function relationships in type III antifreeze protein. PhD thesis, Queen's University, Kingston, Canada.
Steric Mutants of Type III AFP Chao, H., SoÈnnichsen, F. D., DeLuca, C. I., Sykes, B. D. & Davies, P. L. (1994). Structure-function relationship in the globular type III antifreeze protein: identi®cation of a cluster of surface residues required for binding to ice. Protein Sci. 3, 1760 ±1769. Cheng, C. C. & DeVries, A. L. (1991). The role of antifreeze glycopeptides and peptides in the freezing avoidance of cold-water ®sh. In Life Under Extreme Conditions (di Prisco, G., ed.), pp. 1 ± 14, SpringerVerlag, Berlin. Collaborative Computational Project No. 4. (1994). The CCP4 suite: Programs for protein crystallography. Acta Crystallog. sect. D, 50, 760± 763. Davies, P. L. & Hew, C. L. (1990). Biochemistry of ®sh antifreeze proteins. FASEB J. 4, 2460 ±2468. DeLuca, C. I. (1997). The high resolution structure of type III antifreeze protein and its implication for ice binding. PhD thesis, Queen's University, Kingston, Canada. DeLuca, C. I., Chao, H., SoÈnnichsen, F. D., Sykes, B. D. & Davies, P. L. (1996). Effect of type III antifreeze protein dilution and mutation on the growth inhibition of ice. Biophys. J. 71, 2346± 2355. Evans, S. V. (1993). SETOR: hardware lighted threedimensional solid model representations of macromolecules. J. Mol. Graph. 11, 134±138. Frishman, D. & Argos, P. (1995). Knowledge based secondary structure assignment. Proteins: Struct. Funct. Genet. 23, 566± 579. Jia, Z., DeLuca, C. I. & Davies, P. L. (1995). Crystallization and preliminary X-ray crystallographic studies on type III antifreeze protein. Protein Sci. 4, 1236± 1238. Jia, Z., DeLuca, C. I., Chao, H. & Davies, P. L. (1996). Structural basis for a globular antifreeze protein binding to ice. Nature, 284, 285± 288. Jones, T. A., Bergdol, M. & Kjaldgaard, M. (1990). O: a macromolecular modeling environment. In Crystallographic and Modeling Methods in Molecular Design (Bugg, C. E. & Ealick, S. E., eds), pp. 189± 195, Springer-Verlag, New York. Kabsch, W. & Sander, C. (1983). Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers, 22, 2577± 2637. Knight, C. A., Cheng, C. C. & DeVries, A. L. (1991). Adsorption of a-helical antifreeze peptides on speci®c ice crystal surface planes. Biophys. J. 59, 409± 418. Kraulis, J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallorg. 24, 946± 950. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283± 291. Merritt, E. A. & Murphy, M. E. P. (1994). RASTER3D (version 2.0): a program for photorealistic molecular graphics. Acta. Crystallog. sect. D, 50, 869± 873. Nicholls, A., Bharadwaj, R. & Honig, B. (1993). GRASP: graphic representation and analysis of surface properties. Biophys. J. 64, 116± 170. Otwinowski, Z. (1993). Oscillation data reduction program. Proceedings of the CCP4 Study Weekend (Sawyer, L., Issacs, N. & Bailey, S., eds), pp. 80 ± 86, SERC Daresbury Laboratory, Warrington. SoÈnnichsen, F. D., Sykes, B. D., Chao, H. & Davies, P. L. (1993). The nonhelical structure of antifreeze protein type III. Science, 259, 1154± 1157.
Steric Mutants of Type III AFP SoÈnnichsen, F. D., DeLuca, C. I., Davies, P. L. & Sykes, B. D. (1996). Re®ned solution structure of type III antifreeze protein: a possible involvement of hydrophobic groups in the energetics of the protein-ice interaction. Structure, 4, 1325± 1337. Wen, D. & Laursen, R. A. (1992a). A model for binding of an antifreeze polypeptide to ice. Biophys. J. 63, 1659± 1662. Wen, D. & Laursen, R. A. (1992b). Structure-function relationships in an antifreeze polypeptide: the role
525 of neutral, polar amino acids. J. Biol. Chem. 267, 14102± 14108. Wen, D. & Laursen, R. A. (1993). Structure-function relationships in an antifreeze polypeptide: the effect of added bulky groups on activity. J. Biol. Chem. 268, 16401± 16406. Wilson, P. W., Beaglehole, D. & DeVries, A. L. (1993). Antifreeze glycopeptide adsorption on single crystal ice surfaces using ellipsometry. Biophys. J. 64, 1878± 1884.
Edited by I. A. Wilson (Received 21 January 1997; received in revised form 17 June 1997; accepted 9 October 1997)