W hat is the toughest recognition problem in biology? Arguably, it is that faced by antifreeze or thermal hysteresis proteins. These proteins provide a broad range of organisms with protection against freezing damage by depressing, in a noncolligative manner, the freezing point of water. They do this by binding to nascent ice nuclei and inhibiting their growth (1). In other words, antifreeze proteins (AFPs) must distinguish one phase of water, ice, from another phase, liquid. Moreover, the latter is present in great excess at 55 M. There are no chemical differences to key off, just the subtle structural differences, still poorly characterized, that exist between the surface of ice nuclei and liquid water. In addition to the intrinsic interest of the ice recognition problem, AFPs have application to cryopreservation of tissues and organs (2) and to the food industry (3). The work of Davies and colleagues (4) in PNAS represents a significant step forward in understanding the detailed mechanism of AFP action (i.e., a first look, if you will, at an AFP in the act of discriminating between different structural states of water). To put this work into perspective, it is useful to have some background on the study of AFPs, which has a number of paradoxical elements. First, there is the sheer exuberance of structural motifs found in AFPs. AFPs have been isolated and structurally characterized from a wide range of organisms from bacteria, through insects to arctic fish. Fig. 1 illustrates just 3 of the 12 or so different structural motifs discovered in AFPs to date. Each is oriented so that the ice-binding surface (IBS) is uppermost. Reading from the top, there is a small globular protein from sea pout, a single α-helix from winter flounder, and a stack of left-handed PP-II helices from the snow flea. The work of Garnham et al. (4) in PNAS gives a fourth example. MpAFP from the Antarctic bacterium Marinomonas primoryensis adopts a righthanded, parallel β-helix. Discovery of a previously undescribed AFP quite often reveals a unique protein motif. A case in point is the snow flea AFP, discovered by Graham and Davies (5). Structure determination, as described by Pentelute et al. (6), revealed the unique motif in Fig. 1. In a fascinating side note, this was made possible through generation of racemic protein crystals using total chemical synthesis. This simplified the crystallographic solution. Because AFPs recognize an achiral “substrate,” they are one of the few proteins in which both enantiomers would have identical biological activity. Fig. 1 also shows that there is little uniformity of AFP structure within broad classes of organism: Two very different structures come from ocean fish. Thus, the identical function, depression of the freezing point, can be manifested by widely different protein motifs. This raises a number of interesting questions. Do these AFPs all work by the same mechanism? Do they bind to the same plane(s) Fig. 1. ThreeAFP structural motifs. The IBS is oriented uppermost. (Top) Type III AFP from ocean pout (PDB ID code 1MSI) (18). Alanine 16, a key residue in the IBS, is indicated in stick representation. (Middle) Type I AFP fromwinterflounder (PDB ID code 1WFA) (19). (Bottom) AFP from the snowflea (PDB ID code 2PNE) (6).