Marine teleosts at high latitudes can encounter ice-laden seawater that is approximately 1 degrees C colder than the colligative freezing point of their body fluids. They avoid freezing by producing small antifreeze proteins (AFPs) that adsorb to ice and halt its growth, thereby producing an additional non-colligative lowering of the freezing point. AFPs are typically secreted by the liver into the blood. Recently, however, it has become clear that AFP isoforms are produced in the epidermis (skin, scales, fin, and gills) and may serve as a first line of defense against ice propagation into the fish. The basis for the adsorption of AFPs to ice is something of a mystery and is complicated by the extreme structural diversity of the five antifreeze types. Despite the recent acquisition of several AFP three-dimensional structures and the definition of their ice-binding sites by mutagenesis, no common ice-binding motif or even theme is apparent except that surface-surface complementarity is important for binding. The remarkable diversity of antifreeze types and their seemingly haphazard phylogenetic distribution suggest that these proteins might have evolved recently in response to sea level glaciation occurring just 1-2 million years ago in the northern hemisphere and 10-30 million years ago around Antarctica. Not surprisingly, the expression of AFP genes from different origins can also be quite dissimilar. The most intensively studied system is that of the winter flounder, which has a built-in annual cycle of antifreeze expression controlled by growth hormone (GH) release from the pituitary in tune with seasonal cues. The signal transduction pathway, transcription factors, and promoter elements involved in this process are just beginning to be characterized.

Four distinct macromolecular antifreezes have been isolated and characterized from different marine fish. These include the glycoprotein antifreezes (Mr 2.5-33 K), which are made up of a repeating tripeptide (Ala-Ala-Thr)n with a disaccharide attached to the threonyl residues, and three antifreeze protein (AFP) types. Type I is an alanine-rich, amphiphilic, alpha-helix (Mr 3-5 K); type II is a larger protein (Mr 14 K) with a high content of reverse turns and five disulfide bridges; and type III is intermediate in size (Mr 6-7 K) with no distinguishing features of secondary structure or amino acid composition. Despite their marked structural differences, all four antifreeze types appear to function in the same way by binding to the prism faces of ice crystals and inhibiting growth along the a-axes. It is suggested that type I AFP binds preferentially to the prism faces as a result of interactions between the helix macrodipole and the dipoles on the water molecules in the ice lattice. Binding is stabilized by hydrogen bonding, and the amphiphilic character of the helix results in the hydrophobic phase of the helix being exposed to the solvent. When the solution temperature is lowered further, ice crystal growth occurs primarily on the uncoated, unordered basal plane resulting in bipyramidal-shaped crystals. The structural features of type I AFP that could contribute to this mechanism of action are reviewed. Current challenges lie in solving the other antifreeze structures and interpreting them in light of what appears to be a common mechanism of action.