We have reviewed 92 cold-adapted enzymes that were successfully expressed in a heterologous host reported between 2010 and June 2016, which are detailed in Table 1. These enzymes were obtained mainly from psychrophilic or psychrotolerant organisms and bacteria or fungi (Figures 1A,B, respectively). These microorganisms were isolated from different and diverse environments, mainly from Polar Regions and marine environments. As explained later in this review, efficient catalysis at low temperatures requires an increase in protein flexibility, and therefore a reduction on enzyme stability. However, an interesting example of a cold-active enzyme isolated from a psycrophilic organism that had an unexpected high thermostability was reported for the superoxide dismutase DaSOD from Deschampsia antarctica (Rojas-Contreras et al., 2015). The optimal temperature of this enzyme is 20°C, it retains 80% of activity at 0°C and has detectable activity at −20°C, but also DaSOD possess high thermostability, its activity was not affected at 80°C, and the half-life time was 35 min at 100°C.

There are good examples of cold-active enzymes isolated from mesophilic organisms. Most of the time, a high activity at low temperatures is unexpected during the characterization of the catalytic properties of a mesophilic enzyme. This was the case of a Candida albicans lipase (Lan et al., 2011), which shows a low sequence identity with those of known lipases from psychrophilic organisms, but has an optimal temperature of 15°C. Other example is the lipase from Staphylococcus epidermidis, isolated from a car service area, with an optimal temperature of 25°C (Kamarudin et al., 2014). Other interesting case was reported by Monroe et al. (2014), where the cold-active properties from β-amylase 3 from Arabidopsis were inferred from the fact that this enzyme was more active during nighttime, compared to β-amylase 1 that had the opposite behavior. Both enzymes were overexpressed and purified from Escherichia coli confirming that they were differentially thermal adapted. β-amylase 3 had a lower optimal temperature, greater residual activity at low temperatures and less thermal stability than β-amylase 1.

More surprising is to discover a thermophilic enzyme with high activity at low temperatures. This was the case of a β-galactosidase isolated from Pyrococcus furiosus (Dong et al., 2014) with optimal activity at 90°C (130 U/mg). The enzyme was still active at 0°C, retaining 8% of its activity. Despite the decrease in activity compare to its optimal temperature, the lactase activity of P. furiosus at 0°C was still 40% of the optimal activity from the main β-galactosidase use in the food industry (28 U/mg at 50°C and pH 7.0) from K. marxianus. In addition, the lactase activity of P. furiosus at 0°C was 31% of the optimal activity of a cold-active β-galactosidase from Arthrobacter psychrolactophilus strain F2 (33 U/mg at 10°C and pH 8.0).

Several strategies have been suggested to promote proper expression and folding of cold-active enzymes expressed in heterologous host, increasing their solubility, activity, and yield. These strategies are summarized in Figure 5 and include the use of: (i) molecular chaperones; (ii) cold-active promoters; (iii) fusion partners; (iv) psychrophilic hosts, and (v) a combination of these strategies.

i) Molecular Chaperones

Molecular chaperones are ubiquitous proteins that help newly synthesized polypeptides and denatured proteins to reach their native conformation. They are widely distributed in bacteria, yeast, plants, and animals (Evstigneeva et al., 2001). Originally, they were discovered because their encoding genes were expressed under heat shock conditions, therefore this family of molecules was named heat shock proteins (HSP). However, genes encoding molecular chaperones are also induced under other stressful conditions including UV irradiation, hypoxia, and chemical challenges, among others (Whitley et al., 1999). There are various chaperones families, which are named by their molecular size ranging from HSP40 to HSP100 and the small HSPs (Hartl et al., 2011). In E. coli, chaperones with quaternary structure are also known as chaperonins and include the GroEL/GroEL systems, whereas monomeric chaperones include the DnaK/DnaJ systems (Bukau et al., 2000; de Marco, 2007).

In 2004, Ferrer and coworkers reported that expressing chaperones Cpn60 and Cpn10 from the psychrophilic bacterium Oleispira antarctica RB8 in E. coli, lowered its minimal growth temperature below 15°C (Ferrer et al., 2003, 2004a). They used this chaperone-E. coli system to express a heat-labile esterase, reporting for the first time a successfully expression system for heat-sensitive proteins. They demonstrated that the low temperature improved proper folding of the enzyme, enhancing its specific activity 180-fold in comparison to the enzyme purified from the usual E. coli strain grown at 37°C (Ferrer et al., 2004b). Today, a competent E. coli strain that co-express cold-active chaperones Cpn60 and Cpn10 is commercialized by Agilent Technologies under the name of ArcticExpress.

Kim et al. (2015) co-expressed a cold active esterase together with PsyGroELS, a chaperonin from the psychrophilic bacterium Psychrobacter sp. PAMC21119, in an E. coli strain. The expression was performed at 10°C and they compared the enzyme activity using the previously reported chaperones Cpn60 and Cpn10, finding better results with PsyGroELS for this particular enzyme. They conclude that PsyGroELS not only confers cold-tolerance to E. coli, but also is effective for co-expression of stable psychrophilic proteins.

Another example of chaperone co-expression was recently described by Esteban-Torres et al. (2014a) using GroES/GroEL chaperones. First, they cloned the cold-active esterase lp_2631 into the pURI3-TEV expression vector for protein production, but the recombinant protein was expressed as inclusion bodies when E. coli BL21 (DE3) was used as host. To solve this, they used the plasmid pGro7 that produces GroES/GroEL chaperones. When Lp_2631 was co-expressed with the molecular chaperones in the E. coli host, the protein was expressed in the soluble fraction of the cells.

ii) Cold-active promoters

Quing and coworkers developed cold-shock expression vectors (pColdI-IV) harboring the cspA promoter from CspA, the major cold shock protein of E. coli, allowing high expression of several genes upon induction by cold-shock (Qing et al., 2004). They reported that pCold vectors are highly complementary to the widely used pET vectors for the expression of 38 genes. pCold vectors have been used to functionally express various proteins in E. coli at low temperature, most of them from mesophilic organisms including human proteins that were difficult to obtain with other systems (Hayashi and Kojima, 2008). Surprisingly only a few examples have been described for the expression of enzymes from psychrophilic organisms. One of them corresponds to the pCold I vector, used to functionally express a cold-active β-galactosidase (rBglAp) that was found to be extremely heat-labile in E. coli (Nakagawa et al., 2007).

Shuo-shuo and coworkers cloned the cold-active lipase gene Lip-948, from the Antarctic psychrotrophic bacterium Psychrobacter sp. G into the plasmid pColdI and transformed it into E. coli BL21, obtaining substantive expression of lipase LIP-948 with a yield of 39% of total protein, most of which was present as inclusion bodies (Shuo-shuo et al., 2011). Co-expression of pColdI-Lip-948 with chaperone pTf16 and pGro7 decreased the amount of insoluble LIP-948, while the soluble expression was enhanced when pColdI-Lip-948 was co-expressed with “chaperone team” plasmids (pKJE7, pG-Tf2, pG-KJE8), respectively. LIP-948 was most efficiently expressed in soluble form when it was co-expressed with pG-KJE8, which was up to 19.8% of intracellular soluble proteins. Also, pCold vectors have been used for the expression of proteins with fusion partners, as it is described below.

iii) Fusion Partners

Fusion partners are solubility-enhancing tags used to increase both the solubility and expression level of recombinant enzyme expression (Hayashi and Kojima, 2010). They are located at the N- or C-terminus of the target protein and in some cases a specific cleavage site is placed between the tag and the target protein to allow their excision after purification. The most common fusion partners correspond to glutathione-S-transferase (GST), maltose-binding protein (MBP), thioredoxin (Trx), Gβ-1 domain of protein G (Gb1), nascent chain chaperone trigger factor (TF), small ubiquitin-like modifier (SUMO), and N-utilizing substance A (NusA).

Regarding expression of cold-adapted enzymes using fusion partners, Trx fusion tag has been used to obtain high yield of soluble psychrophilic yeast proteins in E. coli host (Illias et al., 2014). Moreover, the effects of seven different N-terminal fusion partners were studied to improve the solubility of proteins from the psychrophilic fish Vibrio salmonicida in E. coli. Among the fusion partners, MBP and NusA showed to be the best for expression yield and protein solubility (Niiranen et al., 2007). In addition, two different expression host strains and three cell culture incubation temperatures were used. Concerning the host strain, E. coli BL21-AI was shown to be superior to BL21(DE3)RIL CodonPlus for protein expression, but the product solubility was not affected by the choice of host. In terms of the incubation temperature for protein expression, the protein yield increased with temperature, although the effect on solubility was the contrary in most cases. They also concluded that small proteins were easier to express.

Another example for the soluble expression of a cold-active enzyme using MBP as the fusion partner was reported for a lipase from marine Antarctic origin (Parra et al., 2008). First, the expression system E. coli BL21(D3E)/pET22b(+) was used but the protein was obtained as inclusion bodies. After using the expression system E. coli TB1/pMAL-c2E, which expressed a fusion MBP-lipase protein, the enzyme was obtained in a soluble an active form. Hayashi and Kojima (2008) used the pCold I vector and modified it in order to express proteins fusioned to a GST tag. They were able to successfully express 9 proteins which they could not obtain using a conventional E. coli expression system. Later, the same authors used the pCold-GST system to successfully express 78 proteins from mesophilic organisms, showing that the primary sequence length of these proteins was not correlated with the expression level in the soluble fraction. They also developed three other cold-shock vectors using the fusion partners GB1, Trx, and MBP, showing that all systems were successful in obtaining soluble fusion proteins, with the pCold-GST system being the preferred and the pCold-MBP system the second choice. Furthermore, they reported that the use of a C-terminal 6 proline tag was successful in inhibiting the degradation of the protein during protein expression and purification, therefore being useful for enzyme stabilization (Hayashi and Kojima, 2010).

iv) Psychrophilic hosts

To overcome the decrease in protein yield and overall process productivity when E. coli strain is cultured at low temperatures, psychrophilic bacteria have been used as expression hosts (Parrilli et al., 2008b).

The most studied psychrophilic host is Pseudoalteromonas haloplanktis TAC125, which uses a modified E. coli cloning vector with psychrophilic molecular signals. This host was reported for the expression of a cold-adapted α-amylase as secretion carrier for extra-cellular protein targeting (Cusano et al., 2006a,b). Later, authors developed a P. haloplanktis TAC125 mutant strain that secreted a reduced number of exo-proteases, therefore reducing the extra-cellular proteolytic activity (Parrilli et al., 2008a).

In other study, an expression vector derived from psychrophilic bacterium Pseudoalteromonas sp. BSi20429 was constructed and Pseudoalteromonas sp. SM20429 was used as the psychrophilic bacterial strain. The system was first reported using a mesophilic promoter from E. coli and used for the active expression of a cold-adapted cellulase at 25–30°C (Zhao D. et al., 2011). Later, the mesophilic promoter was replaced by another from Pseudoalteromonas sp. BSi20429 that acted as a strong promoter at low temperatures and was also inducible by xylan, thus enabling the recombinant expression at lower temperatures. Multiple cloning sites and a His tag were also added to the expression vector, making these system useful for expressing Pseudoalteromonas enzymes that could not be maturely expressed in E. coli (Yu et al., 2015).

Another interesting study is an example to overcome the barrier of studying polyextremophilic enzymes. For halophilic enzymes, a high salt concentration is a requirement to obtain an active protein during overexpression in heterologous hosts. Karan et al. (2013) purified and characterized a halophilic and cold-active β-galactosidase from the cold-adapted haloarchaeon, H. lacusprofundi. They used the haloarchaeon, Halobacterium sp. NRC-1 strain as host in combination with a cold-shock protein gene promoter, cspD2, also from the host. They produced the recombinant β-galactosidase at 20-fold higher levels compared to H. lacusprofundi.

Finally, an example of a eukaryotic expression system for genes codifying cold-active enzymes comes from the work performed by Mao et al. (2015), who developed a novel uracil-deficient Aspergillus oryzae host for heterologous expression. This system was used to express an α-amylase from the psychrophilic fungus Geomyces pannorum.

v) Combination of the above strategies

Combination of these strategies has also been successfully used for the expression and purification of cold-adapted enzymes. A cold-adapted endo-1,4-β-glucanase from the earthworm Eisenia fetida was cloned in the pColdI vector and successfully expressed using the host strain ArcticExpress RT (DE3) (Ueda et al., 2014). Bjerga and Williamson (2015) optimized an expression system for cold-adapted proteins based on the pCold-II vector. They expressed five genes derived from metagenomic DNA from marine Arctic sediments and used three hosts strains including BL21 CodonPlus(DE3)RIL, ArcticExpress(DE3)RIL, and Rosetta2(DE3)pLysS, obtaining the best results using the latter. The yields of soluble protein were increased using fusion partners like MBP, TF, TRX, and SUMO, reporting the best results using large fusion partners like MBP and TF (Bjerga and Williamson, 2015).

The main mechanistic goal of the evolutionary adaptations in cold-active enzymes is to maintain a high catalytic activity at low temperatures. These activities are required to sustain metabolic activity in extremely cold environments, in some cases even near −20°C (Rivkina et al., 2000).

At very low temperatures the kinetic energy is insufficient to allow overcoming the kinetic barriers associated with an enzymatic reaction (Siddiqui and Cavicchioli, 2006). Nevertheless, cold adapted enzymes generally have optimum temperatures of activity and higher reaction rates at lower temperatures than their mesophilic homologs. We can rationalize this behavior if we examine the temperature dependence of the rate of chemical reactions as envisioned by the Arrhenius equation (Laidler, 1984): (1)kcat=Ae-Ea/(RT) Where k_cat is the catalytic rate, E_a is the activation energy of the reaction, R is the gas constant, T is temperature and A is a collision frequency factor. This equation illustrates how the catalytic rate depends on the temperature, such that it decreases upon decreasing temperature. For example, the catalytic rate of a mesophilic enzyme with Ea values ranging 50–75 kJ·mol⁻¹ decreases 2–3-fold upon lowering the temperature every 10°C (Tattersall et al., 2012).

As suggested by the Arrhenius equation, the detrimental effect of lowering the temperature on the catalytic turnover can be countered by decreasing the activation energy, such that the thermal dependence of the catalytic reaction is reduced. Extensive reviews have been made about the changes in activation energies of chemical reactions catalyzed by psychrophilic enzymes and their mesophilic and thermophilic homologs, consistently showing that the activation energies are in fact decreased in cold-adapted enzymes (Lonhienne et al., 2001; Matsuura et al., 2002; D'Amico et al., 2002b, 2003a,b; Mavromatis et al., 2003; Garsoux et al., 2004; Liang et al., 2004; Fedøy et al., 2007; Leiros et al., 2007; Lian et al., 2015).

The temperature-dependence of reaction rates given by the Arrhenius equation explains the increase in k_cat needed for catalyzing reactions at low temperature under saturating substrate concentrations, whereas substrate binding represented by the Michaelis-Menten constant (K_m) could also play an important role in increasing the catalytic efficiency (defined as k_cat/K_m) of cold-adapted enzymes at low temperatures (Feller and Gerday, 1997; D'Amico et al., 2002a). We can rationalize how these changes in the kinetic parameters are tuned within a cold-adapted enzyme through the transition state theory, in which equilibrium between the ground (ES) and a stable activated (ES^‡) enzyme-substrate complex is assumed: (2)E+S⇌ES⇌ES‡⇀E+P In this scenario, the temperature dependence of the catalytic rate can be related to the free energy of activation (ΔG^‡) between the ground and activated that has to be overcome during an enzymatic reaction using the Eyring equation (Eyring, 1935): (3)kcat=kBThe-△G‡RT=kBThe(△S‡R-△H‡RT) Where k_B is the Boltzmann constant, h is the Planck constant and ΔH^‡ and ΔS^‡ are the enthalpic and entropic contributions to the free energy of activation, respectively.

The comparative analysis of the kinetic parameters of a large number of psychrophilic enzymes has shown that their k_cat at low temperatures is similar to those observed for mesophilic enzymes at warm temperatures (D'Amico et al., 2002a; Siddiqui and Cavicchioli, 2006). In that scenario, either ΔH^‡ must decrease or ΔS^‡ must increase, as it is clear from Equation (3) (Lonhienne et al., 2000).

The contribution of ΔH^‡ can be understood in terms of the interactions that are broken while transitioning from the ground enzyme-substrate complex to the transition state of the reaction (Figure 6). Thus, a decrease of the enthalpic contribution translates into a reduction of the number of interactions that must be broken during this process (Siddiqui and Cavicchioli, 2006). This enthalpy decrease for psychrophilic enzymes is consistent with the decrease of the activation energy of the reactions catalyzed by these enzymes, as ΔH^‡ = E_a − RT (Lonhienne et al., 2000).

Compensation of the change of enthalpic contributions for the formation of the transition state is achieved by an antagonist change in ΔS^‡, which explains why k_cat does not increase exponentially upon changing ΔH^‡ as a product of the cold adaptation of psychrophilic enzymes (Lonhienne et al., 2000). This compesation is such that the activation entropy difference between a mesophilic and a psychrophilic enzyme is always negative and the absolute value of their entropy difference, TΔ(ΔS^‡), is always large (Lonhienne et al., 2000). This entropic compensation can be conceptualized in the context of the protein structure as an increase in flexibility of regions of the protein covering the enzyme's active site or other extensive changes in flexibility throughout the protein (Gerday et al., 1997). Extensive evidence from enzyme kinetics has shown that this is true for all cold-adapted enzymes studied so far (Siddiqui and Cavicchioli, 2006). Moreover, as a result of these changes in entropy and enthalpy, a small reduction of the free energy of activation and an increase of the conformational distribution of the ground state of the enzyme-substrate complex occurs (Figure 6). This trade-off between activity and stability is what leads to a small reduction of the free energy of activation in cold-adapted enzymes.

The enthalpic-entropic changes experienced by psychrophilic enzymes and represented in Figure 6 have two different consequences. First, increasing the flexibility of an enzyme through changes in plasticity of the active site leads to increased substrate promiscuity (Nobeli et al., 2009) because substrates with small variations in size and conformation can now fit into the more accesible binding site (Struvay and Feller, 2012; Feller, 2013) as it has been demonstrated for cold-adapted Sporosarcina psychrophila acylaminoacyl peptidase (Brunialti et al., 2011), Shewanella gelidimarina nitrate reductase, Psychromonas ingrahamii serine hydroxymethyltransferase (Angelaccio et al., 2012) and Psychrobacter sp. aminotransferase (Bujacz et al., 2015). This broader substrate utilization can be advantageous for protein engineering strategies focused in enhancing the specificity toward chemical reactions of biotechnological interest (Zhang et al., 2016). Second, these free energy changes, particularly the enthalpic changes, cause a decrease in substrate binding affinity. In this context, cold-active enzymes increase their k_cat at the expense of an increase in K_m (Feller and Gerday, 2003). In fact, stepwise single and multiple mutations engineered on a psychrophilic α-amylase to reconstruct the amino acid substitutions found in a mesophilic homolog exhibit a striking correlation of k_cat and K_m, such that both decrease concomitanly upon increasing the number of mesophilic residues in the cold-adapted enzyme (Cipolla et al., 2011). Nevertheless, some enzymes from psychrophilic organisms that operate under subsaturating substrate concentrations within the cytoplasm exhibit a decrease in this kinetic parameter as an evolutionary strategy for cold adaptation (Bentahir et al., 2000; Hoyoux et al., 2001; Lonhienne et al., 2001).

In summary, cold-adapted enzymes generally exhibit an increase of their catalytic rate (k_cat) allowed by a decrease in enthalpy due to a reduced number of protein-ligand interactions and an increase in entropy due to changes in their stability and flexibility, which can also lead to advantageous properties such as substrate promiscuity. In the following section we rationalize how these changes in stability and flexibility are embodied in the primary, secondary, tertiary (and sometimes quaternary) structure of these enzymes.

Thermophilic enzymes are known for having a higher thermostability than mesophilic enzymes and for being poor biocatalysts at room temperature (Gerday et al., 2000). Such thermostability, which is required to withstand heat denaturation at high temperatures, leads to increased conformational rigidity at temperatures where mesophilic enzymes usually catalyze their reactions (Závodszky et al., 1998). Interestingly, the conformational fluctuations are similar when comparing mesophilic and thermophilic enzymes at their respective optimal activity temperatures in which both K_m and k_cat are also optimal, the so-called “corresponding state” hypothesis (Závodszky et al., 1998). This evidence led to conclude that evolutionary adaptations, in the form of sequence and structure changes, allow a balance between protein stability and conformational flexibility that are responsible of proper function in the environmental niche's temperature of the source organism. In consistency with this idea, it has been argued that the plasticity or flexibility of cold-adapted enzymes is what enables their high specific activity at low temperatures and with a low energy cost (Gerday et al., 2000). It is now broadly accepted that the trade-off between thermostability and activity, and in particular the balance between stability and flexibility, is what evolves in enzymes in order to suit different environmental niches: for enzyme catalysis to be efficient at low temperatures protein flexibility must be increased, otherwise the reduced thermal fluctuations will diminish the conformational mobility and consequently compromise catalytic efficiency (Arnold et al., 2001). Moreover, it has been suggested that the encounter of cold-active enzymes with optimal activities at temperatures higher than their physiological conditions is evidence of an incomplete evolutionary adaptation to low temperatures (Georlette et al., 2004).

Some of the first and most detailed evidences of this apparent increase in conformational flexibility came from the study of A4 lactate dehydrogenases (A4-LDH) from nine Antarctic and three South American notothenioid teleosts, which inhabited niches with temperatures ranging from −1.8 to 10°C (Fields and Somero, 1998). Enzyme activity assays revealed that the catalytic rate of A4-LDH from teleosts inhabiting the coldest environments were higher at 0°C than their homologs, with k_cat decreasing linearly as a function of average body temperature. More importantly, deduction of their amino acid sequences from RT-PCR and DNA sequencing showed that most of the minimal residue substitutions between A4-LDH that led to these catalytic differences were not distributed randomly, but located in two regions in the vicinity of the active site (helix αH and an extended loop connecting an helix with catalytic residues) whose conformational changes are rate-limiting steps for catalysis (Figure 7). Their results suggested that the observed substitutions increased the flexibility of these regions, leading to more rapid conformational changes and thus increasing k_cat (Fields and Somero, 1998).

Similar suggestions of the increased flexibility of cold-adapted enzymes were made based on gene cloning, protein purification and sequence analysis, combined with homology modeling of several enzymes generated using already crystallized mesophilic and thermophilic enzymes as templates (Russell, 2000). Such analysis led to the identification of interesting amino acid substitutions consistently found to occur in several cold-adapted enzymes when compared to their mesophilic and thermophilic homologs, such as the reduction of the number of surface salt bridges due to replacement of basic residues by glutamine or asparagine, changes in the distribution of surface charges, a reduced hydrophobicity of the protein core due to substitutions of bulky aromatic residues by more flexible nonpolar residues, a decrease in the number of hydrogen bonds in the protein structure, an increase in length of loop regions, among others (Davail et al., 1994; Feller et al., 1994; Smalås et al., 1994; Feller and Gerday, 1997; Russell, 2000). However, confirmation of these changes through the resolution of crystal structures of cold-adapted enzymes was lacking, mostly due to the difficulty of crystallizing these proteins (Russell, 2000).

The first crystal structures of cold-adapted enzymes were obtained for elastase (Berglund et al., 1995) and trypsin (Smalås et al., 1994) of Atlantic salmon, whereas the first solved crystal structures of bacterial psychrophilic enzymes corresponded to α-amylase (Aghajari et al., 1998), triose phosphate isomerase (Alvarez et al., 1998), and citrate synthase (Russell et al., 1998). Currently there are more than 50 different cold-adapted enzymes deposited in the Protein Data Bank (Berman et al., 2000), most of them coming from psychrophilic bacteria. The accession codes of most of these solved structures (excluding repeated structures of the same enzymes with ligands or mutations) are provided in Table 2, including a few remarkable examples of the use of metagenomic libraries to collect novel cold-adapted enzymes (Fu et al., 2013). It is worth noting that most of these enzymes correspond to hydrolases (Table 2), which comes as no surprise given that these are the most identified and studied cold-adapted enzymes for biotechnological and industrial applications, as we will see later on in this review.

Comparison of these deposited structures against mesophilic and thermophilic homologs confirmed that only minor structural modifications are needed to adapt warm-adapted enzymes to cold temperatures and that active-site residues involved in the reaction mechanisms are strictly conserved between homologous enzymes adapted to different temperatures (D'Amico et al., 2002a). Such comparisons allow determining the preferred amino acid exchanges and the localization of these changes within the protein structure, the variety of evolutionary strategies toward cold adaptation and also enable the reconstruction of the evolutionary steps that mediate temperature adaptations in the laboratory via rational design (Tsigos et al., 2001; Mavromatis et al., 2003).

In terms of changes in protein sequence, a systematic comparative analysis of multiple sequence and structure alignments containing 21 psychrophilic enzymes belonging to different structural families and 427 homologous mesophilic and thermophilic allowed to create a distance matrix of residue substitutions often found to allow adaptation to low temperatures: charged residues Arg and Glu tend to be replaced at exposed sites on α-helices by Lys and Ala, respectively; Val is replaced by Ala at buried regions in α-helices; and the content of Ala and Asn increases whereas Arg decreases in exposed sites (Gianese et al., 2001). Very similar results were recently obtained using archaeal genome analysis combined with high-throughput homology modeling (Saunders et al., 2003) and also using proteome-wide approaches on six completely sequenced species of psychrophilic and mesophilic bacteria (Metpally and Reddy, 2009). However, the location of these and other substitutions and the number of substitutions vary on a great extent depending on the enzyme under examination, meaning that each protein family adopts different structural strategies to adapt to low temperatures (Gianese et al., 2002).

For example, in the case of M. marina triose phosphate isomerase, a single substitution of an alanine located within a loop that contacts the phosphate moiety of its substrate by a serine that is conserved in mesophilic enzymes is sufficient to increase the thermal stability and decrease the catalytic activity at low temperatures (Alvarez et al., 1998). The same is applicable in some cases for tuning mesophilic enzymes in order to sustain catalytic activities in the cold, as exemplified by the rationally designed single-point mutation I137M of Bacillus subtilis LipJ (Goomber et al., 2016b). Most frequently, evolutionary changes are related to multiple changes that lead to a more accessible and/or a more flexible active site due to substitution of bulky residues, insertions and deletions (Russell et al., 1998; Kim et al., 1999; Schrøder Leiros et al., 2000; Toyota et al., 2002; Aghajari et al., 2003; Van Petegem et al., 2003; Tsuruta et al., 2005, 2008; Leiros et al., 2007; Riise et al., 2007; Jung et al., 2008; Merlino et al., 2010; Jaremko et al., 2011; Malecki et al., 2013; Zheng et al., 2016), which in some cases are accompanied by the introduction of discrete amino acid substitutions in the active site that thermodynamically favor protein-ligand interactions at low temperatures, thus decreasing K_m (Lonhienne et al., 2001). Finally, the most extensive changes involve large portions throughout the protein structure and are related to optimization of the surface electrostatic potential to allow better interactions with the solvent and changes in ion-pair interactions (Bell et al., 2002; de Backer et al., 2002; Leiros et al., 2003; Bae and Phillips, 2004; Kumar and Nussinov, 2004; Arnórsdóttir et al., 2005; Helland et al., 2006; De Vos et al., 2007; Fedøy et al., 2007; Wang et al., 2007; Michaux et al., 2008; Pedersen et al., 2009; Alterio et al., 2010; Arimori et al., 2013; Bujacz et al., 2015), reduction of the number of hydrogen bonds (Matsuura et al., 2002; Bae and Phillips, 2004; Altermark et al., 2008; Michaux et al., 2008; De Santi et al., 2016), changes in loop extension, amino acid content, and flexibility (Bauvois et al., 2008; Helland et al., 2009; Zhang et al., 2011; Fu et al., 2013; Miao et al., 2016; Zheng et al., 2016), introduction or loss of disulfide bonds to modulate local stability (Violot et al., 2005; Helland et al., 2006; Wang et al., 2007), differential flexibility of domains in multidomain enzymes (Watanabe et al., 2005; Bauvois et al., 2008; Angelaccio et al., 2014), and enhanced protein solvation due to increased exposure of hydrophobic residues to the solvent (Aghajari et al., 1998; Russell et al., 1998; Maes et al., 1999; Bell et al., 2002; Van Petegem et al., 2003; Zhao Y. et al., 2012; Zheng et al., 2016). A summary of the most usual modifications responsible for cold-adaptation are shown in Figure 8. It is worth noting that not all of these mechanisms are required to explain the cold-adaptation of a given enzyme (De Maayer et al., 2014), although several proteins exhibit more than one of these mechanisms occurring in parallel (Coquelle et al., 2007), which suggest that comparative analysis within protein families might be better suited to solve the sequence-structure factors that explain the evolutionary adaptations of an enzyme of interest. Although it is rare to find proteins showing other mechanisms of cold adaptation, more extensive changes in protein topology (Tsuruta et al., 2005) or modifications of the oligomerization state that allows to increase the flexibility of solvent-exposed hydrophobic regions while simultaneously stabilizing the native fold of the enzyme (Skalova et al., 2005; Zanphorlin et al., 2016) have been also observed. However, these should be considered as evolutionary alternatives rather than as general mechanisms for enhanced flexibility in cold environments.

While solving the structures of these enzymes helped to deepen our understanding of the molecular mechanisms behind adaptation to cold temperatures, they only provide a static view of the position of these sequence changes within the three-dimensional space, thus lacking of an exploration of protein dynamics, with the only exception of those structures solved by NMR (Jaremko et al., 2011). Therefore, the combination of this structural information with experiments that assess the conformational flexibility of cold-adapted enzymes and the direct use of these structures as inputs for molecular dynamics is crucial to provide a solid framework for further experimental and computational protein engineering approaches.

Among the experiments performed to demonstrate the increased flexibility of cold-adapted enzymes, dynamic quenching of tryptophan fluorescence by increasing concentrations of acrylamide (Eftink and Ghiron, 1975) is commonly used. Acrylamide ascertains the accessibility of tryptophan residues within a protein as a decrease in fluorescence by means of physical contact (Eftink and Ghiron, 1976), thus reflecting the ability of the quencher to penetrate the protein structure and providing information of its permeability (D'Amico et al., 2003b). Typically, the fluorescence quenching constants (as reported by the Stern-Volmer constant) of psychrophilic enzymes are higher than for mesophilic proteins at both low and warm temperatures, thus indicating a more permeable structure (Huston et al., 2008; Tang et al., 2012), and the variation of fluorescence quenching (i.e., the change in the Stern-Volmer constant) within a temperature range where the native state prevails decreases in the order psychrophilic > mesophilic > thermophilic (D'Amico et al., 2003a; Georlette et al., 2003, 2004; Cipolla et al., 2012), thus indicating that cold-adapted enzymes possess higher flexibility. These experiments can also be combined with mutational analysis to explore the interplay between sequence variation, protein flexibility, and catalytic activity (Cipolla et al., 2011; Sigtryggsdóttir et al., 2014; Truongvan et al., 2016).

Further identification of the spots responsible for increased flexibility within a psychrophilic protein requires assessment of local regions of the protein. A successful approach for such task consists of the use of chimeric enzymes, in which a gene encoding for a given psychrophilic protein is divided into several regions that are then replaced by similar regions from a mesophilic homolog (Yoneta et al., 2004; Watanabe et al., 2005). This strategy allows not only to confirm that in some cases the entire protein does not necessarily need to be flexible to achieve high catalytic activity in at low temperatures, but also to identify which protein regions are responsible for the increased flexibility (Yoneta et al., 2004), which can be further combined with mutational analysis to identify the key residues responsible for cold-adaptation (Hayashi et al., 2014). These experiments allowed the identification of the C-terminal region of the cold-adapted isocitrate dehydronease from Colwellia maris as responsible for its psychrophilic characteristics (Yoneta et al., 2004). Another powerful strategy corresponds to amide hydrogen/deuterium exchange mass spectrometry (Balasubramaniam and Komives, 2013), in which the exchange between backbone amide protons and the deuterium from the surrounding solvent is used as a mass probe for the solvent accessibility of a protein, whereas quenching and pepsin digestion of the protein followed by mass spectrometry analysis of the resulting peptides allows to localize the sites of exchange within the protein. In these experiments, highly flexible regions become fully deuterated in a few minutes, whereas well-packed regions such as the hydrophobic core exhibit a low extent of exchange. The advantages of this strategy is that it can be applied to proteins of any size (Balasubramaniam and Komives, 2013), under varying temperature (Ramírez-Sarmiento et al., 2013), and solvent conditions (Medina et al., 2016) and in the absence and presence of ligands (Chalmers et al., 2011). Comparative analysis of deuterium incorporations of local regions of a psychrophilic and a thermophilic alcohol dehydrogenase led to strengthen the notion that only those functional regions related to substrate binding exhibit greater flexibility in the cold-active enzyme than in the warm-adapted homolog, suggesting that local flexibility can be uncoupled from thermal stability (Liang et al., 2004).

The use of solved or homology modeled psychrophilic protein structures in molecular dynamics allows the assessment of dynamical features relevant for cold adaptation with atomistic resolution. Constant temperature simulations of psychrophilic enzymes at several temperatures within the range 10–45°C enabled the identification of loops near active sites that exhibit higher flexibility in comparison with their mesophilic homologs, as in the case of Uracil-DNA glycosylases (Olufsen et al., 2005), elastases (Papaleo et al., 2006), and β-glucosidases (Zanphorlin et al., 2016), and the optimization of ion-pair networks near the active sites of elastases (Papaleo et al., 2007) and serine proteases (Tiberti and Papaleo, 2011). In some cases, these simulations show good correlation between the optimal temperature of catalytic activity and the increased flexibility of functional regions of the protein (Aurilia et al., 2009) and also highlight other loops distant from the active site that exhibit preservation of similar flexibilities between psychro-, meso-, and thermophilic enzymes at their optimal temperature for catalysis (Kovacic et al., 2016). Other quasi-harmonic entropy approximations have been used for comparative analysis of simulations of psychrophilic enzymes in their free and substrate-bound forms, allowing the identification of key determinants of structural flexibility at the residue-level (Kosugi and Hayashi, 2011). More recently, complex molecular simulation strategies have allowed bridging the reduction of the activation energies of enzymatic reactions with the increased flexibility of cold-adapted enzymes. Hybrid quantum-mechanics/molecular-mechanics simulations on a psychrophilic α-amylase from Pseudoalteromonas haloplanctis revealed that formation of the transition state of the enzymatic reaction is accompanied by a rearrangement of a loop neighboring the active site, such that it interacts with the substrate via water-mediated and direct interactions, and is crucial for the reduction of the free energy barrier of the hydrolysis reaction (Kosugi and Hayashi, 2012). Moreover, energetic estimations of peptide hydrolysis by psychrophilic trypsins calculated using free energy perturbation simulations in which the flexibility of the protein surface is systematically reduced through position restraints of different strengths, showed that this protein rigidity is sufficient to increase the activation energy as in mesophilic enzymes, thus strongly suggesting that softness of the protein-water surface is what tunes the temperature adaptation of catalytic rates (Isaksen et al., 2016).

The molecular mechanisms of cold-adaptation and the hotspots of conformational flexibility captured through the application of these experimental and computational approaches are not only compelling evolutionary and theoretical challenges to pursue, but also provide paramount information to integrate in protein engineering and design endeavors. As we will see below, a vast number of rational design and directed evolution approaches used to improve catalysis at low temperatures are proposed based on localized conformational flexibility spots revealed by these types of analysis.

Enzymes used in chemical processes need to be active in the presence of other additives required for these reactions. Organic solvents are widely used, either pure or mixed with aqueous solvents. In general, to obtain some information about cold-active enzymes after purification, residual enzymatic activity was studied in the presence of different additives, such as metal ions, EDTA, DTT, β-mercaptoethanol, and protease inhibitors. In addition, enzyme stability upon addition of organic solvents and salt was also studied for some cold-active enzymes. The effect of different additives was assessed for 76% of the cold-active enzymes examined in Table 1.

In reaction mixtures, organic solvents are used to increase the solubility of hydrophobic substrates, as in biodiesel production through transesterification reactions using lipases and esterases. However, enzyme activity is reduced in organic solvents because water molecules are lost (Doukyu and Ogino, 2010). At low temperatures, cold-active enzymes are able to grasp more tightly to available water molecules because they have a low inherent surface hydrophobicity (Karan et al., 2012). For this reason, cold-active enzymes preserve their catalytic activity in organic solvents because they are able to maintain a tight hydration shell. One example of a cold-active esterase that is active under high concentrations of different solvents and additives is the esterase LipA from Sorangium cellulosum (Cheng et al., 2011). The enzyme retained high level of activity in the presence of 0.1-1% of the commercially available detergents (Tween 20, Tween 80, Triton X-100). Also, the activity was tested after incubation with 16 different solvents. Diethylether, chloroform, benzene, toluene, p-xylene, cyclohexane, n-hexane, n-heptane, and isooctane increase LipA activity and the others have minor negatives effects, conserving always more than 50% activity. Another study of a cold-active enzyme stable in organic solvent was reported for lipase AT2 from S. epidermidis (Kamarudin et al., 2014). This cold-active enzyme was found stable in both hydrophilic and hydrophobic organic solvents. The enzyme displayed stability not only in methanol, ethanol and acetone but the lipolytic activity was also enhanced in the presence of DMSO and diethyl ether. In addition, the enzyme was catalytically active in toluene and n-hexane mixture, which is the preferred solvent in most of the transesterification reactions. One last example of lipases/esterases active in non-aqueous solvent systems is the lipase ReLipA from R. endophyticus (Yan et al., 2016) which exhibited excellent ability to catalyze the synthesis of methyl oleate, ethyl oleate, and butyl oleate in isooctane solvent system with a maximum yield of 82.2%. In addition, the enzyme is stable in different organic solvents.

Since an important number of cold-active enzymes are isolated from marine environments, some of them are also halophiles. One example is a salt-tolerant esterase, Est12 from Psychrobacter celer, which catalyze reactions and degrade organic matters under high salt concentrations (Wu et al., 2013a,b). Est12 was isolated from deep-sea sediments and showed enhanced activity and stability in 4.5 M NaCl, with Km decreasing from 0.069 to 0.033 mM p-NB and kcat doubled to around 9.21 s⁻¹ compared to the enzyme without salt. Moreover 0.5 and 1% (v/v) non-ionic detergents (Tween 20, Tween 80, Triton 100 and CHAPS) significantly enhanced the activity, in some cases up to 200%. After incubation with 5-30% (v/v) ethanediol, methanol, DMSO as well as 5-20% isopropanol and ethanol for 1 h, Est12 retained more than half of its activity.

More in depth, based on the remarkable activity of Pul-SH3 in the presence of SDS, two commercial detergents, Rika (7.5% v/v) and Fadisheh (2.5% w/v), were used to assess the potential application of the enzyme for washing purposes. The results showed that the enzyme was highly active in the presence of these detergents by 80.4 and 93.7%, respectively. In addition, the stability of the enzyme against the commercial detergents was interestingly high, so that the remaining activity after a 10-day holding at room temperature with Rika (7.5% v/v) and Fadisheh (2.5% w/v) was about 54.5 and 85%, respectively (Rajaei et al., 2015).

From the cold-active enzymes reviewed in Table 1, most of them claim potential uses in industrial processes. However, only a few went a step further and made at least a small trial in a real application, commonly for the food industry. These examples include three β-galactosidases for the hydrolysis of lactose in milk, one methylesterase for fruit firming, one polygalacturonase for juice industry, and a glycogen branching enzyme with biotechnological potential in bread production.

The first example was previously mentioned in this review and reported by Dong and coworkers (Dong et al., 2014) and corresponds to a cold-active but thermostable β-galactosidase. The enzyme was expressed as both a soluble protein and in the form of inclusion bodies. The active inclusion bodies of β-galactosidase were easily isolated by nonionic detergent treatment and directly used for lactose conversion in a repetitive batch mode. The enzyme lost ~5% (90°C) or 1% (10°C) activity after each reaction cycle. More than 54% (90°C) or 88% (10°C) of the original enzyme activity was retained after 10 conversion cycles under optimum conditions. These results suggest that the recombinant thermostable β-galactosidase may be suitable for the hydrolysis of lactose in milk processing, with the advantages of being active at low temperatures and cost-convenient. The second example is the attractive activity of Lactococcus lactis β-galactosidase at low temperatures, for which its efficiency as biocatalyst to bioconvert lactose within milk during storage was explored. For this purpose, they performed lactose hydrolysis in milk at 4 and 10°C. Using the free enzyme or immobilized cells, bioconversion rates of nearly 98% were achieved after 7 and 6 h of incubation, respectively. The immobilized cells were recycled and used several times, followed by enzyme activity measurements. Using immobilized E. coli NovaBlue cells expressing the β-galactosidase, more of the 96% of the initial activity was retained after 10 cycles of use at 4°C (Vincent et al., 2013). One last example of a β-galactosidase with potential applications in milk and dairy product industry is the enzyme from Antarctic Arthrobacter sp. 32cB, which has the capacity to hydrolyzed 90% of the lactose in 1 mL of milk at 10°C in 24 h (Pawlak-Szukalska et al., 2014).

Another example of enzymes with potential applications in food industry is an acidic and cold-active pectin methylesterase PE8F46 that was identified from P. chrysogenum and successfully expressed in P. pastoris (Pan et al., 2014). This enzyme was shown to significantly improve the firmness of pineapple dices in combination with calcium lactate, compared with a commercial pectinase complex. Thus, it represents an excellent candidate for food processing in the fruit and vegetable industry, considering the requirement of low-temperature to keep fruit quality.

An example concerning the juice industry is polygalacturonase, Endo-PG I, which was shown to reduce the viscosity of papaya juice by 17.6%, and increased its transmittance by 59.1% (Tu et al., 2013). When combined with a commercial pectin methylesterase, it showed higher efficiency with a synergy degree of more than 1.25. Currently, the widely used polygalacturonase has a pH optimum of 3.5, which is lower than the papaya juice pH (5.7). Endo-PG I have a slightly acid pH optimum (6.0), is cold active and stable in a large range of temperatures, properties required for potential applications in the juice industry.

The last example of application in food industry is a glycogen branching enzyme (RmGBE) from the thermophilic fungus Rhizomucor miehei that showed interesting cold-adapted characteristics (Wu et al., 2014). Addition of RmGBE to wheat bread resulted in a 26% increase in specific volume and a 38% decrease in crumb firmness in comparison with the control. Besides, the retrogradation, determined by measuring the crumb firmness and chewiness of bread, was significantly retarded along with the enzyme reaction. These properties make RmGBE highly useful in the food and starch industries.

Two nice examples of cold-active enzymes with potential uses in biomedicine were also reported for a α-galactosidase and a nitroreductase. The possibility to generate a universal blood type from B-type blood for application in transfusion therapy has been studied using enzymes. Some α-galactosidases are capable of removing the antigenic component from surface carbohydrates of group B red blood cells. One example is the cold-active α-galactosidase from Pseudoalteromonas sp. strain KMM 701 that showed to convert B red blood cells into blood type O cells at neutral pH (Balabanova et al., 2010). The activity of the enzyme was first observed when it was purified from its natural host producer. In view of its application, this cold-active enzyme was then overproduced in a heterologous host (Bakunina et al., 2014). An example of a potential cold-active enzyme for prodrug therapy was described using a cold-active nitroreductase, Ssap-NtrB (Çelik and Yetis, 2012). Despite Ssap-NtrB derived from a mesophilic bacterium, it showed optimal activity at 20°C against cancer prodrugs. Authors comment that the cold-activity of this novel enzyme will be useful for therapies in combination with crymotherapy, exposing the target tissue to low temperatures in order to trigger the enzyme activity to activate the drug only where is required. Moreover, the enzyme could also be used for bioremediation of compounds of explosive and volatile nature in regions where high activity at low temperatures is needed.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.