M. M. Harding, School of Chemistry, The University of Sydney, NSW 2006, Australia. Fax: + 61 29351 6650, E-mail: firstname.lastname@example.org
Antifreeze glycoproteins (AFGPs) constitute the major fraction of protein in the blood serum of Antarctic notothenioids and Arctic cod. Each AFGP consists of a varying number of repeating units of (Ala-Ala-Thr)n, with minor sequence variations, and the disaccharide β-d-galactosyl-(1→3)-α-N-acetyl-d-galactosamine joined as a glycoside to the hydroxyl oxygen of the Thr residues. These compounds allow the fish to survive in subzero ice-laden polar oceans by kinetically depressing the temperature at which ice grows in a noncolligative manner. In contrast to the more widely studied antifreeze proteins, little is known about the mechanism of ice growth inhibition by AFGPs, and there is no definitive model that explains their properties. This review summarizes the structural and physical properties of AFGPs and advances in the last decade that now provide opportunities for further research in this field.
High field NMR spectroscopy and molecular dynamics studies have shown that AFGPs are largely unstructured in aqueous solution. While standard carbohydrate degradation studies confirm the requirement of some of the sugar hydroxyls for antifreeze activity, the importance of following structural elements has not been established: (a) the number of hydroxyls required, (b) the stereochemistry of the sugar hydroxyls (i.e. the requirement of galactose as the sugar), (c) the acetamido group on the first galactose sugar, (d) the stereochemistry of the β-glycosidic linkage between the two sugars and the α-glycosidic linkage to Thr, (e) the requirement of a disaccharide for activity, and (f) the Ala and Thr residues in the polypeptide backbone. The recent successful synthesis of small AFGPs using solution methods and solid-phase chemistry provides the opportunity to perform key structure-activity studies that would clarify the important residues and functional groups required for activity.
Genetic studies have shown that the AFGPs present in the two geographically and phylogenetically distinct Antarctic notothenioids and Arctic cod have evolved independently, in a rare example of convergent molecular evolution. The AFGPs exhibit concentration dependent thermal hysteresis with maximum hysteresis (1.2 °C at 40 mg·mL−1) observed with the higher molecular mass glycoproteins. The ability to modify the rate and shape of crystal growth and protect cellular membranes during lipid-phase transitions have resulted in identification of a number of potential applications of AFGPs as food additives, and in the cryopreservation and hypothermal storage of cells and tissues.
Many plants, insects, animals and other organisms have evolved with unique adaptive mechanisms that allow them to survive in harsh environments at the extremes of temperature [1–4]. Nearly two-thirds of the surface of the earth is comprised of water, with the average surface temperature of seas and oceans varying from −2 °C to 30 °C depending on latitude . Within the polar regions, seawater temperatures are consistently below the freezing point of physiological solutions, which themselves have freezing points below the freezing point of pure water, 0 °C at 1 atmosphere, due to dissolved sugars and salts. The effect of these subzero temperatures on the cells of plants, animals, bacteria and fungi can be extremely harmful, if not deadly .
Scholander [6,7] and DeVries [8,9] were the first to investigate the mechanisms by which species of fish inhabiting the polar oceans at temperatures that are frequently below that of the freezing point of pure water, are able to survive. Analysis of the blood plasma of these fish showed that while the concentrations of salts and small ions in the body fluids are somewhat higher relative to fish in temperate waters, these salts are only responsible for 40–50% of the observed freezing point depression. The remainder of the protective effect was attributed to the presence of a series of relatively high molecular mass glycoproteins and proteins [10–13].
‘Antifreeze’ proteins (AFPs) and ‘antifreeze’ glycoproteins (AFGPs) have since been identified in the body fluids of many species of polar fish. Four classes of structurally diverse AFPs, classified as type I [14,15], type II [16,17], type III [18,19] and type IV [20,21] have now been identified along with a single class of glycosylated protein denoted AFGP [22–24]. The principal characteristics of these compounds, which are compared in a number of articles [3,25–27], are summarized in Fig. 1. In contrast to many solutes, these compounds kinetically depress the temperature at which ice grows in a noncolligative manner, and hence exhibit thermal hysteresis, i.e. a positive difference between the equilibrium melting point and the ice growth temperature (the temperature at which seed ice crystals will grow in the solution). This property allows fish to survive in the subzero waters at temperatures colder than the equilibrium freezing point of their blood and other internal fluids, by modifying or suppressing ice crystal growth and by protecting cell membranes from cold-induced damage [3,28]. These versatile properties have attracted significant interest for their potential applications in medicine and industry where low temperature storage is required and ice crystallization is damaging . Applications include improved protection of blood platelets and human organs at low temperatures , increasing the effectiveness of the destruction of malignant tumors in cryosurgery , and improvement of the smooth texture of frozen foods .
Most research has focused on the type I AFPs and a number of reviews summarizing progress in this area have been published [3,25–27,33–35]. Studies of the more complex type II and III AFPs are now being addressed [19,36–40]. In contrast to the AFPs, the AFGPs present in cold water fish have been much less studied. This is due to their structural complexity compared to AFPs (Fig. 1), and the difficulties in accessing sufficient quantities of pure material to allow detailed studies to be performed.
This review will focus on new research published in the last decade on AFGPs. Several reviews have already summarized the AFGP literature published in the 1970s and 1980s [1,41–44] and hence this work will be only briefly mentioned in this article. Recent new insights into the mechanism of action of type I AFPs, as well as studies on type II and III AFPs, have provided new clues about the crucial interactions that occur between AFPs and the ice/water interface, which need to be considered in the mechanism of action of AFGPs. Other recent progress that is significant in the field includes detailed characterization of the solution conformation of AFGPs, the development of methodology to allow the production of synthetic AFGPS, and molecular evolutionary studies on the origin of AFGPs.
Structure and classification of glycoproteins
Antifreeze glycoprotein is a collective name that has been used widely in the literature to refer to a group of at least eight structurally related glycoproteins that constitute the major fraction of protein in the blood serum of Antarctic notothenioids and Arctic cod. Each AFGP consists of a number of repeating units of (Ala-Ala-Thr)n, with minor sequence variations and the disaccharide β-d-galactosyl-(1→3)-α-N-acetyl-d-galactosamine joined as a glycoside to the hydroxyl oxygen of the Thr residues (Fig. 2A). The glycoproteins isolated from the notothenioids  have been further classified as AFGP1–8 on the basis of their relative rates of electrophoretic migration . There are eight distinct classes of glycopeptides, which range in relative molecular mass from 33.7 kDa (n = 50) to 2.6 kDa (n = 4) (Fig. 2A). For convenience these are generally further classified as large (AFGP1–5) and small (AFGP6–8).
In addition to these molecular mass size variations, there is some minor difference in the amino-acid composition in AFGPs 6–8 in which the first Ala in some of the repeats is replaced by Pro (Fig. 2B) [11,46,47]. Thus, while the notothenioid AFGPs have a simple primary structure, they exhibit significant size and some amino-acid variation.
AFGPs have also been identified in several Arctic and north Atlantic cods [48–50]. These glycoproteins are remarkably similar to those present in the unrelated notothenioids, with the exception that Thr is occasionally replaced by an Arg residue (Fig. 2C) and hence the glycopeptide lacks a disaccharide at this position.
While Fig. 2 shows the most common AFGP structures, there is evidence that further amino-acid substitution can be tolerated. A novel AFGP containing the carbohydrate residue N-acetylglucosamine and the amino acids Asn, Gln, Gly, Ala and traces of Arg, Val, Leu and Thr has been isolated from the Antarctic fish species Pleuragramma antarcticum.
The general abbreviation AFGP has been widely used in the literature although many other intermediate sizes of glycoproteins than those shown in Fig. 2 have been identified as a result of better protein resolution techniques . This fact has been highlighted in a recent study in which AFGPs were isolated and purified from the blood plasma of the rock cod Gadus ogac with additional purification and characterization using electrospray mass spectrometry . This allowed more accurate mass identification and showed multiple isoforms for AFGPs within a particular mass range. For example, glycoproteins classified as AFGP6 on the basis of their overall molecular mass, were further subdivided into two mass fractions of 6026–9784, containing 14 different isoforms and 3865, which contained a single sequence. Thus, the abbreviations AFGPx (x = 1–8) does not always refer to a single compound, but in many cases a mixture of glycopeptides in an approximate mass range.
The use of the generic term AFGP to refer to all of the structures represented in Fig. 2 has led to confusion in some literature reports where it is not clear whether a pure glycoprotein or a mixture of different molecular mass glycoproteins have been used. In addition, as studies are now addressing the molecular level mechanism of ice growth inhibition, the exact amino-acid composition is also important, and the presence of any minor sequence variations in the Ala-Ala-Thr backbone needs to be established. Hence we propose an expanded list of abbreviations (Fig. 2) in order to clarify the amino-acid composition of the glycoprotein being studied. As the exact number and positions of the Pro and Arg residues in AFGP-Pro and AFGP-Arg are frequently unknown, these abbreviations simply subclassify whether the tripeptide repeat is constant Ala-Ala-Thr or not. For example, AFGP-Arg8 would refer to a tripeptide repeat where n = 4 with Arg substituted for some of the Thr residues, and an approximate molecular mass of 2.7 kDa. In glycoproteins in which the exact number and positions of the Pro or Arg residues are known, a full sequence and unique abbreviation is required.
Origin and evolution of glycoproteins
Table 1 summarizes the phylogenetic relationship of teleost fish that produce AFGPs, adapted from Cheng . AFGPs have been isolated from both Antarctic notothenioid fish as well as from a northern gadid in the Labrador, the rock cod, Gadus ogac and other high-latitude northern cods belonging to the family Gadidae[43,55]. The most studied AFGPs are from the Antarctic fish, Trematomas borgrevinki and Dissostichus mawsoni, and from a northern fish, Boreogadus saida. In both Trematomas borgrevinki and Dissostichus mawsoni the total AFGP concentration is about 25 mg·mL−1 of which approximately 25% is due to AFGP1–5 with the remaining 75% containing the smaller AFGP6–8.
Table 1. Summary of phylogenetic relationship of teleost fish that produce AFGPs adapted from Cheng .
A long standing issue regarding the evolutionary origin of AFGPs was recently resolved in elegant work by Chen, DeVries and Cheng [52,56,57]. The high degree of structural similarity between AFGPs found in the two geographically and phyologenetically distinct Antarctic notothenioids and Artic cods (Table 1) has been noted for many years. Chen et al. showed that the AFGP gene from the Antarctic notothenioid Dissostichus mawsoni derives from a gene encoding a pancreatic trypsinogen via a unique mechanism that does not involve the more common recycling of existing protein genes. The novel portion of the AFGP gene which encodes the ice-binding function derives from the recruitment and iteration of a small region spanning the boundary between the first intron and second exon of the trypsinogen gene. Expansion and iterative duplication of this new segment produces 41 tandemly repeated segments, with sequences at either end that are nearly identical to trypsinogen. The small sequence divergence between notothenioid AFGP and trypsin genes indicates that the transformation of the protein gene into the novel ice-growth inhibition gene occurred about 5–15 million years ago, which is consistent with the estimated times of freezing of the Antarctic Ocean. This conversion is unique and shows how an old protein gene spawned a new gene for an entirely new protein with a new function.
In a related study, the sequence for the Arctic cod, Boreogadus saida was compared with the notothenioid gene . While the Boreogadus saida AFGP genes have a similar polyprotein structure to the notothenioid genes in which multiple copies of the AFGP coding sequences are linked by small cleavable spacers, molecular evidence from detailed comparative analyses argue strongly for independent evolution of the cod AFPG genes. This evidence includes (a) different signal peptide sequences, (b) different spacer sequences that link the encoded AFGP molecules in the polyprotein, invoking different mechanisms of processing of the polyprotein precursors, (c) distinct codon bias of the nine nucleotide sequence for the AFGP tripeptide, and (d) different genomic loci of the AFGP gene loci in the cod and notothenioid AFGPs. Thus, the near-identical AFGPs of these two unrelated fish is a rare example of protein sequence convergence, i.e. the development of a similar protein from different parents under similar environmental pressure. Furthermore these studies established that every AFGP isoform is distinctly encoded as individual copies within polyprotein genes, i.e. the various lengths of AFGPs shown in Fig. 2A are not due to protein processing through splicing small AFGPs or cleaving large ones into small ones. The high concentration of the AFGPs in blood (35 mg·mL−1) also suggest that a large family of polyprotein genes must escalate the gene dosage.
AFGPs accumulate at certain faces of the ice/water interface, and modify the rate and shape of crystal growth. The terms ‘antifreeze’ activity, ice growth inhibition and hysteresis, and definitions and labelling of the different ice planes are illustrated in our earlier review of type I AFPs .
A characteristic property of AFPs and AFGPs is thermal hysteresis, which is determined by measurement of the kinetic ice growth point and subtraction of the equilibrium melting (= freezing) point of a solution . In the presence of an AFGP, the measured melting point depression is as expected on the basis of colligative properties, i.e. it is proportional to the molar fraction of molecules in solution. The depression of the ice growth point (the temperature at which ice starts to grow from a seed ice crystal) is, however, very much greater than this. Figure 3 shows the concentration-dependent thermal hysteresis exhibited by AFGPs, the magnitude of which depends on the length of the polymer chain. Maximum hysteresis is observed with AFGP1–5, compared with the lower molecular mass AFGP6–8 . These values are comparable to the thermal hysteresis exhibited by many type I AFPs .
Other phenomena associated with ice growth inhibition include accumulation at specific faces of the ice crystal, detected by hemisphere etching , and modification of the crystal habit when ice is grown in a thermal gradient. Ice may exist in many polymorphic forms, with ice 1 h the most stable form at 1 atmosphere below zero °C. The hexagonal ice 1 h lattice unit may be characterized by four axes, a1, a2, a3 and c with the surface of the hexagonal unit comprising eight faces, two basal faces normal to the c-axis and six prism faces . As it is normal to the c-axis, the basal face is known as the c-face or (0001). Directions and vectors within the ice lattice are also described in terms of the four axes and are distinguished by the types of brackets that encloses the coordinates. For example, designates the group of 12 equivalent surfaces of a hexagonal bipyramid of which is one specific surface.
Raymond et al. showed that single ice crystals suspended in solutions of AFGP1–5 at temperatures within the hysteresis gap form hexagonal pits on the basal plane, while in the presence of AFGP7–8, c-axis growth occurred to a greater extent and the edges of the basal plane formed bipyramidal faces . Figure 4 illustrates the effect of blood serum from Dissostichus mawsoni on ice crystal growth, showing the formation of the ‘pits’ on a flat basal surface of a growing seed crystal of ice 1 h. The equilibrium melting/freezing point of the solution is measured to be −1.21 °C. The three images are taken approximately 30 s apart, in order to show the growth of hexagonal pits. These pits eventually cover the entire exposed surface of ice, which then stops growing (even though below the equilibrium melting/freezing temperature of the solution) until the temperature is decreased even further, well below temperatures the fish encounter in the ocean. This modification of the ice crystal habit by AFGPs is quite different to the AFPs which typically inhibit growth along the a-axis resulting in accelerated growth primarily along the crystallographic c-axis to give bipyramidal crystal forms .
Using hemisphere etching, a simple test to determine which crystal planes (if any) are recognized by a compound, AFGP7 and AFGP8 were shown to accumulate at the primary prism planes by Knight [61,62], while at very low concentrations (<0.03 mg·mL−1) AFGP1–5 accumulate at the plane, changing to the plane at higher concentrations . Elegant ellipsometry measurements with AFGP7 and AFGP8 have shown that that the AFGPs accumulate at the basal and prism planes of single ice crystals . The particular faces at which specific AFGPs accumulate were determined elegantly by Knight and colleagues .
Effect of molecular mass
As shown in Fig. 3, the molecular mass of the different AFGPs is important with the longer polymers (AFGP1–5) having enhanced thermal hysteresis properties compared to the shorter polymers (AFGP6–8). The small molecular mass forms (AFGP7 and 8) comprise most of the circulating antifreeze  but show only two-thirds of the antifreeze activity of the larger molecular mass AFGPs . However, comparison of the effect of a synthetic dimer of AFGP6 to the monomeric AFGP6 did not show substantially greater activity when the molecular mass was doubled . The dimer was prepared by carbodiimide coupling of methylated AFGPs, followed by HPLC purification and cleavage of the O-acyl bonds. The synthetic dimer contains a different peptide sequence to the natural AFGP with a Pro following three Ala residues. This sequence may affect the AFGP conformation and hence the ability to inhibit ice growth resulting in no increase in thermal hysteresis.
A more rigorous study of the hysteresis values of a series of highly purified AFGPs from the rock cod Gadus ogac showed that they could be grouped into two distinct classes. AFGPs with molecular mass >13 kDa gave approximately three to four times higher hysteresis values than the smaller Pro–containing AFGPs (molecular mass <10 kDa) .
The hysteresis values of AFGP from cod has been compared with the values obtained for different AFPs . Due to the structural differences between the AFGPs and AFPs (see Fig. 1) these results are not directly comparable, but in terms of molecular mass it was noted that the type I AFPs from the winter flounder and shorthorn sculpin had greater activity than did glycoproteins of similar size. However AFGPs with a molecular mass of 10 kDa or higher had activities which exceeded those of any known AFP.
Structural modification of sugars
There are limited studies on the structural requirements of the disaccharide that are required for activity. This is directly related to difficulty in the synthesis of AFGPs and derivatives (discussed in a later section). Hence the only data available is on standard carbohydrate degradation studies. The key derivatives that have been prepared, and the effects of these structural modifications are summarized in Fig. 5. However, it should be noted that in most cases the derivatives were not isolated and purified.
The glycopeptide structure is important as β-elimination of the saccharides and loss of the Thr hydroxyl functionality removes all antifreeze activity [23,68]. Acetylation of the sugar hydroxyls to give derivative 1, or periodate oxidation of the terminal galactose sugar to give derivative 2, removed the hysteresis properties of the AFGP consistent with the requirement of at least some, if not all, of the hydroxyls on the galactose sugar . Oxidation with d-galactose oxidase to give the bisaldehyde 3 had no effect on hysteresis showing that the hydroxyl group at C6 of galactose is not essential for activity. However, conversion of the newly formed aldehydes to negatively charged groups by oxidation to the acid 5 or by addition of bisulfite to give 4 removed activity [22,69]. Thus the type of functional groups present on the sugars are important and the loss of activity of both 4 and 5 compared with 3 suggest that the negative charge is not tolerated. Addition of 0.15 m sodium borate to the AFGP eliminated hysteresis. This reagent complexes cis-hydroxyl groups and hence gives rise to a mixture of products including 6a and 6b. This reaction is pH dependent and could be reversed to give fully active AFGP [22,70].
More recently, oxidation of the C-6 hydroxyls of a mixture of AFGPs from Pagtothenia borchgrevinki to the aldehyde with galactose oxidase and catalase, produced peptides with an average of 75% of the activity of the native AFGP, but with some batches dropping to 30% activity . These peptides were then reductively alkylated with a variety of amino acids or short peptides and cyanoborohydride and the antifreeze activity reported relative to the oxidized starting material. Glycopeptides with Gly to (Gly)4 substitution all contained activity >60% that of the oxidized starting material, indicating that bulky substitution at the C-6 position is not detrimental to activity while the lowest activities were reported for the Gly-Glu (13%), Gly-Gly-Phe (30%) and Arg (30%) derivatives.
Taken together, these degradation results support a requirement for at least some of the sugar hydroxyl groups for activity. The C-6 hydroxyl group does not appear to be required for activity, and the C6-position tolerates a range of substituents with the exception of charged groups. The importance of the following structural elements for antifreeze activity has not been established: (a) the number of hydroxyls required, (b) the stereochemistry of the sugar hydroxyls (i.e. the requirement of galactose sugars), (c) the acetamido group on the first galactose sugar and (iv) the stereochemistry of the β-glycosidic linkage between the two sugars and the α-glycosidic linkage to Thr, and (d) the requirement of a disaccharide for activity.
Modification of the peptide backbone
As discussed above and shown in Fig. 2, the most common tripeptide in AFGPs is Ala-Ala-Thr, while in the smaller glycopeptides Pro or Arg substitutes occasionally for Ala. A series of glycopeptides of approximately the same molecular mass but containing different amounts of Pro and Arg were prepared by Edman degradation of AFGPs isolated from different species of fish . The very similar hysteresis values measured on solutions of these different AFGPs at a range of concentrations indicate that the amino-acid composition does not have a significant effect on noncolligative freezing point depression. Of note is the fact that substitution of Arg for Thr removes the disaccharide from one of the tripeptide units, but this structural modification does not affect antifreeze activity. However, a systematic study of the number of substitutions of Ala for Pro or Thr for Arg that can be tolerated in a given molecular mass AFGP has not been carried out. Cleavage of the peptide backbone with subtilopeptidase A, as expected, removed activity .
There have been no systematic investigations into the role of the Ala and Thr residues in activity. Outstanding questions include whether Ser could be substituted for Thr, which would simplify synthetic production of AFGP analogues, and the role of the Ala sidechains. The evolution of Ala in the tripeptide could be due to the hydrophobic nature of the sidechain, or the small sidechain, which has unique secondary structure preferences. It is interesting to note that in the type I AFPs, Thr and Ala are critical amino acids required for activity in these α-helical proteins. In particular, the β-methyl group of Thr, along with surrounding hydrophobic residues, including Ala, provide a hydrophobic face of the helix which is oriented towards the ice/water interface . The role of hydrophobicity in the mechanism of action of AFGPs has not yet been considered (discussed in a later section) but in this context, the effect of mutation of the Thr and Ala residues in the Ala-Ala-Thr tripeptide repeats of AFGPs would be highly informative.
Synthesis of antifreeze glycoproteins
AFGPs are presently only available from natural sources in limited amounts. Difficulties in isolation from natural sources in analytically pure quantities for commercial development, as well as the fact that harvesting of fish is necessary, require the development of an alternative source of compounds. Hence recent research has focused on the development of an efficient synthetic route to AFGPs and analogues. The synthesis of glycoproteins and carbohydrates is significantly more demanding than for protein synthesis. While automated solid-phase peptide synthesis or molecular biology techniques allow the routine production of AFPs, as well as the incorporation of mutations and isotopic labels into AFP sequences (see for example [27,73]), these methods are not widely applicable to the preparation of AFGPs.
The first and only synthesis of a naturally occurring AFGP was reported in 1996 and is summarized in Fig. 6A. The key glycotripeptide was polymerized using diphenylphosphoryl azide to give a polymer with an estimated molecular mass of 6000–7300, i.e. n = 10–12. A full paper describing the experimental details and testing of the synthetic AFGPs for activity has not been reported. In principle, modification of this synthetic scheme should allow the production of synthetic AFGP analogues in which the number and relative stereochemistries of the hydroxyls are varied in each sugar, and hence provide access to compounds which would allow key structure activity studies to be performed. Other potential synthetic routes to AFGPs which involve glycosidation of Thr as the last step in the synthesis [75,76] are currently restricted to model tripeptides.
An alternate route to high molecular mass AFGPs using solid-phase peptide synthesis has recently been reported (Fig. 6B) using Fmoc-chemistry and standard protecting groups to produce AFGPs where n = 4 and 8 . Related solid-phase synthesis [78,79] of AFGPs containing single sugars should allow access analogues for structure-activity studies.
The advantage of using solid-phase synthesis (Fig. 6B) is the ability to generate oligomers of defined length and sequence variation, including mutations of the Ala residues at one or more sites in the sequence and modification of the structure of each sugar by the preparation of a different Fmoc-protected building block. In contrast, the solution phase route (Fig. 6A) will always produce mixtures of oligomers which need to be separated, and require the use of a single tripeptide unit for the polymerization reaction.
Given the synthetic difficulties outlined above, analogues of AFGPs that are synthetically more accessible by the replacement of Thr with Lys and the formation of the more stable C-glycosides in place of O-glycosides have been reported [80,81]. However, the effect of these drastic structural modifications on hysteresis has not been published.
A detailed knowledge of the solution conformation of AFGPs is clearly essential in understanding the molecular mechanism of ice growth inhibition. A range of techniques have been used to study the solution conformation of different AFGPs including CD, Raman spectroscopy, light scattering measurements and NMR spectroscopy.
Early CD studies of one AFGP  concluded that the compound had a random coil conformation. Due to the similarity of the CD spectrum of a random coil and a left-handed 3-residue-per-turn helix, the temperature dependence of the CD spectra of the AFGP from Trematomus borchgrevinki and Eliginus gracilis were measured . The lack of a sharp transition in the spectra was consistent with an unordered conformation in solution. Independent CD studies , quasielastic light scattering  and Raman spectroscopy measurements  all suggested the presence of some folded structure.
Natural abundance 13C NMR spectroscopy of an aqueous solution of AFGP3–6 from Dissostichus mawsoni, including measurement of relaxation times, nOes and variable temperature experiments were consistent with the AFGPs existing as predominantly flexible random coil polymers . Early 1H NMR data (300 MHz) of AFGP1–4  provided a more detailed picture of the conformation and, along with conformational energy calculations, it was proposed that the hydrophobic surfaces of the disaccharide side chains are wrapped closely against a threefold left handed helical backbone. A comparison of the solution conformation of AFGP1–4 and AFGP8 suggested that both AFGPs adopt similar conformations , and hence the differences in their ice growth inhibition properties (see Fig. 3) are not due to a structural difference. 2D NMR studies (300 MHz)  allowed further refinement of this data and measurement of amide exchange rates which ruled out significant strong hydrogen bonding involving the amide protons in aqueous solutions. Comparison of AFGP amide vibrational frequencies with those observed and calculated for beta and gamma-turns in other peptides suggests that AFGPs contain substantial turn structure  while NMR studies on model glycopeptides showed an intramolecular hydrogen bond between the amide proton of N-acetylgalactosamine and the carbonyl oxygen of the Thr to which the sugar is attached .
The most detailed insight into the global conformation of AFGPs has been provided from two recent complementary papers from the same group [92,93]. A combination of high field NMR (500 MHz) and IR spectroscopies, along with molecular dynamics calculations were performed on the 14 amino-acid residue Thr-Pro-Ala glycoprotein AFGP8 (i.e. AFGP-Pro8 in Fig. 2B), and a mixture of AFGP1–5, which contains no Pro residues. While AFGP-Pro8 has no long-range order, it displays significant local order. In contrast, AFGP1–5 was reported to be a dynamically disordered molecule that shows neither significant long or short range order. The somewhat unexpected result that AFGP-Pro8 lacks long range order [92,93], has prompted a closer study of this Pro-containing AFGP. Using an initial model derived from 10 NMR structures, molecular dynamics simulations along with free energy calculations using a continuum solvation model were performed to gain insight into the nature of the conformations and motions in this AFGP-Pro8 . While the presence of the Pro residues does induce adoption of a poly proline helix, the glycoprotein exists in a number of structurally distinct, but energetically equivalent conformers. Hydrogen bonding between the N-acetyl groups and the peptide backbone were also identified as making a significant contribution to the overall stability of the AFGP .
13C NMR spectroscopy and FTIR spectroscopy have been used to probe the dynamics and conformations of an N,N-dimethylated AFGP from the Greenland cod in the presence of ice . Overall the study concluded that the AFGP adopts a similar type of three-dimensional fold in the presence of ice and in a freeze-dried state but, as in related studies, the molecule is highly flexible accessing a large number of conformers.
Despite these recent NMR studies, no three-dimensional solution structure of any AFGP has been published. The torsional flexibility in the sequence, as well as the fact that a large number of conformers are available do not allow a definitive structure to be produced. This contrasts with the well-defined secondary and tertiary structures present in the type I-IV AFPs (Fig. 1).
There is currently no mechanism that explains the ice growth inhibition properties of AFGPs. Just as with early proposed models for the mechanism of type I AFPs (summarized in ), in the case of AFGPs many erroneous conclusions were drawn from vacuum/ice models at the absolute zero of temperature, which have little or nothing in common with ice/water interfaces at or near the melting point. A hydrogen-bonding dominated mechanism, that involves insertion of the disaccharide hydroxyls of AFGPs into the vacuum/ice lattice, was proposed by analogy with a model for the type I AFP from the winter flounder that relied on hydrogen bonding involving the hydroxyl groups in the Thr residues . However, structure-activity studies have now shown conclusively that hydrophobic interactions provided by the β-methyl group of the Thr residues are crucial to the ice growth inhibition mechanism in type I AFPs [73,96–99]. Lavalle, DeVries and colleagues  have recently studied adsorption of AFGP1–5 on surfaces other than ice, namely two silicate minerals. While not directly relevant to the behavior in water, they conclude that their results argue ‘against a crucial role of hydroxyl matching in the antifreeze action’, and cite the companion story in type I AFPs .
To date, apart from the work of Lavalle, DeVries and colleagues , the ice/vacuum mechanism for AFGPs involving hydrogen bonding has not been revisited. In light of the recent new insights into the mechanism of action of type I AFPs, including the important role of hydrophobic interactions, new mechanisms for the molecular action of AFGPs need to be considered. The chemistry of modification of the hydroxyl groups (including stereochemistry), as well as the hydrophobic amino-acid sidechains, which illuminated the interactions of type I AFPs with the ice/water interface, is obviously more difficult for the AFGPs, as described in the Synthesis section above. It will be of interest to see whether hydrophobicity is a dominant interaction in the mechanism of action of both AFGPs and AFPs. While AFGPs are unstructured in solution, it has been noted that in a three-fold left-handed helical conformation, the glycoprotein contains a hydrophilic side face and a hydrophobic face in which most of the Ala side-chains are located [3,34,87]. Whether this conformation is a significant contributor to the alignment of AFGP molecules with specific surfaces at the ice/water interface is unknown and will need to be tested through structure-activity relationship studies.
Both AFGPs and AFPs exhibit a number of unique properties which protect biological systems in vitro and have been investigated for potential applications in medicine, biotechnology and the food industry. A comprehensive review summarizing the effects of AFPs and AFGPs on low temperature preservation processes has recently been published . The ability to change the normal growth habit of ice, the capacity to inhibit recrystallization and the protection of cell membranes are all properties of AFPs and AFGPs that may be tailored for a range of low temperature processes.
The ability of AFGPs to aid in the cryopreservation and hypothermal storage of cells and tissues was noted by Rubinsky et al. . The effect of the addition of a mixture of AFGP1–8 (one part AFGP1–5 to three parts AFGP7–8), or separate solutions of AFGP1–8 and AFGP7–8, on the storage of pig oocytes which cannot survive hypothermic temperatures as high as 10 °C, was evaluated . Protection of the oocytes was monitored by measurements of the membrane potential across the oolemma, and it was proposed that AFGP1–8 protect the cell membranes and inhibit ion leakage. Later studies proposed a more detailed mechanism of cellular protection by both AFPs and AFGPs involving blocking of the potassium and calcium ion channels during cooling [103,104]. In contrast, AFGPs failed to enhance storage of isolated rat hearts at hypothermic temperatures and caused increased damage under freezing conditions regardless of AFGP concentration , and samples of ram spermatozoa were not stabilized in the presence of AFGPs when chilled and rewarmed .
In an effort to understand the apparent different properties of AFGPs discussed above, Crowe and coworkers performed a series of studies on liposomes as a model for studying the effects of lipid-phase transitions. The effects of AFGPs on the leakage of a trapped marker from liposomes during chilling were monitored . While cooling of these liposomes through the transition temperature resulted in leakage of approximately 50% of their contents, addition of less than 1 mg·mL−1 of AFGP prevented up to 100% of this leakage, both during chilling and warming through the phase transition. Thus it was concluded that the stabilizing effects of AFGPs on intact cells during chilling reported in earlier studies [103,104] was possibly be due to a nonspecific effect on the lipid components of native membranes [107,108]. The importance of performing studies with purified AFGPs was also highlighted, with contaminants from other blood proteins present shown to also associate with liposomes, leading to defects in the bilayer and thus leakage . An independent study on the effect of AFGPs from the rod cod Gadus ogac showed that all AFGPs with molecular mass 2.6–24 kDa prevented leakage from model liposomes as they were cooled through their phase transition temperature, with the larger molecular mass compounds being about four times as effective as the smaller ones .
In support of the hypothesis that AFGPs protect cellular membranes during lipid-phase transitions, improved storage of chilled blood platelets was demonstrated [30,108]. In contrast to liposomes, only the AFGPs provided a protective mechanism with nonglycosylated AFPs and ovotransferrin having no beneficial effects. The internal calcium concentration of human platelets was shown to increase during chilling  but AFGPs did not eliminate this rise in concentration.
More recently the effects of AFGPs on different membrane compositions has been studied . The effects of freezing spinach thyalkoloid membranes and model membranes of varying lipid compositions in the presence of AFGPs showed that the lower molecular mass AFGP8 offers a limited degree of protection during freezing and does not induce membrane fusion at concentrations up to 10 mg·mL−1. This behavior is quite distinct from that exhibited by AFPs , or the larger molecular mass AFGP1–5 or AFGP3–4, which are cryotoxic to thyalkaloids and liposomes.
AFGPs and AFPs have been identified as useful in cryosurgery, increasing the destruction of solid tumors through mechanical damage to cells caused by the growth of bipyramidal ice crystals . However specific applications are limited to AFPs [113,114] and the effectiveness of AFGPs in this field has yet to be demonstrated.
Both AFPs and AFGPs have attracted significant interest as potential food additives that inhibit ice recrystallization and hence the formation of large ice crystals in the storage of frozen foods [29,32,115,116]. Unfortunately the use of the generic term antifreeze proteins or compounds to refer to both AFPs and AFGPs in many papers and reviews makes it difficult to establish exactly which AFGPs have been tested. Studies of the effect of AFGP1–8 (Dissostichus mawsoni) in the quality of frozen meat have shown reduced tissue damage due to freezing , and improved drip loss and sensory properties of thawed meat from lambs that had been administered AFGPs prior to slaughter . However, efficient and cost-effective methods of using these compounds as additives are required for commercial applications.
While significant progress had been made in the structural characterization and properties of AFPs and AFGPs from cold water fish, the molecular level detail of how each class of compounds is able to inhibit ice growth is still not fully understood. The key structural features required for ‘antifreeze’ activity by type I AFPs have been identified through structure-activity studies on analogues accessible using either synthetic or molecular biology techniques. In contrast, the lack of a feasible synthetic route to AFGP analogues has hampered progress with this class of compound, and the understanding of the accumulation of AFGPs at certain ice/water interfaces stands at roughly the same point as type I AFPs were in the early 1990s. A concerted attempt at the routine production of AFGP analogues is warranted, as difficult as this may be, to provide essential data regarding the mechanism of ice growth inhibition. In addition, such studies have the potential to identify simpler AFGP analogues that are less difficult to produce.
A second avenue ripe for exploration is the interaction of AFGPs (and AFPs) with membranes, both synthetic and natural [110–112]. While potential applications in the storage and preservation of low temperature biological samples has been demonstrated, systematic studies are still required to establish how each class of compound interacts with membranes and other biomolecules in order to tailor new AFPs and AFGPs for specific applications.
M. M. H. acknowledges financial support from the University of Sydney Sesqui Research and Development Scheme and the Australian Research Council, and travel funds from the Australian Academy of Science. A. D. J. H. thanks the Welch Foundation for support at the University of Houston where part of this review was written, and NSF for use of the Crary Research Laboratory, McMurdo Sound, where the data for Fig. 4 were collected. A. D. J. H. acknowledges many helpful conversations on this topic over the years with Drs Art DeVries, Chris Cheng, Charlie Knight and Peter Wilson.