An Antifreeze Protein Folds with an Interior Network of More Than 400 Semi-Clathrate Waters

See allHide authors and affiliations

Science  14 Feb 2014:
Vol. 343, Issue 6172, pp. 795-798
DOI: 10.1126/science.1247407

This article has a correction. Please see:


When polypeptide chains fold into a protein, hydrophobic groups are compacted in the center with exclusion of water. We report the crystal structure of an alanine-rich antifreeze protein that retains ~400 waters in its core. The putative ice-binding residues of this dimeric, four-helix bundle protein point inwards and coordinate the interior waters into two intersecting polypentagonal networks. The bundle makes minimal protein contacts between helices, but is stabilized by anchoring to the semi-clathrate water monolayers through backbone carbonyl groups in the protein interior. The ordered waters extend outwards to the protein surface and likely are involved in ice binding. This protein fold supports both the anchored-clathrate water mechanism of antifreeze protein adsorption to ice and the water-expulsion mechanism of protein folding.

Folding When Wet

Most globular proteins release water as they fold to form a dry hydrophobic core. In contrast, Sun et al. (p. 795; see the Perspective by Sharp) report a high-resolution structure showing that the antifreeze protein Maxi retains about 400 water molecules in its core. Maxi is a dimer in which two helical monomers each bend in the middle to form a four-helix bundle. The helices are spaced slightly apart to accommodate two intersecting polypentagonal monolayers of water. The pentagons form cages around inward pointing side chains to stabilize the structure. The ordered waters extend to the protein surface where they are likely to be involved in ice binding.

A driving force for protein folding is formation of a hydrophobic core (1, 2). As aliphatic and aromatic side chains pack together in the protein core, they release constrained waters into the surrounding solvent with an overall gain in entropy that helps power the folding process. Therefore, almost all globular proteins reported to date have a dry protein core. There are two main schools of thought for how a protein’s hydrophobic core is formed. In the dewetting mechanism, waters collectively evaporate from the partially formed core. This is followed by the spontaneous collapse of the core, which stabilizes the protein by reducing the solvent-accessible surface area of core residues (3). In the expulsion mechanism, an initial structural collapse forms a near-native intermediate with a partially solvated hydrophobic core. This is followed by water expulsion from the hydrophobic core to form the native state (4). In this model, waters are thought to function as a lubricant that helps the hydrophobic core find its optimally packed state (4). Here we report the crystal structure of an antifreeze protein (AFP) with a water-rich core that offers an alternative view on the role of water in protein folding.

Antifreeze proteins are a class of proteins that adsorb to the surface of ice crystals to prevent their growth. We determined the crystal structure of the AFP Maxi, a large isoform of the 3-kD, alanine-rich type I AFP from winter flounder (Pseudopleuronectes americanus). Type I AFP is a monomeric α helix with an 11-residue periodicity (5) that binds to a pyramidal ice plane (6). Its ice-binding residues (Thr, i+4 Ala, i+8 Ala) are arrayed along one face of the helix (7). Ice-binding residues are thought to organize and stabilize ordered waters to merge with, and freeze to, the quasi-liquid layer of water next to the ice lattice (810). Maxi is five times as long as type I and forms a homodimer with no stable monomer form, but is otherwise similar in alanine richness (65%), helicity (>95% α helix), and 11-residue periodicity (1113).

The crystal structure of Maxi was determined at 1.8 Å resolution from needle-shaped crystals (table S1) grown in arginine buffer at pH 9.6. Maxi is a rodlike four-helix bundle with a length of 145 Å and an average diameter of 22 Å (Fig. 1A). Both 290 Å–long helix monomers fold exactly in the middle through 180° so that their N and C termini are side by side. In the dimer, the two hairpins are aligned in an antiparallel manner without overlap such that the four-helix bundle has a twofold rotational axis of symmetry (indicated by the central curved arrow). The two N-terminal helices lie adjacent to each other in an antiparallel orientation, as do the two C-terminal helices. At the secondary-structure level, Maxi is composed of tandem 11-residue repeats (T/IxxxAxxxAxx, where x is any residue) that each form three helical turns with an average of 3.7 residues per turn (R1-3 and R1′-3′), as opposed to 3.6 residues per turn in the classic α helix. The only departures from this pattern are the central sections of two seven-residue segments (two helical turns each) and the terminal capping regions. The capping regions comprise the N and C termini of one monomer and the hairpin loop region of the other (Fig 1B). Notably, the two chains of Maxi associate with minimal protein-protein interactions, as do the two arms of the hairpins (Fig. 1C). This is in sharp contrast to a standard four-helix bundle protein like “Repressor of primer” (Rop), where the longer aliphatic and aromatic side chains form a compact hydrophobic core with only two waters inside (Fig. 1D). Rop was selected for comparison because, like Maxi, it is also a dimer of hairpin α helices with a nonoverlapping antiparallel alignment (14).

Fig. 1 Four-helix bundle structure of Maxi has an open core.

(A) Four-helix bundle structure of Maxi in ribbon format, showing how the antiparallel hairpin monomers 1 and 2 align. N and C termini are indicated. The dyad axis of symmetry is shown by the arrow C2. Capping regions, 11-residue repeats, and center sections are labeled as Cap, R, and Center, respectively. Residues in van der Waals contact in the cap and central regions are indicated by their side chains. (B) End-on view of the capping structure of Maxi; hydrogen bonds are illustrated by black dotted lines. (C) End-on view of a cross section of Maxi in surface representation marked in (A) by the red box. Red spheres are crystallographic waters present in the 20 Å–deep section. (D) Similar-sized section through Rop, a representative four-helix bundle structure.

Direct contact between the Maxi monomers is largely limited to the capping structures and the center region. In the capping structures, Leu94 and Tyr103 near the hairpin loop of one monomer, together with Ile2 at the N terminus and Phe191 at the C terminus of the other monomer, pack together to form a local hydrophobic core with stacking of the two aromatic side chains (Fig. 1B). On the hairpin loop, Lys100 donates hydrogen bonds to cap the backbone carbonyl groups of Ala191, Ala192, and Ala194 on the C terminus, whereas the carbonyl groups of Leu94 and Ile97 and the side chain of Asn98 accept hydrogen bonds from Arg7. In the center region, Ile51 and Ile54 from both N-terminal helical arms of Maxi form a hydrophobic cluster, as does Val144 and Val148 from both C-terminal helical arms. However, the intervening helical repeats pack loosely through sparse van der Waals interactions, as shown by a representative cross section through one repeat (Fig. 1C).

The apparent loose packing of the four helices in the 11-residue repeat regions generates internal space that is just wide enough to accommodate a single layer of water (Fig. 1C and Fig. 2, A and B). The water molecules that occupy the gap form an extensive polypentagonal network around the inner-projecting residues (mostly Ala and Thr, fig. S1), with an occasional tetragonal and hexagonal water ring. In the 11-residue repeat regions, the network comprises two monolayers of polypentagonal waters (Fig. 2, A and B) that cross each other at ~90° (Fig. 3 and fig. S3). The intrachain sheet, which lies between the N-terminal arms and the C-terminal arms of Maxi, contains 19 pentagonal rings per 11-residue repeat (labeled A to R in Fig. 2A). The pattern for each repeat is almost identical, and three repeats spanning nine helical turns are shown. Occasional large holes in the sheet are generated where side chains like Thr132 and Ala68 from the same monomer form close protein-protein contacts. The interchain water web that forms the interface between the two monomers contains seven pentagonal rings (labeled S to Z in Fig. 2B). Again, a few large holes appear in this sheet due to rare interchain side-chain contacts like those between Thr122 and Ala173 from different monomers. A molecular dynamics simulation of Maxi in a box of water showed that the predicted interior water densities match extremely well with the ones in the protein core of the crystal structure (fig. S2).

Fig. 2 The interior water structure associated with the repeat regions of Maxi.

(A) Sheet of pentagonal intrachain water rings labeled A to R. Waters are showed as red spheres and hydrogen bonds by black dashes. Residues forming intrachain contacts are identified and shown in stick representation. Waters that hydrogen bond to the carbonyl groups of Maxi in R3 are numbered. (B) A sheet of interchain pentagonal waters at right angles to the sheet in (A), as indicated in the schematic diagrams on the right. Notations are the same as in (A). The pentagonal rings in this sheet are labeled from S to Z. In addition, rings G, H, and K at the intersection of two sheets are also labeled.

Fig. 3 Interior water network in Maxi.

Section through Maxi showing network of interior waters as spheres colored by association with the yellow or orange monomers. Hydrogen bonds are indicated by black dotted lines. Waters that make multiple matches to the ice lattice and are not sterically hindered by the protein are boxed in red.

Fifty years ago, Scheraga et al. theorized that a partial cage of pentagonal water rings will form when hydrophobic side chains in α helices or β sheets are separated by a single layer of water molecules (15). The formation of pentagonal rings has negative free energy when two hydrated aliphatic side chains approach each other during protein folding to a distance where they are separated by a single water layer. This proposal has been borne out in the present work, where the spaces inside Maxi’s four-helix bundle are just wide enough to allow a single layer of waters to fit in and form a semi-clathrate structure. Before this study, five pentagonal “ice-like” water clusters had been reported in crystals of crambin [Protein Data Bank (PDB) code: 1CRN], a 46-residue seed storage protein, but only as a result of intermolecular packing (16). It was predicted that these localized pentagonal clusters might occur not only at intermolecular hydrophobic contacts, but also around adjacent, hydrophobic side chains in α helices or β sheets. This prediction has been supported by a molecular dynamics simulation study of streptavidin (17), where a five-membered water ring was formed in between two hydrophobic groups in the binding cavity. The crystal structure of Maxi has revealed the natural association of pentagonal water clusters within a protein on a large scale.

Two-dimensional phases of water are produced by nanoscale confinement between nonpolar materials (nanopores) (18). These thin water layers exhibit unusual properties compared to bulk water and are typically studied by molecular simulation. Maxi contains a monolayer of amorphous ice/water between mainly hydrophobic surfaces. The pattern of the water sheet in Maxi is different from those observed by molecular simulations and experimental approaches (19, 20). This is mainly due to the inner-surface architecture of the protein. Because most nanopores have flat surfaces, water molecules inside tend to arrange themselves in layers parallel to the surface (21). By contrast, the inward-pointing side chains that form the inner surface of Maxi are molecularly rough, so the five-membered water rings form cages on individual residues, illustrating their semi-clathrate hydrate structure (Fig. 4D). In addition, unlike most nanopores used for simulations, the inner surface of Maxi interacts with about one in four of the waters through hydrogen bonds, in addition to van der Waals interactions (as described below).

Fig. 4 Water interactions with Maxi.

(A) Cross section through R2, illustrating the interior water network and the intersection of the intra- and intermolecular water sheets. Waters are shown as red spheres, and hydrogen bonds are shown as green dotted lines. (B) A section of helix from Maxi in stick format showing bifurcation of intramolecular hydrogen bonds. Hydrogen bonds between water (red spheres) and carbonyl O atoms are shown as green dotted lines; intrahelical hydrogen bonds are shown as black dashed lines. N atoms are blue and O atoms, red. (C) A section of helix from Rop in stick format showing normal intramolecular hydrogen bonds. (D) An example of semi-clathrate hydrate structure in the internal space is shown as a close-up view of the green boxed region in (A).

The 3.7-residue repeat of the helices in Maxi results in altered Φ and Ψ angles compared to those of classic α helices. This deviation likely causes the backbone carbonyl groups to be slightly tilted outwards (13), allowing them to form the key intrahelical hydrogen bonds while also hydrogen bonding with solvent waters (Fig. 4B), a duality not seen in Rop and other typical α helices (Fig. 4C). In addition, Maxi is an alanine-rich protein, and these small side chains make the backbone carbonyl groups more exposed to solvent. Therefore, most of the carbonyl groups in Maxi are involved in hydrogen-bonding interactions with waters, which helps to keep this rather hydrophobic protein highly solvated and freely soluble in flounder blood. For typical protein structures determined at 2.0 Å resolution, the number of water molecules located by crystallography is roughly equal to the number of amino acid residues (22). By contrast, in the asymmetric unit of Maxi, the ratio of water molecules to amino acid residues is almost three times this value (2.9), even though Maxi does not contain many polar residues to hydrogen bond with waters. Because most of the residues in the internal space are also alanine, hydration occurs around all four helices. Maxi uses the interior bifurcated carbonyl groups to help anchor the more than 400 internal ordered waters, ~25% of which are directly involved in backbone hydrogen-bonding interactions. The hydrogen-bonding interactions in R3 are listed in table S2.

The hydrated protein core of Maxi suggests that this protein uses the water-expulsion folding mechanism but does not complete the water-discharge step. The alanine richness of the protein core may have provided an ideal opportunity to see retention of waters during the folding. These waters serve to glue the four helical arms of Maxi together through a stabilizing network of hydrogen bonds that are anchored to backbone carbonyl groups in the interior (23). Because water-mediated protein association is less stable than direct protein association (4), this could explain why this antifreeze protein is thermolabile and irreversibly denatures at temperatures above 16°C.

In addition to the internal waters, a second unexpected feature of Maxi’s structure was that the putative ice-binding residues (Thr/Ile, i+4 Ala, i+8 Ala) occur on the inward-pointing surfaces of all four helices (colored pink in fig. S1). These residues were identified by homology to the small, monomeric type I AFP isoform that binds ice through this highly conserved hydrophobic face using regularly spaced methyl groups from the Ala and Thr/Ile (7). The same ice-binding residues are well conserved in each 11-residue repeat of Maxi and occupy equivalent quadrants of the helices throughout the molecule. Their side chains have a function similar to that of ice-binding residues because they cooperate to form and anchor the interior ordered waters that stabilize the protein structure (Fig. 4A and table S2). The average B-factor for all waters in the crystallographic asymmetric unit is 26 Å2. However, most of the waters in the inner network have B-factors that are lower than 20 Å2 (colored blue and green in fig. S4). Outer waters (colored red) are more disordered.

Three lines of evidence suggest that the crystal structure of Maxi is representative of the solution form and that the protein does not undergo a conformational change to bind to ice crystals through the Thr/Ile, i+4 Ala, i+8 Ala residues. First, Olijve et al. have determined the solution structure of Maxi at low temperature using small-angle x-ray scattering (24). In solution, Maxi has a cylindrical shape and dimensions that are consistent with the four-helix bundle crystal structure and eliminate a previously proposed model where the two helices form a fully extended helix dimer (13). Second, the measurement of intrinsic fluorescence transfer between Tyr and Phe suggests that the exquisite capping structure seen in the crystal (Fig. 1B) is also present in solution (fig. S3). Lastly, cross-linking experiments show that the juxtaposition of residues on neighboring chains of the helix bundle is the same in solution as it is in the crystal (figs. S4 and S5). Moreover, where these cross-links prevent the opening of the helix bundle to expose the “ice-binding residues,” there is no appreciable loss of antifreeze activity (table S3)

How then does Maxi bind to ice? Close examination of Maxi’s crystal structure shows positioned waters extending outwards between all four helices from the core to the surface. At the periphery, they form a network of ordered waters that are unobstructed by the helices and available to merge and freeze with the quasi-liquid layer on the surface of ice (810). It is possible to fit clusters of crystallographic surface waters at the face formed by the N- and C-terminal helices (boxed regions in Fig. 3) into the ice lattice on numerous planes, three of which are shown in fig. S6, A to C. Consistent with this result, when we used fluorescence-based ice plane affinity analysis (25) to determine which ice planes adsorb Maxi, all surfaces of the single ice-crystal hemisphere were bound by the fluorescently tagged AFP (fig. S7D). Unlike other AFPs characterized to date, which have ice-binding residues located on their surface, Maxi is the only one in which these residues are buried inside. This further supports the anchored-clathrate water mechanism by which AFPs adsorb to ice (10) because it suggests that this AFP cannot directly bind to its ligand but must do so through the ordered surface waters.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S10

Tables S1 to S3

References (2635)

References and Notes

  1. Acknowledgments: We are grateful to D. McLeod and K. Munro of the Protein Function Discovery Facility at Queen’s University for mass spectrometry and intrinsic fluorescence analyses, respectively. We thank E. Lazo and V. Stojanoff of the X6A beamline at Brookhaven National Laboratories for help with x-ray data acquisition and processing. This research was funded by the Canadian Institutes of Health Research. P.L.D. holds the Canada Research Chair in Protein Engineering. J.S.A holds a Canada Research Chair (Tier 2) in Structural Biology and is an Ontario Early Researcher Award recipient. The coordinates and x-ray data for Maxi have been deposited with the PDB with accession code 4KE2.
View Abstract

Navigate This Article