Molecular Simulation of the Antifreeze Mechanism of Antifreeze Proteins

Weijia Zhang; Xueguang Shao; Wensheng Cai

doi:10.7536/PC200912

Progress in Chemistry ›› 2021, Vol. 33 ›› Issue (10) : 1797-1811 DOI: 10.7536/PC200912

research-article

Scan to read & Share

Molecular Simulation of the Antifreeze Mechanism of Antifreeze Proteins

Author information +

History +

PDF

Abstract

Antifreeze proteins enable organisms to survive in subzero environments. Owing to this unique property, antifreeze proteins have great potential application in a variety of fields. Antifreeze proteins have been extensively studied, but the antifreeze mechanism is still fragmentary, due to the limitation of experimental means. Molecular dynamics, as a method to simulate the interaction between molecules at the atomic level, has been widely used in the study of the mechanism of antifreeze proteins in recent years. In this paper, the functional and structural characteristics of antifreeze proteins, and the research progress of antifreeze mechanisms are described and reviewed from the perspective of structure. Furthermore, the structural characteristics of 29 wild-type antifreeze proteins with known crystal structure, and the hydrophobicity of the residues distributed on the protein surface and at the ice-binding site are analyzed. Both the affinity of the interaction between hydrophilic residues and water and the specificity of the interaction between hydrophobic residues and ice-like water are found either on the surface or at the ice-binding site of antifreeze proteins. The relationship between the secondary structure, the hydrophobicity of the residues at ice-binding site and the antifreeze activity is discussed. Finally, the mechanisms of antifreeze proteins and the factors affecting the antifreeze activity are discussed from the perspective of structure, and the progress in the design and application of bioinspired antifreeze materials with low toxicity and cost is briefly summarized.

Contents

1 Introduction

2 Functional characteristics of antifreeze proteins

3 Structure and mechanisms of antifreeze proteins

3.1 Fish AFPs

3.2 Insect AFPs

3.3 Plant AFPs

3.4 Microorganism AFPs

3.5 Similarities and differences of antifreeze mechanisms of different antifreeze proteins

3.6 Structural factors affecting antifreeze activity

4 Bioinspired antifreeze materials

5 Conclusion and outlook

Graphical abstract

Keywords

antifreeze proteins / bioinspired antifreeze materials / thermal hysteresis activity / adsorption-inhibition mechanism / ACW mechanism / molecular dynamics simulations

Cite this article

Download citation ▾

Weijia Zhang, Xueguang Shao, Wensheng Cai. Molecular Simulation of the Antifreeze Mechanism of Antifreeze Proteins. Progress in Chemistry, 2021, 33(10): 1797-1811 DOI:10.7536/PC200912

登录浏览全文

4963

注册一个新账户忘记密码

1 Introduction

antifreeze proteins (AFPs) are a class of proteins evolved by organisms to resist low temperature and freezing injury, which have attracted much attention because they are one of the most efficient biological antifreeze agents^[1⇓⇓~4]. There are two ways for organisms to resist freezing injury: avoiding freezing and tolerating freezing^[5]. Freezing resistant organisms use the thermal hysteresis activity of their AFPs, that is, the difference between the freezing point and the melting point caused by the binding of AFPs to ice, to non-colligatively lower the freezing point and prevent body fluids from freezing at low temperatures^[6,7]. However, the thermal hysteresis activity of AFPs is very low for the freezing tolerant organisms, which mainly helps the organisms to tolerate freezing by inhibiting the recrystallization of ice crystals^[1,3]. Antifreeze protein has great potential application value because of its unique antifreeze properties, but it is difficult to be produced and applied on a large scale because of its difficult extraction, high synthesis cost and certain cytotoxicity. Therefore, it has been a hot topic in the field of chemistry to reveal its antifreeze mechanism and design and synthesize low-cost and low-toxicity antifreeze protein biomimetic materials based on it^[8⇓⇓~11].

In 1969, animal physiologists De Vries and Wohlschlag first discovered and purified AFP in the plasma of Antarctic fish. So far, many kinds of AFPs have been found in fish, insects, plants and microorganisms^[1]^[1,2,12,13]^[6,7]^[3]^[14,15]. Because AFPs have evolved from different species and have no genetic homology, they have significant differences in sequence, structure and ice-binding site (IBS). So do the ice binding sites of these AFPs have common characteristics? How to distinguish water and ice with different structures of the same substance, identify and bind to ice crystals from water? These problems have attracted wide attention of researchers. In recent years, the amino acid sequence and crystal structure of antifreeze proteins have been reported, but the antifreeze mechanism is still controversial^[14,15]^[16⇓~18]. At present, the generally accepted mechanism of action of antifreeze proteins is the adsorption-inhibition mechanism, that is, antifreeze proteins irreversibly adsorb on the specific surface of ice crystals and inhibit the growth of ice crystals^[19]. However, the existing experimental methods are difficult to directly detect the details of the interaction of antifreeze proteins in ice-water system. molecular dynamics (MD) simulation is a computer simulation of the interaction between molecules at the atomic level. At present, MD simulation has been widely used to study the mechanism of various physical and chemical processes in biological and abiotic systems^[20,21]^[22,23]. The simulation trajectory based on MD can not only obtain the stable state and transition state structure of the system in the process of change, but also obtain the thermodynamic statistical results such as free energy change of physical and chemical processes, which can further reveal the mechanism of complex chemical and biological processes^[24⇓~26]. In recent years, MD has also been used to simulate the behavior of antifreeze proteins in ice-water mixtures. For example, by calculating and comparing the diffusion coefficients of water at the ice-water interface in the presence and absence of antifreeze proteins, it is found that antifreeze proteins can broaden the thickness of the ice-water interface^[27]. By calculating the binding free energy between antifreeze proteins and their mutants and the ice-water interface, the key groups in the binding process were found^[28,29]. By studying the density and hydrodynamics of solvated water around antifreeze proteins, the structure and properties of solvated water were obtained^[30,31]. Based on the results of experiments and simulations, researchers further proposed the anchored clathrate water (ACW) mechanism on the basis of the adsorption-inhibition mechanism for antifreeze proteins without sugar groups, which holds that the hydration layer of antifreeze proteins recognizes and binds to the specific surface of ice crystals, and finally freezes on ice^[32]. However, some researchers have observed that some antifreeze proteins can cause local melting of ice through MD simulation, and proposed an antifreeze mechanism based on local melting^[33]. There are also different views on the mechanism of antifreeze glycoproteins (AFGPs) in fish. Some reports claim that the adsorption process is reversible, and some reports suggest that AFGPs do not need to adsorb to ice, but through long-distance disturbance of water solvent, which makes water not conducive to freezing^[29]^[31,34,35]. To sum up, although the mechanism of action of antifreeze proteins has been widely studied, it is still controversial. In this paper, the functional properties and structural characteristics of antifreeze proteins are described, and the theoretical and experimental research progress on the antifreeze mechanism of antifreeze proteins is reviewed from the structural point of view. On the other hand, based on the hydrophobicity of the surface residues and the ice-binding sites of 29 wild-type antifreeze proteins in the PDB database, it was found that there were hydrogen bonds between hydrophilic residues and water and specific binding between hydrophobic residues and ice-like structures on the surface of antifreeze proteins and at the ice-binding sites. Then, the relationship between the secondary structure of antifreeze proteins, the proportion of hydrophobic residues in the ice binding sites and antifreeze activity was discussed, and the factors affecting the antifreeze activity of proteins were further speculated.

2 Functional Properties of Antifreeze Protein

(1) Thermal hysteresis activity

Antifreeze proteins can lower the freezing point but hardly affect the melting point, resulting in a difference between the freezing point and the melting point, which is called thermal hysteresis. The difference between the melting point and the freezing point is called thermal hysteresis. The greater the difference, the greater the thermal hysteresis activity^[36]. The thermal hysteresis activity of the protein can be measured using a nanoliter osmometer, a differential scanning calorimeter, or the like^[37]^[38]. However, it is difficult to accurately compare the thermal hysteresis activity of different antifreeze proteins because the results depend on the method and time. Generally speaking, insect AFPs and microbial AFPs are considered as hyperactive AFPs (hypAFPs), and their thermal hysteresis activity can reach 5 ℃ or more, while fish AFPs and plant AFPs are considered as moderate AFPs, and their thermal hysteresis activity is generally less than 1 ℃^[39,40]^[41]. The thermal hysteresis activity of moderately active antifreeze proteins mainly depends on their binding rate and concentration in solution, and some of them can even reach superactivity with the increase of concentration.For superactive antifreeze proteins, although the thermal hysteresis activity also increased with the increase of AFPs concentration in the solution within a certain range, it was more dependent on the concentration of AFPs on the surface of ice crystals^[42,43]^[44]^[8,45]^[42,46,47].

(2) Growth Morphology of Modified Ice Crystal

All natural ice in nature is hexagonal ice (Ih), consisting of multiple layers of water molecules arranged in a hexagonal grid, in which each single layer of ice crystal is called the basal plane, the normal of the basal plane is called the "C axis" of the ice crystal, and the direction of the a axis is shown in Figure 1. In the absence of antifreeze proteins, ice crystals grow rapidly along the a-axis, but slowly along the c-axis, and eventually form dendritic crystals. It is generally believed that antifreeze proteins with moderate activity can be adsorbed on the prismatic and/or pyramidal surfaces of ice crystals, inhibit the growth of ice crystals in the a-axis direction, and make ice crystals form a double pyramid shape, as shown in Figure 1 (B)^[48].

The superactive antifreeze protein can be adsorbed not only on the prismatic surface and/or the conical surface, but also on the basal surface, and simultaneously inhibits the growth of the ice crystal in the a-axis and the c-axis, so that the ice crystal is modified into a hexagonal disc shape, as shown in fig. 1 (C)^[48]. However, it has been reported that the characteristic ice crystal morphology observed in most superactive antifreeze protein solutions is formed during the melting process^[49]. In the thermal hysteresis range, the superactive antifreeze protein can be adsorbed on the basal plane and at least one crystal plane orthogonal to the basal plane, so that the shape of ice does not change; and in the melting process, due to the strong affinity of the superactive antifreeze protein for the basal plane, the melting rate of ice crystals along the c-axis is reduced, and "lemon-like" ice is formed. This affinity of superactive antifreeze proteins for the basal surface is believed to be a determinant of thermal hysteresis activity^[39,49]. However, some researchers have found that LpAFP from plants and FcIBP from microorganisms can be adsorbed on the basal surface of ice crystals, but they belong to moderate activity antifreeze proteins^[50]^[51]. This may be due to the fact that their ice binding sites are composed of threonine, serine and other residues, which are not complementary to the geometric shape of the ice crystal surface, resulting in a loose ACW, weakening the binding capacity and reducing the thermal hysteresis activity. This indicates that the binding strength of antifreeze proteins to ice crystals is also the determinant of their thermal hysteresis activity.

(3) ice recrystal lization inhibition (IRI)

Recrystallization of ice crystals refers to the regeneration and redistribution among the formed ice crystal particles, that is, the larger the ice crystals grow, the smaller the ice crystals become until they disappear. This phenomenon can cause fatal damage to organisms^[52]. Antifreeze proteins can inhibit the recrystallization of ice crystals, make the crystals formed by ice crystals small and uniform, and prevent the mechanical damage to cell membranes caused by the increase of ice crystals^[53]. All the known antifreeze proteins have the activity of inhibiting ice recrystallization, among which AFGPs from fish are the most effective inhibitors, which may be due to their disaccharide motifs^[39]^[54].

(4) Other characteristics

AFPs secreted by marine psychrophilic microorganisms can structure the external ice (structuring of external ice), as shown in Figure 2 (a), keeping the water channels around their cells open, ensuring the supply of nutrients needed by the cells and the discharge of waste products^[18,55]. In addition, in recent years, it has been found that AFPs (MpAFP) secreted by bacteria have the function of ice adhesin, that is, one end of AFPs is connected to the surface of bacteria, and the other end is adhered to the ice through the ice binding region, as shown in fig. 2 (B), so that bacteria can adhere to the lake surface or under the sea ice to ensure the supply of oxygen and nutrients^[56].

3 Structure and Action Mechanism of Antifreeze Protein

In 1977, Raymond and DeVries proposed an adsorption-inhibition mechanism, as shown in Fig. 3, suggesting that antifreeze proteins are irreversibly adsorbed on certain crystal faces of ice crystals, and water molecules can only continue to crystallize between the adsorbed antifreeze proteins^[19]. Thus, in the presence of antifreeze proteins, the newly formed ice surface is curved, according to the Gibbs-Thomson effect (Kelvin effect) — an increase in surface curvature leads to a decrease in the freezing point, thus activating antifreeze activity. This statement is currently recognized as the mechanism of action of antifreeze proteins, but the specific force of binding between antifreeze proteins and ice crystals is not clear. In order to activate the antifreeze activity of antifreeze proteins, the binding force between them and ice crystals is very stable, otherwise the growing ice crystals can push these proteins away from the front of the ice, thus preventing the proteins from binding to the ice crystals.

The interaction between antifreeze proteins (AFPs) and ice crystals has been extensively studied since then. At present, the interaction models of AFPs can be divided into two categories, the "hydrogen bonding" model and the "receptor-ligand" model^{[48,57⇓⇓~60]}. The "hydrogen bonding" model is subdivided into the "lattice matching" model, the "dipole-dipole" model, and the "lattice occupation" model^[57]^[58]^[59]. These "hydrogen bonding" models are mainly proposed for fish AFPs. For AFPⅠ in fish AFPs (see Section 3.1 for classification), DeVries proposed a "lattice matching" model in 1983, which suggests that the hydroxyl group on threonine forms hydrogen bonds with water molecules on specific crystal planes due to the α-helix structure of AFPⅠ, as shown in Figure 4 (a)^[57]. The "dipole-dipole" model assumes that the dipole of AFP I can interact with the dipole of water molecules surrounding the ice crystal^[58]. The "lattice occupation" model suggests that hydroxyl groups in AFGPs occupy the positions of oxygen atoms on the surface of ice crystals, thus forming multiple hydrogen bonds with adjacent oxygen atoms^[59]. However, these "hydrogen bonding" models have been widely disputed, and why antifreeze proteins can preferentially bind to the surface of ice crystals through hydrogen bonding mode rather than forming hydrogen bonds with water^[32]. The "hydrogen bonding" model was also negated by the AFPⅠ mutation experiment^[60]. With the discovery of more and more antifreeze proteins and the analysis of their amino acid composition and structure, it has been found that antifreeze proteins usually show hydrophobicity.This indicates that the hydrophobic groups may be involved in the binding of antifreeze proteins to ice, and the increase of hydrophobic groups enhances the antifreeze activity of antifreeze proteins on the basis of maintaining the structural match between proteins and ice crystals^[61⇓~63]^[18]^[64]. In 2002, Jia and Davies proposed a "receptor-ligand" model that can be applied to the action of antifreeze proteins that have been found so far, in which antifreeze proteins are ligands and ice is the receptor^[48]. The "receptor-ligand" model does not negate the contribution of hydrogen bonding, but suggests that more intermolecular forces are involved in the process, such as hydrophobic interactions and van der Waals forces.

These studies indicate that both hydrophobic and hydrophilic interactions are important in the binding of antifreeze proteins to ice, suggesting that solvated water is involved in the process. Therefore, although researchers have a general consensus on the mechanism of action of antifreeze proteins, it is difficult to regard it as a simple "receptor-ligand" model. It has been reported that hydrophobic dehydration may be the driving force for the binding of antifreeze proteins to ice, and the release of solvated water from the ice binding site of antifreeze proteins into the solvent increases entropy, as shown in Figure 4 (B)^[65,66]. However, studies have shown that the solvated water at the ice-binding sites of many antifreeze proteins forms an ice-like layer, rather than releasing water molecules when bound to ice^{[35,67⇓⇓~70]}. Yang and Sharp proposed that the water in the active surface solvation layer of antifreeze proteins is structurally similar to ice, and proposed an affinity-specificity paradox based on the interaction of hydrogen bonds and hydrophobic groups.That is, hydrophilic groups can form hydrogen bonds with water and thus have a good affinity for water, but are relatively less inclined to bind to ice-like structures, thus resulting in low specificity with ice crystals^[67,68]^[71]. In contrast, hydrophobic groups tend to bind to ice-like structures with good specificity but low affinity, so it is possible that both forces are involved in AFPs-ice binding. This hypothesis was confirmed by Garnham et al., who clearly obtained three-dimensional crystal pictures of the interaction between antifreeze proteins and ice crystals, showing the hydration layer structure of bacterial antifreeze proteins (MpAFP, PDB ID:Therefore, a possible binding mechanism is proposed: antifreeze proteins carry their own "ice" (ACW) to bind with ice crystals, that is, antifreeze proteins make water molecules arrange in a lattice way through hydrophobic interaction.The water molecular lattice is anchored by hydrogen bonds, and the anchored water molecular lattice is matched with the specific ice crystal surface to combine the antifreeze protein with the ice crystal, and finally the hydration layer is frozen on the ice crystal surface^[32]. This mechanism is called the ACW mechanism and is shown in Figure 4 (C). However, some MD simulations and NMR studies have shown that there is no evidence of ice-like structure in the liquid water around the antifreeze protein^[72,73]^[74]. The role of antifreeze proteins on non-ice-binding surfaces has also been extensively studied^[69,75,76]. Liu et al. Revealed the "Janus" effect of antifreeze protein on ice nucleation by MD simulation, that is, the ice binding surface promotes ice nucleation, while the non-ice binding surface inhibits ice nucleation^[75]. Because the non-ice binding surface contains more hydrophilic residues, it can be stably combined with water molecules in the first hydration layer, so that the fluidity of water molecules is reduced, which is not conducive to the formation of hydrogen bonds with subsequent water molecules. In this case, water is not conducive to icing, which prevents the whole antifreeze protein from being frozen in ice^[69]. Therefore, the affinity-specificity co-existence characteristic exists not only in the ice binding sites of antifreeze proteins, but also in the whole surface of antifreeze proteins.

In the above, we have reviewed the research progress of the common mechanism of action in antifreeze proteins. Next, we will analyze the surface properties of 29 wild-type antifreeze proteins in the current PDB database, and compare them with the non-antifreeze protein ubiquitin to compare the proportion of hydrophobic residues in their surface residues. Firstly, the amino acid residues with a relative solvent accessibility of more than 20% were regarded as surface residues, and the proportion of hydrophobic groups in the surface residues was counted.The results showed that the proportion of hydrophobic residues on the surface of 26 antifreeze proteins was greater than that of ubiquitin, and the proportion of hydrophilic residues was greater than 30%, which supported the coexistence of affinity and specificity on the surface of these antifreeze proteins^[77]. Then, we counted the proportion of hydrophobic residues in the ice binding sites of antifreeze proteins, and found that most of them had a high proportion of hydrophobic residues, ranging from 30% to 60%, and some of them had a low proportion of hydrophilic residues in the ice binding sites, but they were rich in threonine, which could provide both hydroxyl and hydrophobic methyl. Therefore, the coexistence of affinity and specificity also exists in the ice binding sites of antifreeze proteins, which supports the ACW mechanism. Finally, we summarized the research progress on the specific mechanism of action of each class of antifreeze proteins, and discussed the relationship between the secondary structure of proteins, the proportion of hydrophobic residues in the ice-binding site and antifreeze activity.

3.1 Fish AFPs

According to the sequence and structural characteristics of AFPs in fish, AFPs can be divided into five categories: AFPⅠ, AFPⅡ, AFPⅢ, AFPⅣ and AFGPs. AFP I is an alanine-rich α-helix monomer structure, and its ice-binding sites are threonine and alanine in the TxxxAxxxAxx repeat sequence (T is threonine and A is alanine)^[78]. AFP I was originally thought to bind to ice through the hydroxyl group of threonine, but mutation experiments rejected this "hydrogen bonding" model^[60]. When the threonine in the ice binding site was mutated to serine, the antifreeze activity was almost completely lost, although the hydroxyl group was retained, but there was no hydrophobic methyl side chain.When the threonine in the ice binding site was mutated to valine, the hydrogen bonding group was lost, but the antifreeze activity was only partially lost. Then, Chakraborty and Jana used MD simulation method to explain the above experimental phenomena from the perspective of conformation and solvated water, and confirmed that AFPⅠ was a hydration-mediated adsorption mechanism^[79]. They used umbrella sampling method to calculate the potential of mean force (PMF) curve of the binding process between the helix structure of wild type AFP I and its valine mutant and the ice surface.The structure of the binding stable state corresponding to the lowest free energy was further simulated, and the results showed that both wild type AFPⅠ and valine mutant were not bound directly by hydrogen bonds, but adsorbed on the ice surface through an ordered water cage around threonine methyl^[26]. They further used metadynamics and equilibrium simulation to explore the stable state of the helix structure of wild type AFP I and its serine and valine mutants.The water structure around the stable state was analyzed, and the results showed that the loss of activity of the serine mutant was due to the destruction of α-helix structure and the loss of cage-like water structure around the ice-binding surface^[26]. Although the valine mutation affects the α-helix conformation, which is dynamically balanced between the long straight and curved helical structures, the long straight helical structure can still form an orderly cage-like water structure around the methyl group of the valine residue, thus adsorbing on the ice surface similar to the wild type AFP I. This mechanism involves only the action of hydrophobic groups and is therefore different from the ACW mechanism. However, Calvaresi et al. Observed that AFPⅠ can cause local melting of ice through MD simulation, and proposed a new antifreeze mechanism: antifreeze proteins recognize and bind to the ice surface, causing local melting of ice to form supercooled water.The contraction of supercooled water and the influx of adjacent, denser liquid water cause the antifreeze proteins to stay in place, and ice continues to grow between the curved surfaces, causing the freezing point to drop through the Gibbs-Thomson effect, as shown in Figure 5^[33]. This mechanism suggests that AFPs are adsorbed at the ice-water interface, which can explain some apparently contradictory experimental results, that is, the binding of AFPⅠ to ice is both quasi-permanent and reversible, but the force of AFPⅠ inducing ice melting is not indicated^[80]^[81]. Since then, AFPⅢ and insect AFPs have been reported to induce the transformation of ice into liquid water molecules at low temperatures^[82,83]. The α-helix structure of AFPⅠ is essential for either antifreeze mechanism.

As shown in Figure 6, we compared the structural characteristics of AFP I and non-antifreeze protein ubiquitin, and found that the proportion of hydrophobic residues on the surface of AFP I was significantly higher than that of ubiquitin, and hydrophilic residues accounted for about 30%, which supported the characteristics of affinity-specificity coexistence on the surface of antifreeze protein. According to the TAA ice-binding site, the proportion of hydrophobic residues on the ice-binding surface is also as high as 67%, and threonine residues can provide hydrogen bond donors, but it has been reported that it does not belong to the ACW mechanism.The reason may be that the hydroxyl group on the threonine residue of AFP Ⅰ is poorly matched with the ice crystal lattice and can not be directly bonded to the ice crystal surface through hydrogen bonding, so AFP Ⅰ is adsorbed on the ice crystal surface through cage-like water around the methyl group^[79].

AFPⅡ is a globular protein composed of α-helix, β-sheet and five disulfide bonds, which can be further divided into Ca²⁺- dependent AFPⅡ and Ca²⁺- independent AFPⅡ according to the requirement of 1 mol Ca²⁺ for thermal hysteresis activity^[84]. About 40% of the amino acid sequences of the two AFPⅡs are identical. X-ray studies show that they are similar to each other and are homologous proteins of C-type lectin^[85]^[86]. At present, the ice binding site of AFPII is not completely clear. According to the ice binding plane predicted by mutation experiment, as shown in Figure 7, the binding site of Ca²⁺- dependent AFPII is near the Ca²⁺ binding loop^[84]^[87]. AFP II is considered to belong to the ACW mechanism. Arai et al. Revealed the role of Ca²⁺. Ca²⁺ does not affect the overall structure of AFP II, but changes its side chain conformation and the rigidity of its Ca²⁺ binding ring, thus promoting the formation of cage-like structure of solvated water and ice binding^[84,88]^[84].

The hydrophobicity of three wild-type AFPII surface residues and ice-binding site residues in the PDB database was analyzed, as shown in fig. 8, for one Ca²⁺- independent (PDB ID: 2ZIB) and two Ca²⁺- dependent (PDB ID: 2PY2, 6JK4). The proportion of hydrophobic groups in the surface residues of the three AFPⅡs was more than 30%, which was significantly higher than that of ubiquitin. The ice binding sites of 2ZIB and 6JK4 have fewer hydrophobic residues, but they are rich in threonine, which can provide both hydroxyl and hydrophobic methyl groups^[84]. To sum up, both the surface and ice binding sites of AFPⅡ have the characteristics of affinity-specificity coexistence.

AFPⅢ is also a spherical protein with an ice-binding plane consisting of two relatively flat and hydrophobic adjacent segments at an angle of about 150 °^{[89⇓⇓⇓~93]}. However, the polar groups around the non-polar residues of AFPⅢ show the hydration of the non-polar residues and tend to form ice-like structures^[67]. Solid-state NMR studies show that when AFP ⅲ binds to ice crystals, its hydration layer is also frozen, thus allowing direct contact between the protein and the ice^[92]. The MD simulation study of Grabowska et al shows that the antifreeze activity depends on two factors: the transformation of solvated water to ice-like structure and the matching with the geometric shape of ice lattice^[93]. Chakraborty and Jana performed MD simulations of AFPⅢ (PDB ID: 1MSI) in low temperature aqueous solution and found that the peak position of the radial distribution function of water molecules around it appeared at 3.77 Å, which was larger than the distance between the hydrogen bond donor-acceptor (33 Å), indicating that water molecules were not bonded to IBS through hydrogen bonding, and found that water molecules around IBS tended to form a more ordered structure as the temperature decreased^[94]. They further calculated the PMF curves of AFPⅢ binding with the cone, prism and basal surfaces of ice crystals, and simulated the equilibrium of AFP-ice complexes corresponding to the minimum values.The results show that although the number of hydrogen bonds formed between AFP Ⅲ and the substrate is the largest, the binding between AFP Ⅲ and the substrate is the weakest, so the main driving force for the adsorption of AFP Ⅲ on ice crystals is hydrophobic interaction, rather than hydrogen bonds between IBS and ice surface. Recently, Kumari et al. 'S MD simulation study on AFPⅢ and its low activity mutants also showed that AFPⅢ tends to adsorb on the ice facet because the ice facet is the most hydrophobic.The preferential adsorption of AFP on the basal surface may be due to other factors affecting its binding to the ice-water interface, and they found that there was no difference in the ability of AFP III and its mutants to bind to ice crystals, but the difference was its ability to adsorb to the ice-water interface^[95]. All these results indicate that the hydrophobic group plays an important role in the binding process. The results of our analysis of the structural characteristics of AFPIII are shown in Figure 9. Both the surface and the ice binding site of AFPIII are characterized by the coexistence of affinity and specificity, and the proportion of hydrophobic residues at the ice binding site is as high as 50%.

AFP Ⅳ is rich in glutamic acid and contains four antiparallel helical bundles. Up to now, the crystal structure of AFP Ⅳ has not been reported, and the understanding of its antifreeze mechanism is still based on the adsorption-inhibition theory. The homology between AFPⅣ and membrane apolipoprotein is more than 20%, and its homologues also exist in many temperate or subtropical fish that do not need antifreeze proteins. In general, its concentration in the blood of fish is too low to produce antifreeze effect, so researchers generally believe that the antifreeze activity of AFP IV is acquired by chance, rather than the result of selective evolution^[96⇓~98].

AFGPs are very different from other antifreeze proteins in that their thermal hysteresis activity is very low, but they are the most effective inhibitors of ice recrystallization at present^[54]. AFGPs are usually composed of 4 to 55 repeating tripeptide glycan units Ala-Ala-Thr, and are divided into two subtypes, AFGP1-6 and AFGP7-8, according to the degree of repetition. Sometimes the repeating units of AFGPs are mutated, for example, alanine is replaced by proline. It has no clear secondary structure and shows strong flexibility^[99]. It has been reported that antifreeze glycoprotein is adsorbed on the surface of ice crystal through the side chain of sugar group, because the addition of borate which can combine with sugar group will weaken its antifreeze activity^[100,101]. Mochizuki and Molinero suggested that AFGPs were bound to ice crystals by reversible adsorption of hydrophobic groups of peptides and sugar groups^[29]. Through MD simulation, they confirmed that AFGPs exist in the form of PPII helix at low temperature, which leads to the spatial separation of hydrophobic groups and hydrophilic groups, which is the premise for AFGPs to bind to ice.The simulation results showed that the binding of AFGPs to ice was due to the adsorption of the methyl groups of peptides and sugar groups in the holes on the surface of ice crystals, and the decrease of antifreeze activity after adding borate might be due to the destruction of the PPⅡ helix structure of AFGPs after the binding of sugar groups to borate. However, in 2018, Meister et al. Confirmed that both subtypes of AFGPs were irreversibly adsorbed on ice crystals^[17]. In addition to the adsorption-inhibition theory, there are also some reports that AFGPs do not need to adsorb to ice to affect ice growth, but through long-distance disturbance of water solvent, the hysteresis of water movement makes it unfavorable for ice formation^[31,34,35]. Compared with other AFPs, which are generally considered to belong to the ACW mechanism based on the adsorption-inhibition mechanism, the antifreeze mechanism of AFGPs is more controversial, mainly because its molecular flexibility hinders the study of simulation, and makes the experimental results difficult to analyze, and the role of its unique disaccharide group in it is not clear.

3.2 Insect AFPs

Insect AFPs can be divided into two groups, one is snow flea antifreeze protein (SfAFP), which is composed of six antiparallel PPⅡ helices, one side of which is hydrophilic and the other is hydrophobic, and the hydrophobic surface is considered to be the ice-binding site of the protein.Todde et al. Found that SfAFP, like AFPⅠ, can induce local thawing of ice, so SfAFP is considered to belong to the above antifreeze mechanism based on local thawing, and explained that the reason why SfAFP is different from AFPⅠ and belongs to superactive antifreeze protein is that it can irreversibly adsorb on the basal surface of ice crystals^[102,103]^[104].

The other is insect AFPs with TxT repeats on the β-sheet plane as the ice binding plane, and X is mostly hydrophobic residues, thus constituting an amphipathic binding site, as shown in fig. 10^[105]. The results of mutation experiments and MD simulations show that hydrogen bonding and hydrophobic interaction are the necessary intermolecular forces for the binding of this antifreeze protein to ice^{[4,106⇓~108]}. In 2018, Hudait et al. Proposed that hypAFPs of insects are also adsorbed on ice crystals through ACW, but their ACW motif is different from that of MpAFP^[109]. The ice-binding site of MpAFP is composed of a Thr-Gly-Asx motif, which is topologically flat, allowing all three amino acids to form hydrogen bonds with ice. However, only threonine in the Thr-x-Thr binding site of hypAFP can form hydrogen bonds with ice, so there are differences in ACW between hypAFP and ice, indicating that hypAFP can recognize ice through different ACW motifs, and its structure is controlled by the ice crystal plane and the amino acid sequence of the binding site. Then, they used MD simulations to illustrate the role of hydrogen bonds and hydrophobic groups in the formation and stabilization of ACW motifs, and the role of methyl groups is to maintain the conformational rigidity of threonine residues at binding sites.O that the hydroxyl group and the methyl group are arranged to the positions required for binding with the ice, and the size of the methyl group is most suitable for stabilizing the ice-like structure, and too large or too small can lead to the reduction of binding free energy^[28]. However, Zanetti-Polzi et al denied the existence of ice-like structure by studying the hydration layer density of two antifreeze proteins (PDB ID: 1HG7, 1M8 N) in solution.They suggest that the combined effect of the hydrophobicity of the ice-binding surface, which contributes to the binding to ice, and the hydrophilicity of the non-ice-binding surface, which contributes to the prevention of ice growth, and the resulting imbalance in the local water density of the antifreeze protein, impedes ice growth and thus contributes to an increase in antifreeze activity^[16]. Hudait et al. Recently showed that the ice-binding sites of AFPs did not induce the formation of ice-like structures in aqueous solution, but they did when facing the ice-water interface.That is to say, hypAFPs recognize ice through their slow diffusion in solution. When hypAFPs find the right direction relative to the ice surface, they quickly form ACW at the ice binding site and then bind to the ice crystal surface^[73]. Grabowska et al. Studied the stress of antifreeze protein at different positions from the ice surface and the properties of the interfacial water between antifreeze protein and ice, and proposed that the arrangement of water near the ice binding surface of antifreeze protein during adsorption is a "fluent clathrate" structure, which is different from the "anchored clathrate" proposed by Garnham^[76,110]^[32]. The former is plastic and can undergo structural changes during crystallization to match the different crystal planes of the ice. In addition, Grabowska et al. proposed that only when the distance and orientation between the antifreeze protein and the ice are appropriate, the structure of "fluent clathrate" will be generated, otherwise the structure will be deformed or even destroyed, thus generating a force to push or pull the antifreeze protein away from or to the ice surface, so that the distance and orientation between the antifrost protein and the ice surface meet the conditions for forming "fluent clathrate", and then the adsorption is completed^[110]. The results of our analysis of the structural characteristics of this antifreeze protein show that the hydrophobic groups in the surface residues of this antifreeze protein account for 15% -25%, and the hydrophobic residues in the ice binding sites account for 20% -44% (see Table 1).These two ratios are less than those of the spherical fish antifreeze proteins mentioned above, and the proportion of hydrophobic groups in the surface residues is close to that of non-antifreeze proteins, so it is speculated that the flatness of the ice-binding surface is one of the main factors affecting the antifreeze activity.

3.3 Plant AFPs

The thermal hysteresis activity of plant antifreeze proteins is low, but they have strong ability to inhibit the recrystallization of ice crystals^[53,111,112]. Since the crystal structure of only one plant AFP is known so far, the antifreeze mechanism of plant AFPs is rarely reported. Many plant AFPs are homologous to pathogenesis-related (PR) proteins, so they are considered as proteins with dual functions, such as β-1,3-glucanase, chitinase and so on^[113]^[114]. There is evidence that the refolding of PR protein in cold environment may be the reason for its antifreeze activity^[113]. However, there is no clear report about how plant antifreeze proteins exert their antifreeze activity.

The antifreeze glycoprotein extracted from Solanum lyratum Thunb. Has a reduced antifreeze activity when borate is combined with the sugar group^[3]. However, the antifreeze activity of other plant antifreeze glycoproteins did not change significantly even if the glycosyl group was removed, indicating that the role of glycosyl group in different plant antifreeze glycoproteins was different^[115]. LpAFP is the only known crystal structure of plant antifreeze protein (PDB ID: 3ULT), which is composed of eight flat β-sheets, four of which are short β-sheets and four of which are long β-sheets. The two flat β-sheet planes are called "a" and "B" planes^[50,116]. Initially, some researchers proposed that both planes would combine with ice crystals, and then mutation experiments proved that only the "a" plane participated in the combination, as shown in Figure 11^[117]. Compared with insect hypAFPs with the same β-sheet plane, LpAFP is not a regular TxT sequence, in which part of threonine is replaced by serine or valine. Although the proportion of hydrophobic residues in the ice binding site increases to 50%, the complementarity with the ice crystal surface decreases, and the network structure of ACW is looser, which can explain why LpAFP has moderate antifreeze activity^[50]^[50].

3.4 Microbial AFPs

Microbial antifreeze proteins include fungal and bacterial antifreeze proteins. At present, there are 8 microbial antifreeze proteins with known crystal structures in the PDB database, including LeIBP (3UYU) from Arctic yeast, ColAFP (3WP9) from Antarctic sea ice bacteria, FfIBP (4NU2) from Flavobacterium antarcticum, TisAFP6 (3VN3) and TisAFP8 (5B5H) from Nivibacter nivalis, FcIBP (6A8K) from Bacillariophyta, EfcIPP (6E^[14]^[118]^[119]^[120]^[121]^[51]^[15]^[32]. With the exception of MpAFP, which has a β-helix structure, all others have a similar tertiary structure (IBP-1 fold), that is, they contain both β- and α-helices, and their ice-binding sites are located on one side of the β-helix, possibly involving a nearby loop region^[43]^[44].

In 2011, Garnham et al. Demonstrated that MpAFP binds to ice through the ACW mechanism, and the ice binding sites are two rows of threonine and aspartic acid or asparagine, and then researchers successively demonstrated that other microbial antifreeze proteins also belong to the ACW mechanism, and widely discussed why the antifreeze activities of microbial antifreeze proteins with similar structures are very different^[32]^[119⇓~121]. For example, the sequence similarity between LeIBP and FfIBP is as high as 56%, but the antifreeze activity of FfIBP is 10 times higher than that of LeIBP. Do et al believe that this difference is due to the fact that FfIBP exists in the form of monomer in solution, and the ice binding site is composed of T-a/G-x-T/N motif, where G is glycine and N is asparagine, while LeIBP exists in the form of dimer in solution, and the ice binding site is complex^[119]. Hanada et al reported the crystal structures of TisAFP6 and TisAFP8. The two antifreeze proteins have the same high structural similarity, but their antifreeze activities are quite different. The results show that the ice binding site of TisAFP8, especially the loop region, has higher hydrophobicity and better complementarity^[120,121].

It can be seen from Table 1 that the surface and ice binding sites of microbial antifreeze proteins also have the characteristics of affinity-specificity coexistence. The proportion of hydrophobic residues in the ice binding sites of the seven IBP-1-folded microbial antifreeze proteins was less than that of the fish spherical antifreeze proteins, which also indirectly explained the importance of the flatness of the ice binding sites.

3.5 Similarities and differences of antifreeze mechanisms of different AFPs

In conclusion, the antifreeze mechanism of AFGPs in fish is not clear due to their unique disaccharide groups due to their high molecular flexibility, while other AFPs are generally considered to be ACW mechanisms based on the adsorption-inhibition mechanism, and their ice binding sites are also flat, rigid and rich in threonine. Comparing different kinds of AFPs, the similarities and differences of their antifreeze mechanisms and the existing disputes are as follows:

(1) The common mechanism of different AFPs recognizing ice is the ice recognition mechanism mediated by the ordered hydration layer. However, the structural differences of different types of AFPs on IBS lead to the formation of different ordered hydration layers, and the roles of hydrophobic interaction and hydrogen bonding in the binding process are also different. For example, AFPⅠ and AFPⅢ are considered to be the main driving force, and hydrophobic interaction and hydrogen bonding are considered to be equally important in insect hypAFPs^[50,109]^[79,94]^[28]. Hydrophobic interaction makes the water molecules at IBS arrange into ice-like structures, and the main function of hydrogen bonds is to anchor these water molecules at IBS, so the lattice matching degree between AFPs and crystal faces determines whether hydrogen bonds are formed or not. Therefore, for hypAFPs with TxT motifs, both hydrogen bonding and hydrophobic interactions are equally important. However, for AFP Ⅲ, which has no regular hydroxyl group arrangement, recent studies have shown that the main driving force for its adsorption on ice crystals is hydrophobic interaction. There are regular threonine residues in the IBS of AFP Ⅰ, but it has been reported that it binds to ice crystals through hydrophobic groups, possibly due to the poor lattice matching between the hydroxyl groups of threonine residues in AFP Ⅰ and ice crystals^[95]^[79].

(2) It has been reported that AFPⅠ belongs to the "local melting" mechanism. Although it is based on the adsorption-inhibition mechanism, it is completely different from the ACW mechanism in explaining the role of AFPs at the ice-binding sites. The ACW mechanism believes that the hydrophobic groups such as methyl groups at the ice-binding sites can arrange the solvated water around them into ice-like structures.The "local melting" mechanism suggests that the ice binding sites induce ice melting to form supercooled water. Although the force of AFPs inducing ice melting is not specified, the "local melting" mechanism can explain the contradictory phenomenon observed in experiments, that is, the binding between AFPⅠ and ice is both quasi-permanent and reversible^[33]^[80]^[81].

(3) It has been reported that AFPs in aqueous solution, including hypAFPs, can form an ice-like layer in the solvated water at the ice binding site, so it is considered that the binding of AFPs to ice crystals is a process of pre-ordering and then recognition and binding.However, recent studies have shown that AFPs do not need to pre-order water to recognize ice, but diffuse to the corresponding ice crystal surface in solution, and then quickly form ACW at the ice binding site to bind to the ice crystal surface^{[68⇓~70,122]}^[104,123].

3.6 Structural factors affecting antifreeze activity

The proportion of hydrophobic residues of antifreeze proteins is significantly higher than that of non-antifreeze proteins, and the most important structural feature of superactive antifreeze proteins is their regular helical structure. Therefore, our research results on their antifreeze mechanisms from the structural point of view are summarized in Table 1.Based on the hydrophobicity of the surface residues of antifreeze proteins and the residues in the ice binding sites, the mechanism of antifreeze proteins and the factors affecting antifreeze activity were discussed.Conclusions are as follows, in order to provide a theoretical reference for further study of the interaction between different antifreeze protein and ice (or interface water).

(1) The hydrophobicity of the surface residues of 29 wild-type antifreeze proteins and the residues at the ice-binding sites were analyzed and compared with those of ubiquitin. The results showed that the proportion of hydrophobic residues on the surface of 26 antifreeze proteins was higher than that of ubiquitin, with an average proportion of 30% (16% for ubiquitin) and an even higher proportion of 39% at the ice-binding sites. The average proportion of charged (strongly hydrophilic) residues in surface residues was 18%, significantly lower than that of ubiquitin (41%), and the average proportion of charged residues in ice-binding sites was even lower, at 7%. It can be seen that the relatively hydrophobic surface and ice binding sites are important features of the structure of antifreeze proteins^[18]. From the macroscopic point of view, the ice binding sites are relatively hydrophobic, while the non-ice binding sites are relatively hydrophilic, which makes the ice binding surface of antifreeze protein promote ice nucleation, while the non-ice binding surface inhibits ice nucleation, which is in line with the characteristics of affinity-specificity coexistence on the surface of antifreeze protein, and supports the "Janus" effect of antifreeze proteins on ice nucleation^[75].

(2) The average proportion of hydrophobic residues is 39% and the proportion of hydrophilic residues is more than 40% in the ice binding sites, which makes the ice binding sites bind to both water and ice-like structures, which is in line with the characteristics of affinity-specificity coexistence and supports the ACW mechanism^[32].

(3) For the antifreeze proteins with TxT motif, the ice-binding sites are flat β-sheet planes, and the proportion of hydrophobic residues is relatively high, between 20% and 40%, so they belong to superactive antifreeze proteins, while for the spherical antifreeze proteins of fish, although their ice-binding sites are not as flat as those of TxT motif, the proportion of hydrophobic residues is significantly higher. This suggests that antifreeze proteins with a flat ice binding plane, a higher proportion of hydrophobic residues and a regular amino acid arrangement are more conducive to ice binding^[18,94].

(4) The ice binding site of plant antifreeze protein (PDB ID: 3ULT) is also a flat β-sheet plane, and the proportion of hydrophobic residues is as high as 50%.However, due to the poor complementarity between the amino acids at the ice binding sites and the geometric shape of the ice crystal surface, the antifreeze protein has only moderate antifreeze activity, which indicates that the antifreeze protein with good surface complementarity with the ice crystal can play a better antifreeze activity^[50].

4 Antifreeze protein biomimetic material

Antifreeze proteins (AFPs) have been widely used in medical treatment and aquaculture due to their high antifreeze characteristics.However, it is difficult to produce natural AFPs on a large scale due to their high synthesis cost, and the potential immunogenicity of AFPs may induce harmful immune responses^[124,125]^[126,127]^[128]. Therefore, it has been a hot topic in the field of chemistry to reveal the antifreeze mechanism and design and synthesize antifreeze protein biomimetic materials with mass production, good biocompatibility and high stability^{[8⇓⇓~11,129⇓⇓⇓~133]}. In recent years, thanks to the rapid development of high-performance GPU parallel computing architecture and MD simulation methods, especially enhanced sampling technology, large-scale microsecond MD simulation has been used to explore the properties of the hydration layer of antifreeze proteins and the binding free energy of antifreeze proteins with ice crystals, which further systematically reveals the interaction mechanism of various antifreeze proteins with ice crystals at the atomic level^[24,134]^[16,30,73]^[28,29]^[79,95,109].

With the in-depth study of antifreeze proteins, inspired by their chemical composition, three-dimensional structure and functional properties, researchers have designed and synthesized a series of biological or biomimetic materials that can regulate ice nucleation, control ice growth morphology and inhibit ice recrystallization.As anti-icing coatings, natural gas hydrate inhibitors or cryopreservation and transportation of enzymes, cells and organs, these materials usually have the advantages of low cytotoxicity and environmental friendliness^{[11,135⇓⇓⇓⇓⇓⇓~142]}. For example, AFGPs are considered to be the most effective IRIs, but their cost and cytotoxicity are high, and it is hoped that their functions can be simulated by polyvinyl alcohol (PVA), the most effective biomimetic material. Jin et al. Revealed that adequate diffusion of PVA at the ice-water interface contributes to its IRI properties^[138]. Naullage and Molinero used molecular dynamics simulations to clarify the influence of the degree of polymerization and functionalization of PVA on its IRI and the molecular characteristics that give AFGPs excellent IRI performance, which will help to design new flexible polymers with higher IRI activity^[139]. In addition, antifreeze proteins can also be combined with other new materials to improve their antifreeze ability. For example, inspired by antifreeze (sugar) proteins, Lee et al. First attached oligopeptides to the surface of gold colloids to prepare antifreeze gold colloids with precise shape control, as shown in Figure 12 (a)^[135]. The experimental and MD simulation results show that the modified gold colloid can effectively inhibit the growth and recrystallization of ice, and the anti-freezing performance is related to the shape of the gold colloid. This achievement not only expands the design and application of gold colloids, but also enriches the diversity of cryoprotectants.

In addition to simulating the chemical composition of antifreeze proteins to design antifreeze biomimetic materials, the three-dimensional arrangement of some important groups in antifreeze proteins is also very important for their antifreeze properties, and a series of high-performance antifreeze biomimetic materials have been designed inspired by this phenomenon^[14,120]. For example, the formation of hydrogen bonds between natural antifreeze proteins and ice lattices is believed to play an important role in inhibiting the growth of ice. Therefore, the spatial matching of hydrogen bonding sites with ice lattices is very important. Based on this, Bai et al designed and synthesized oxidized quasicarbonitride quantum dots OQCNs, as shown in Figure 12 (B), with a distance of 7.4242 Å between adjacent tertiary nitrogen atoms.Similar to the repeated spacing (7.3535 Å) between oxygen atoms along the c-axis in the ice Ih lattice, they can form hydrogen bonds, so OQCNs bind to the surface of ice crystals, show thermal hysteresis activity, modify the growth morphology of ice crystals and inhibit the recrystallization of ice crystals, and are successfully applied to cryopreservation of cells as cryoprotectants^[28,32]^[105]^[136]. Similarly, based on the idea of precisely locating hydroxyl groups, Drori et al. Found that a supramolecule formed by self-assembly of a synthetic dye can reversibly bind to ice crystals and inhibit the growth of ice crystals. Its structure is shown in Figure 12c. The structure of this aggregate presents alternating amino and methyl groups, which is similar to the arrangement of hydroxyl and methyl groups at the IBS of insect TmAFP^[11]. Recently, Xue et al designed and synthesized a biomimetic ice growth inhibitor based on self-assembled peptides (as shown in Figure 12 (d)), which mimics the regular spatial arrangement of threonine residues in antifreeze proteins and forms a good match with the ice lattice.The results show that it has a moderate inhibitory effect on ice crystal growth and recrystallization, which confirms the importance of threonine residues and their spatial arrangement in the binding of peptides to ice^[137].

For example, Kozuch et al. Used molecular dynamics simulation to study a group of antifreeze proteins with known activity, and analyzed the geometric structure of the protein and the dynamic behavior of surface water.A method for automatically detecting the binding surface of AFPs was proposed, and the activity of 17 unknown antifreeze proteins was accurately predicted by using the artificial neural network based on these data, which laid the foundation for the rational design of antifreeze proteins with superior performance and biomimetic cryoprotective materials^[41]. However, more 3D structures of antifreeze proteins are needed for further study.

5 Conclusion and prospect

In this paper, the functional properties and structural characteristics of antifreeze proteins were described, and the antifreeze mechanism and the factors affecting the antifreeze activity were discussed from the structural point of view. The mechanism of action of antifreeze proteins has been studied for nearly 50 years, but there are still controversies and some unsolved mysteries related to it. With the development of new methods of high-performance MD simulation, especially the combination of enhanced sampling frontier technology to explore the physical and chemical processes related to anti-freezing mechanism, it will become a powerful tool to reveal this magical phenomenon in nature. In addition, based on the revealed antifreeze mechanism, it is a very important but still challenging development direction to design biocompatible and low-cost antifreeze protein biomimetic materials or ice growth inhibitors by means of molecular simulation and machine learning.This reasonable design strategy, which is similar to computer-aided drug design, can provide guidance for the synthesis of such biomimetic materials and reduce the blindness of experiments.

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	DeVries A L, Wohlschlag D E. Science, 1969, 163(3871): 1073.

[2]	Marshall C B, Fletcher G L, Davies P L. Nature, 2004, 429(6988): 153.

[3]	Duman J G. Biochim. Et Biophys. Acta BBA Protein Struct. Mol. Enzymol., 1994, 1206(1): 129.

[4]	Graether S P, Kuiper M J, GagnÉ S M, Walker V K, Jia Z C, Sykes B D, Davies P L. Nature, 2000, 406(6793): 325.

[5]	Duman J G. J. Exp. Biol., 2015, 218(12): 1846.

[6]	Li C M, Guo X R, Jia Z C, Xia B, Jin C W. J. Biomol. NMR, 2005, 32(3): 251.

[7]	Hakim A, Nguyen J B, Basu K, Zhu D F, Thakral D, Davies P L, Isaacs F J, Modis Y, Meng W Y. J. Biol. Chem., 2013, 288(17): 12295.

[8]	He Z Y, Liu K, Wang J J. Acc. Chem. Res., 2018, 51(5): 1082.

[9]	Lv J, Song Y L, Jiang L, Wang J J. ACS Nano, 2014, 8(4): 3152.

[10]	Deville S, Viazzi C, Leloup J, Lasalle A, Guizard C, Maire E, Adrien J, Gremillard L. PLoS One, 2011, 6(10): e26474.

[11]	Drori R, Li C, Hu C H, Raiteri P, Rohl A L, Ward M D, Kahr B. J. Am. Chem. Soc., 2016, 138(40): 13396.

[12]	Sicheri F, Yang D S C. Nature, 1995, 375(6530): 427.

[13]	Kwan A H Y, Fairley K, Anderberg P I, Liew C W, Harding M M, Mackay J P. Biochemistry, 2005, 44(6): 1980.

[14]	Lee J H, Park A K, Do H, Park K S, Moh S H, Chi Y M, Kim H J. J. Biol. Chem., 2012, 287(14): 11460.

[15]	Mangiagalli M, Sarusi G, Kaleda A, Bar Dolev M, Nardone V, Vena V F, Braslavsky I, Lotti M, Nardini M. FEBS J., 2018, 285(9): 1653.

[16]	Zanetti-Polzi L, Biswas A D, Del Galdo S, Barone V, Daidone I. J. Phys. Chem. B, 2019, 123(30): 6474.

[17]	Meister K, DeVries A L, Bakker H J, Drori R. J. Am. Chem. Soc., 2018, 140(30): 9365.

[18]	Davies P L. Trends Biochem. Sci., 2014, 39(11): 548.

[19]	Raymond J A, DeVries A L. PNAS, 1977, 74(6): 2589.

[20]	Cai W S, Chipot C. Acta Chimica Sin., 2013, 71(2): 159.

[21]	( 蔡文生, Chipot Christophe. 化学学报, 2013, 71(2): 159.)

[22]	Zong Z Y, Li Q Y, Hong Z Y, Fu H H, Cai W S, Chipot C, Jiang H F, Zhang D Y, Chen S L, Shao X G. J. Am. Chem. Soc., 2019, 141(36): 14451.

[23]	Liu P, Shao X G, Cai W S. Prog. Chem., 2013, 25(5): 692.

[24]	( 刘鹏, 邵学广, 蔡文生. 化学进展, 2013, 25(5): 692.)

[25]	Wang T, Cai W S, Shao X G. Prog. Chem., 2010, 22(5): 803.

[26]	( 王腾, 蔡文生, 邵学广. 化学进展, 2010, 22(5): 803.)

[27]	Fu H H, Zhang H, Chen H C, Shao X G, Chipot C, Cai W S. J. Phys. Chem. Lett., 2018, 9(16): 4738.

[28]	Fu H H, Shao X G, Cai W S, Chipot C. Acc. Chem. Res., 2019, 52(11): 3254.

[29]	Chen H C, Fu H H, Shao X G, Cai W S. Prog. Chem., 2018, 30(7): 921.

[30]	( 陈淏川, 付浩浩, 邵学广, 蔡文生. 化学进展, 2018, 30(7): 921.)

[31]	Todde G, Hovmöller S, Laaksonen A. J. Phys. Chem. B, 2015, 119(8): 3407.

[32]	Hudait A, Qiu Y Q, Odendahl N, Molinero V. J. Am. Chem. Soc., 2019, 141(19): 7887.

[33]	Mochizuki K, Molinero V. J. Am. Chem. Soc., 2018, 140(14): 4803.

[34]	Kuffel A, Czapiewski D, Zielkiewicz J. J. Chem. Phys., 2014, 141(5): 055103.

[35]	Mallajosyula S S, Vanommeslaeghe K, MacKerell A D. J. Phys. Chem. B, 2014, 118(40): 11696.

[36]	Garnham C P, Campbell R L, Davies P L. PNAS, 2011, 108(18): 7363.

[37]	Calvaresi M, Höfinger S, Zerbetto F. Biomacromolecules, 2012, 13(7): 2046.

[38]	Ebbinghaus S, Meister K, Born B, DeVries A L, Gruebele M, Havenith M. J. Am. Chem. Soc., 2010, 132(35): 12210.

[39]	Meister K, Ebbinghaus S, Xu Y, Duman J G, DeVries A, Gruebele M, Leitner D M, Havenith M. PNAS, 2013, 110(5): 1617.

[40]	DeVries A L. Science, 1971, 172(3988): 1152.

[41]	Braslavsky I, Drori R. J. Visualized Exp., 2013, 72: e4189.

[42]	Amornwittawat N, Wang S, Duman J G, Wen X. Biochim. Et Biophys. Acta BBA Proteins Proteom., 2008, 1784(12): 1942.

[43]	Olijve L L C, Meister K, DeVries A L, Duman J G, Guo S Q, Bakker H J, Voets I K. PNAS, 2016, 113(14): 3740.

[44]	Drori R, Celik Y, Davies P L, Braslavsky I. J. R. Soc. Interface, 2014, 11(98): 20140526.

[45]	Kozuch D J, Stillinger F H, Debenedetti P G. PNAS, 2018, 115(52): 13252.

[46]	Bar Dolev M, Braslavsky I, Davies P L. Annu. Rev. Biochem., 2016, 85(1): 515.

[47]	Drori R, Davies P L, Braslavsky I. Langmuir, 2015, 31(21): 5805.

[48]	Mahatabuddin S, Hanada Y, Nishimiya Y, Miura A, Kondo H, Davies P L, Tsuda S. Sci. Rep., 2017, 7: 42501.

[49]	Marshall C B, Daley M E, Graham L A, Sykes B D, Davies P L. FEBS Lett., 2002, 529(2-3): 261.

[50]	Drori R, Davies P L, Braslavsky I. RSC Adv., 2015, 5(11): 7848.

[51]	Celik Y, Drori R, Pertaya-Braun N, Altan A, Barton T, Bar-Dolev M, Groisman A, Davies P L, Braslavsky I. PNAS, 2013, 110(4): 1309.

[52]	Jia Z C, Davies P L. Trends Biochem. Sci., 2002, 27(2): 101.

[53]	Bar-Dolev M, Celik Y, Wettlaufer J S, Davies P L, Braslavsky I. J. R. Soc. Interface, 2012, 9(77): 3249.

[54]	Middleton A J, Marshall C B, Faucher F, Bar-Dolev M, Braslavsky I, Campbell R L, Walker V K, Davies P L. J. Mol. Biol., 2012, 416(5): 713.

[55]	Kondo H, Mochizuki K, Bayer-Giraldi M. Phys. Chem. Chem. Phys., 2018, 20(39): 25295.

[56]	Mazur P. Am. J. Physiol., 1984, 247(3): 125.

[57]	Yu S O, Brown A, Middleton A J, Tomczak M M, Walker V K, Davies P L. Cryobiology, 2010, 61(3): 327.

[58]	Budke C, Dreyer A, Jaeger J, Gimpel K, Berkemeier T, Bonin A S, Nagel L, Plattner C, DeVries A L, Sewald N, Koop T. Cryst. Growth Des., 2014, 14(9): 4285.

[59]	Raymond J A, Janech M G, Fritsen C H. J. Phycol., 2009, 45(1): 130.

[60]	Guo S Q, Garnham C P, Whitney J C, Graham L A, Davies P L. PLoS One, 2012, 7(11): e48805.

[61]	DeVries A L. Annu. Rev. Physiol., 1983, 45(1): 245.

[62]	Yang D S C, Sax M, Chakrabartty A, Hew C L. Nature, 1988, 333(6170): 232.

[63]	Knight C A, Driggers E, DeVries A L. Biophys. J., 1993, 64(1): 252.

[64]	Zhang W, Laursen R A. J. Biol. Chem., 1998, 273(52): 34806.

[65]	Sönnichsen F D, DeLuca C I, Davies P L, Sykes B D. Structure, 1996, 4(11): 1325.

[66]	Haymet A D J, Ward L G, Harding M M. J. Am. Chem. Soc., 1999, 121(5): 941.

[67]	Baardsnes J, Davies P L. Biochim. Et Biophys. Acta BBA Proteins Proteom., 2002, 1601(1): 49.

[68]	Zhou Y X, Zhang Y, Tan H W, Jia Z C, Chen J G. Chem. J. Chinese U., 2007, 28(03): 144.

[69]	( 周艳霞, 张勇, 谭宏伟, 贾宗超, 陈光巨. 高等学校化学学报. 2007, 28(03): 144.)

[70]	Jorov A, Zhorov B S, Yang D S C. Protein Sci., 2004, 13(6): 1524.

[71]	Baardsnes J, Kondejewski L H, Hodges R S, Chao H M, Kay C, Davies P L. FEBS Lett., 1999, 463(1-2): 87.

[72]	Yang C, Sharp K A. Biophys. Chem., 2004, 109(1): 137.

[73]	Yang C, Sharp K A. Proteins, 2005, 59(2): 266.

[74]	Nutt D R, Smith J C. J. Am. Chem. Soc., 2008, 130(39): 13066.

[75]	Midya U S, Bandyopadhyay S. J. Phys. Chem. B, 2014, 118(18): 4743.

[76]	Sharp K A. PNAS, 2011, 108(18): 7281.

[77]	Naullage P M, Lupi L, Molinero V. J. Phys. Chem. C, 2017, 121(48): 26949.

[78]	Hudait A, Moberg D R, Qiu Y Q, Odendahl N, Paesani F, Molinero V. PNAS, 2018, 115(33): 8266.

[79]	Modig K, Qvist J, Marshall C B, Davies P L, Halle B. Phys. Chem. Chem. Phys., 2010, 12(35): 10189.

[80]	Liu K, Wang C, Ma J, Shi G, Yao X, Fang H, Song Y, Wang J. PNAS, 2016, 113(51): 14739.

[81]	Grabowska J, Kuffel A, Zielkiewicz J. J. Phys. Chem. B, 2019, 123(38): 8010.

[82]	Tien M Z, Meyer A G, Sydykova D K, Spielman S J, Wilke C O. PLoS One, 2013, 8(11): e80635.

[83]	Knight C A, Cheng C C, DeVries A L. Biophys. J., 1991, 59(2): 409.

[84]	Chakraborty S, Jana B. Phys. Chem. Chem. Phys., 2017, 19(18): 11678.

[85]	Pertaya N, Marshall C B, DiPrinzio C L, Wilen L, Thomson E S, Wettlaufer J S, Davies P L, Braslavsky I. Biophys. J., 2007, 92(10): 3663.

[86]	Ba Y, Wongskhaluang J, Li J B. J. Am. Chem. Soc., 2003, 125(2): 330.

[87]	Wang C, Liu J J. JIMU, 2019, 50(1): 65.

[88]	( 王超, 刘俊杰. 内蒙古大学学报, 2019, 50(1): 65.)

[89]	Shi M H, Liu J J, Zhao X J, Liu J P. JIMU, 2017, 48(5): 515.

[90]	( 史明卉, 刘俊杰, 赵雪珺, 刘俊平. 内蒙古大学学报, 2017, 48(5): 515.)

[91]	Arai T, Nishimiya Y, Ohyama Y, Kondo H, Tsuda S. Biomolecules, 2019, 9(5): 162.

[92]	Nishimiya Y, Kondo H, Takamichi M, Sugimoto H, Suzuki M, Miura A, Tsuda S. J. Mol. Biol., 2008, 382(3): 734.

[93]	Ewart K V, Li Z J, Yang D S C, Fletcher G L, Hew C L. Biochemistry, 1998, 37(12): 4080.

[94]	Liu Y, Li Z J, Lin Q S, Kosinski J, Seetharaman J, Bujnicki J M, Sivaraman J, Hew C L. PLoS One, 2007, 2(6): e548.

[95]	Chakraborty S, Jana B. Metallomics, 2019, 11(8): 1387.

[96]	Siemer A B, McDermott A E. J. Am. Chem. Soc., 2008, 130(51): 17394.

[97]	Garnham C P, Natarajan A, Middleton A J, Kuiper M J, Braslavsky I, Davies P L. Biochemistry, 2010, 49(42): 9063.

[98]	Garnham C P, Nishimiya Y, Tsuda S, Davies P L. FEBS Lett., 2012, 586(21): 3876.

[99]	Siemer A B, Huang K Y, McDermott A E. PNAS, 2010, 107(41): 17580.

[100]

Grabowska J, Kuffel A, Zielkiewicz J. J. Chem. Phys., 2016, 145(7): 075101.

[101]

Chakraborty S, Jana B. Phys. Chem. Chem. Phys., 2019, 21(35): 19298.

[102]

Kumari S, Muthachikavil A V, Tiwari J K, Punnathanam S N. Langmuir, 2020, 36(9): 2439.

[103]

Deng G, Laursen R A. Biochim. Et Biophys. Acta BBA Protein Struct. Mol. Enzymol., 1998, 1388(2): 305.

[104]

Gauthier S Y, Scotter A J, Lin F H, Baardsnes J, Fletcher G L, Davies P L. Cryobiology, 2008, 57(3): 292.

[105]

Lee J K, Kim H J. Fish. Aquat. Sci., 2016, 19(8): 33.

[106]

Tsvetkova N M, Phillips B L, Krishnan V V, Feeney R E, Fink W H, Crowe J H, Risbud S H, Tablin F, Yeh Y. Biophys. J., 2002, 82(1): 464.

[107]

Ahmed A I, Yeh Y, Osuga Y Y, Feeney R E. J. Biol. Chem., 1976, 251(10): 3033.

[108]

Giubertoni G, Meister K, DeVries A L, Bakker H J. J. Phys. Chem. Lett., 2019, 10(3): 352.

[109]

Pentelute B L, Gates Z P, Tereshko V, Dashnau J L, Vanderkooi J M, Kossiakoff A A, Kent S B H. J. Am. Chem. Soc., 2008, 130(30): 9695.

[110]

Lin F H, Graham L A, Campbell R L, Davies P L. Biophys. J., 2007, 92(5): 1717.

[111]

Todde G, Whitman C, Hovmöller S, Laaksonen A. J. Phys. Chem. B, 2014, 118(47): 13527.

[112]

Liou Y C, Tocilj A, Davies P L, Jia Z C. Nature, 2000, 406(6793): 322.

[113]

Bar M, Celik Y, Fass D, Braslavsky I. Cryst. Growth Des., 2008, 8(8): 2954.

[114]

Lee H. J. Mol. Graph. Model., 2019, 87: 48.

[115]

Midya U S, Bandyopadhyay S. J. Phys. Chem. B, 2018, 122(40): 9389.

[116]

Hudait A, Odendahl N, Qiu Y Q, Paesani F, Molinero V. J. Am. Chem. Soc., 2018, 140(14): 4905.

[117]

Grabowska J, Kuffel A, Zielkiewicz J. J. Mol. Liq., 2020, 306: 112909.

[118]

Sidebottom C, Buckley S, Pudney P, Twigg S, Jarman C, Holt C, Telford J, McArthur A, Worrall D, Hubbard R, Lillford P. Nature, 2000, 406(6793): 256.

[119]

Worrall D, Elias L, Ashford D, Smallwood M, Sidebottom C, Lillford P, Telford J, Holt C, Bowles D. Science, 1998, 282(5386): 115.

[120]

Gupta R, Deswal R. PLoS One, 2014, 9(3): e91723.

[121]

Hon W C, Griffith M, Mlynarz A, Kwok Y C, Yang D S C. Plant Physiol., 1995, 109(3): 879.

[122]

Gupta R, Deswal R. J. Proteome Res., 2012, 11(5): 2684.

[123]

Kuiper M J, Davies P L, Walker V K. Biophys. J., 2001, 81(6): 3560.

[124]

Middleton A J, Brown A M, Davies P L, Walker V K. FEBS Lett., 2009, 583(4): 815.

[125]

Hanada Y, Nishimiya Y, Miura A, Tsuda S, Kondo H. FEBS J., 2014, 281(16): 3576.

[126]

Do H, Kim S-J, Kim H J, Lee J H. Acta Crystallogr., 2014, 70(4): 1061.

[127]

Kondo H, Hanada Y, Sugimoto H, Hoshino T, Garnham C P, Davies P L, Tsuda S. PNAS, 2012, 109(24): 9360.

[128]

Cheng J, Hanada Y, Miura A, Tsuda S, Kondo H. Biochem. J., 2016, 473(21): 4011.

[129]

Meister K, Strazdaite S, DeVries A L, Lotze S, Olijve L L C, Voets I K, Bakker H J. PNAS, 2014, 111(50): 17732.

[130]

Marks S M, Patel A J. PNAS, 2018, 115(33): 8244.

[131]

Amir G, Rubinsky B, Kassif Y, Horowitz L, Smolinsky A K, Lavee J. Eur. J. Cardio-Thorac. Surg., 2003, 24(2): 292.

[132]

Amir G, Rubinsky B, Horowitz L, Miller L, Leor J, Kassif Y, Mishaly D, Smolinsky A K, Lavee J. Ann. Thorac. Surg., 2004, 77(5): 1648.

[133]

Khanna H K, Daggard G E. Plant Cell Rep., 2006, 25(12): 1336.

[134]

Holmberg N, FarrÉs J, Bailey J E, Kallio P T. Gene, 2001, 275(1): 115.

[135]

Graham B, Bailey T L, Healey J R J, Marcellini M, Deville S, Gibson M I. Angew. Chem. Int. Ed., 2017, 56(50): 15941.

[136]

Voets I K. Soft Matter, 2017, 13(28): 4808.

[137]

Xiang H, Yang X H, Ke L, Hu Y. Int. J. Biol. Macromol., 2020, 153: 661.

[138]

Kreder M J, Alvarenga J, Kim P, Aizenberg J. Nat. Rev. Mater., 2016, 1(1): 15503.

[139]

Deller R C, Vatish M, Mitchell D A, Gibson M I. Nat. Commun., 2014, 5: 3244.

[140]

Mizrahy O, Bar-Dolev M, Guy S, Braslavsky I. PLoS One, 2013, 8(3): e59540. DOI:10.1371/journal.pone.0059540.

[141]

Bonati L, Zhang Y Y, Parrinello M. PNAS, 2019, 116(36): 17641.

[142]

Lee J, Lee S Y, Lim D K, Ahn D J, Lee S. J. Am. Chem. Soc., 2019, 141(47): 18682.

[143]

Bai G Y, Song Z P, Geng H Y, Gao D, Liu K, Wu S W, Rao W, Guo L Q, Wang J J. Adv. Mater., 2017, 29(28): 1606843.

[144]

Xue B, Zhao L S, Qin X H, Qin M, Lai J C, Huang W M, Lei H, Wang J J, Wang W, Li Y, Cao Y. ACS Macro Lett., 2019, 8(10): 1383.

[145]

Jin S L, Yin L K, Kong B, Wu S W, He Z Y, Xue H, Liu Z, Cheng Q, Zhou X, Wang J J. Sci. China Chem., 2019, 62(7): 909.

[146]

Naullage P M, Molinero V. J. Am. Chem. Soc., 2020, 142(9): 4356.

[147]

Dou R M, Chen J, Zhang Y F, Wang X P, Cui D P, Song Y L, Jiang L, Wang J J. ACS Appl. Mater. Interfaces, 2014, 6(10): 6998.

[148]

Li S X, Lv R, Yan Z S, Huang F, Zhang X R, Chen G J, Yue T T. ACS Sustainable Chem. Eng., 2020, 8(10): 4256.

[149]

Qin Q Y, Zhao L S, Liu Z, Liu T, Qu J X, Zhang X W, Li R, Yan L Y, Yan J, Jin S L, Wang J J, Qiao J. ACS Appl. Mater. Interfaces, 2020, 12(16): 18352.