Structural Basis for the Superior Activity of the Large Isoform of Snow Flea Antifreeze Protein<xref rid="fn1"></xref><xref rid="fn2"></xref>

The snow flea (Hypogastrum harveyi) is protected from freezing at sub-zero temperatures by a glycine-rich antifreeze protein (AFP) that binds to seed ice crystals and prevents them from growing larger. This AFP is hyperactive and comprises two isoforms [Graham, L. A., and Davies, P. L. (2005) Science 310, 461]. The larger isoform (15.7 kDa) exhibits several-fold higher activity than the smaller isoform (6.5 kDa), although it is considerably less abundant. To establish the molecular basis for this difference in activity, we determined the sequence of the large isoform. The primary sequences of these two isoforms are surprisingly divergent. However, both contain tripeptide repeats and turn motifs that enabled us to build a threedimensional model of the large isoform based upon the six-polyproline helix structure of the small isoform. Our model contains 13 polyproline type II helices connected by proline-containing loops stacked into two flat sheets oriented antiparallel to one another. The structure is strictly amphipathic, with a hydrophilic surface on one side and a hydrophobic, putative ice-binding surface on the other. The putative ice-binding site is approximately twice as large in area as that of the small isoform, providing an explanation for the difference in activity that is consistent with other examples noted. By tagging the recombinant AFP with green fluorescent protein, we observed its binding to multiple planes of ice, especially the basal plane. This finding supports the correlation between AFP hyperactivity and basal plane binding first observed with spruce budworm AFP. Antifreeze proteins (AFPs) are structurally diverse macromolecules found in a variety of organisms that live in freezing habitats (1-5). AFPs make crucial contributions to the existence of life in these environments, where organisms can be killed by uncontrolled ice crystal growth (6, 7). AFPs function at the ice crystal surface by an absorption-inhibition mechanism (8), creatingmicrocurvatures in the ice surface that, due to theKelvin effect, energetically disfavor the further addition of water molecules (9, 10). This results in the depression of a solution’s freezing temperature below its equilibrium melting point, a phenomenon that is termed “thermal hysteresis” (TH) (11). In this way, AFPs inhibit the growth of ice crystals within organisms and enhance their ability to survive at sub-zero temperatures. To date, AFPs have been isolated from many species of fish, insects, plants, and microorganisms. For proteins that serve the same function, AFPs have a remarkable diversity of structures, including a number of folds that are novel (3). One common feature, however, appears to be an ice-binding face that is relatively flat andmore hydrophobic than other surfaces of theAFPs and that in several cases contains a regular array of threonine residues (12-14). Nevertheless, the molecular basis of AFP binding to ice is still not well understood, and a major driving factor for the modeling and elucidation of new AFP structures is the potential for discerning clues about the exact mechanism by which AFPs exert their effect. A highly potent, glycine-richAFPwas discovered in snow fleas (Hypogastrum harveyi) collected from Eastern Ontario, Canada, in late winter (15). Crude extracts from these primitive arthropods were able to depress the freezing temperature of a solution in the presence of ice ∼6 C below its melting temperature. The substantial TH activity exhibited by the extracts places snow flea AFP in the class of hyperactive AFPs, with activity significantly higher than those found in most fish and plants (16). This feature potentially helps snow fleas avoid freezing in the exposed, snowladen ground from where they forage. The AFP responsible for the observed TH activity was subsequently purified from the extracts by two rounds of ice affinity purification (17). Reversedphase high-performance liquid chromatography (HPLC) resolved the protein into two isoforms, each with a single mass (15). The larger (15.7 kDa) of the two isoforms was considerably less abundant than the smaller (6.5 kDa) one. However, the 15.7 kDa isoform exhibited significantly greater antifreeze activity. Both isoforms had an unusual amino acid composition, in that glycine made up∼45%of the residues. The smaller isoformcontained no aromatic or long chain aliphatic amino acids, while the larger isoform had a small number of these residues. The small isoform contained four internally disulfide bonded cysteines, while the large isoform had two. When clones from a snow flea cDNA library were randomly sequenced, four of 57 sequences matched the amino acid composition of the small isoform and had a predicted mass (after P.L.D. holds aCanadaResearchChair in Protein Engineering. F.-H. L. is funded by an Ontario Graduate Scholarship. Y.C. acknowledges support from the Condensed Matter and Surface Science program at OhioUniversity. This researchwas funded by a grant to P.L.D. from the Canadian Institutes for Health Research and by a grant to I.B. from the National Science Foundation (Grant CHE-0848081). The DNA sequence of the large isoform of snow flea antifreeze protein has been submitted toGenBank (accession numberGU169329). *To whom correspondence should be addressed. E-mail: peter.davies@ queensu.ca. Telephone: (613) 533-2983. Fax: (613) 533-2497. Abbreviations: AFP, antifreeze protein; GFP, green fluorescent protein; Cy5, cyanine-5; PPII, polyproline type II; rmsd, root-meansquare deviation. 2594 Biochemistry, Vol. 49, No. 11, 2010 Mok et al. removal of a signal sequence) equal to that experimentally determined for the naturally isolated AFP (15). None of the 57 sequences encoded the large isoform. The small isoform had a prominent tripeptide repeat pattern where glycine was present in the first position. The second position was occupied by either glycine or an amino acid with a small side chain. The third position was the most variable, but even here there was a pattern, where runs of charged and hydrophilic residues alternated with hydrophobic residues separated by discontinuities in the tripeptide repeat pattern that often included a proline residue. These observations led to the development of a model for the threedimensional (3D) structure of the small isoform in which the 81-residue protein was folded into six polyproline type II helices that alternated in direction at the proline-containing turns (18). The core of the protein was stabilized by hydrogen bonds between the backbone amides which were made possible by the regular pairing of glycines on opposing faces of the internal coils. The model predicted a novel structure that had several of the attributes of AFPs. These include amphipathic character and an elongated, relatively flat hydrophobic surface with surface regularity that could potentially dock to ice. A high-resolution crystal structure of the small isoform obtained via complete chemical synthesis of the protein (19) was recently reported and is in excellent agreement with the model (20). The large isoform of snow flea AFP is of interest because of its superior antifreeze activity compared to that of the small isoform. To determine the molecular basis of this difference in activity, we set out to sequence the large isoform. Given the similarity in amino acid composition between the twoAFPs, we hypothesized that the 15.7 kDa protein was an isoform of the smaller AFP and would also fold into polyproline type II antiparallel sheets to give a larger version of the small isoform structure. In this paper, we present the primary structure of the large isoform as well as a 3D model based on the small isoform structure of snow flea AFP. The primary sequences of the two isoforms are surprisingly divergent, but the 3Dmodels are highly similar. The hydrophobic face of the structure is at least twice as large as the equivalent face on the small isoform. Chemical synthesis of the gene for the large isoform then allowed us to express the protein recombinantly in sufficient amounts to confirm its superior activity. By tagging this protein with green fluorescent protein (GFP), we found that it binds tomultiple planes of ice, including the basal plane, which is consistent with its hyperactivity. MATERIALS AND METHODS N-Terminal and Internal Sequencing of the 15.7 kDa Protein.N-Terminal sequence analysis was performed byEdman degradation on an Applied Biosystems Procise Sequencer at the Peptide Sequencing Facility of theAdvanced Protein Technology Centre (Hospital for Sick Children, Toronto, ON) using protein from the 15.7 kDaHPLCpeak.A trypticmass fingerprint, aswell as a second, internal protein sequence determined by MS/MS sequencing, were obtained from the Protein Function and Discovery facility (Queen’s University). Cloning of the DNA Sequence Using the Polymerase Chain Reaction. Snow flea mRNA (100 ng per reaction) was reverse transcribed with either oligo(dT)17-CSX or random hexamer primers using the Thermoscript RT-PCR System (Invitrogen, Carlsbad, CA) by following the manufacturer’s instructions. PCR was performed using 2 μL of cDNA (/10 of that in the reaction mentioned above) in a 50 μL volume using Taq DNA polymerase with Q-solution (for GC-rich templates) (Qiagen, Germantown, MD), together with partially degenerate primers (final concentration of 0.4 μM) corresponding to the peptide sequences DGRSNGE (50-GAYGGNAGRAGYAAYGGCGAA-30) and RGGDGAN (50-TTNGCNCCRTCNCCTCCACG-30). The cDNA was denatured at 95 C for 5 min, followed by 30 cycles at 95 C for 1 min, 60 C for 1 min, and 72 C for 2 min. An aliquot (1 μL) was reamplified as described above for 25 cycles. All of the PCR bands described were TA cloned (TOPO TA-cloning kit, Invitrogen) and sequenced (Cortec, Kingston, ON). Note that the degeneracy of the primers described above was reduced on the basis of the sequence of short PCR fragments obtained using degenerate primers based on the N-terminal sequence APNGADG (50-GCWCCHAAYGGWGCWGAYGG-30) and the internal sequence GGDGANG (50-NCCRTTNGCNCCRTCNCCNCC-30). The 30 end of the cDNA was cloned from an aliquot of the cDNA library (15) by anchor PCR. The gene-specific primer (50-ACAACCCGGTGGTAATGGTGGAAAC-30) and the T7 extendedprimer (50-ACGACTCACTATAGGGCGAATTGG-30) were used in the first reaction, and the second reactionwas nested at the 50 end using a second gene-specific primer (50-CGGGACTAGGTGGTGATAGTGTAA-30). Cloning of the 50 Portion of the Gene by Inverse PCR. Aliquots (2 μg) of snow flea genomic DNA were digested in a 250 μLvolumewith 40 units of eitherHindIII,AseI, orHaeIII for 4 h at 37 C and then overnight following the addition of an identical aliquot of enzyme. Following denaturation of the restriction enzymes at 68 C, 200 ng of digestedDNAwas ligated in a 100 μL volume at 16 C overnight. PCR was performed as previously described but with an annealing temperature of 65 C. The reaction was nested at the 30 end using the same primers used in anchor PCR and at the 50 end using the following two primers, 50-CAGACCCTGTTCCAGGGAACCCAT-30 and 50-ATCACAACCATTTGCCCCAGCAGTA-30. A second portion of genomic sequence was obtained using genomic DNA in PCRs as described above, using one of the second gene-specific primers described above and a primer overlapping the stop codon (50-TTACGCTCCACCGCCGCCACC-30). The sequences of the two isoforms were aligned using DNAMAN version 6 (Lynnon Corp., Pointe-Claire, QC) followed by minor manual adjustments. ModelingBuilding andMolecularDynamics Simulation. Modeling of the large isoform employed the same concepts and strategy described for the small isoform (18) and was an intuitive, iterative process. Briefly, it began with the recognition of a tripeptide-repeating pattern throughout the protein suggestive of a 3-fold helical repeat. It continued with the identification of regularly spaced discontinuities in the repeat pattern that might correspond to bends or turns in the chains, thus dividing the sequence into 13 segments. The presence of a disulfide bond and irregular distribution of Gly-Gly pairs were also taken into account. A physical model was constructed using the HGS Biochemistry Molecular Model from Himomoto Plastics (Tokyo, Japan) before the model was rendered with PyMOL (21). To perform molecular dynamics simulations, the model was computationally solvated in a 8.4 nm 6.3 nm 5.2 nm box of water containing 8409 water molecules and had a single net positive charge. To neutralize the charge of the system and to provide an effective concentration of 0.1 M NaCl, 33 water molecules were replaced with 16 Naþ and 17 Cl ions. The system was subject to energy minimization by steepest descents, Article Biochemistry, Vol. 49, No. 11, 201