Stabilization of Multimeric Proteins via Intersubunit Cyclization

ABSTRACT

Proteins with high catalytic efficiency and selectivity under mild conditions have long been appreciated by industrial and medicinal fields. These proteins, which are commonly multimeric, often possess low stability, impeding wider application. Currently, strategies to improve the stability of multimeric proteins concentrate on enhancing the interaction at internal interface of the subunits. In this report, we confirmed that the largely underestimated subunit terminal ends are as significant as the internal interface for protein stability. By connecting both the terminal ends and internal interface of subunits, the tetrameric Leifsonia alcohol dehydrogenase (LnADH) protein can been cyclized into a rigid form with significantly improved thermostability and resilience. The improvement in the temperature at which enzyme activity is reduced to 50% after a 15-min heat treatment (T₅₀¹⁵) and melting temperature (T_m) of the modified protein was 18°C and 23.3°C, respectively, which is superior to the results achieved by normal protein engineering. Our study provided a novel strategy to effectively improve the stability of multimeric proteins, which is suitable not only for the short-chain dehydrogenase/reductase (SDR) family but also other classes of proteins with close terminal ends.

IMPORTANCE Industrially interesting proteins are generally multimeric proteins; however, their applications are often restricted due to low stability caused by the natural tendency of subunit disassociation. Current approaches targeting this problem mainly focus on enhancing the internal interfaces of the subunits to avoid their disassociation. In this study, we identified and confirmed the external interface to be significant for improving the stability of multimeric proteins. By connecting the terminal ends and internal interface with disulfide bonds, we found that the multimeric protein LnADH cyclized into a robust monomeric-like form, resulting in superior thermostability compared to traditional protein engineering. This intersubunit cyclization approach is efficient and easy to perform, providing a novel method for engineering many important classes of multimeric proteins.

INTRODUCTION

Enzymes (catalytic proteins) are selective and display high activity under mild experimental conditions. The use of enzymes as catalysts or therapeutic reagents has been long appreciated by industrial and medicinal fields due to their low environmental impact, economic viability, sustainability, and high selectivity (1, 2). In the synthetic biology field, proteins encoded by their corresponding genes are widely used in heterologous systems to afford specific functions. However, many proteins have drawbacks that make their implementation difficult, most notably their naturally low stability (3). Thus, improvement of protein robustness and stability is often a prerequisite for their application (4).

Protein stability is significantly affected by internal structural flexibility and unfavorable surface electrostatic interactions. Enhanced stability can be achieved by rigidifying the backbone or optimizing the charge distribution via protein engineering (4–6). The stability of monomeric proteins can be improved by introducing beneficial mutations in flexible sites and surface electrostatic sites (4–8) or by covalently circularizing the backbone N- and C-terminal ends via a disulfide bond or other connection (3–5, 9, 10). In multimeric proteins, the first step of protein inactivation is the dissociation of subunits; thus, prevention of their disassociation is the priority for stabilization of these complex enzymes (3). The main approach for this purpose concentrates on the introduction of stabilizing mutations to form or enhance interactions, such as disulfide bonds, hydrophobic interactions, and salt bridges, at the internal interface of protein subunits (3, 11, 12). Multimeric proteins are often more thermolabile than monomeric proteins due to the natural tendency toward subunit disassociation; however, their valuable and diverse functions, such as oxidoreduction and transamination, make them attractive for industrial and synthetic biology applications (3). The development of new effective approaches to improve their fitness and stability is of great practical interest.

The terminal ends of proteins are usually the most flexible parts of the backbone and easily become targets of proteolytic enzymes (10). In monomeric proteins, connection of terminal ends has been demonstrated to dramatically improve protein stability and resilience (10). Interestingly, we observed that in a number of important multimeric protein families, such as short-chain dehydrogenase/reductases (SDRs), subunits are assembled in an ordered way in a head-to-head or head-to-tail fashion, with the terminal ends positioned within a reachable range by covalent connection (13). We hypothesized that circularization of protein subunits via covalent connection at both the internal interface and terminal ends can transform these multimeric proteins into monomeric-like proteins whose subunit associations and backbone rigidity would be extensively improved. We investigated this intersubunit cyclization (ISC) stabilization strategy of multimeric proteins using Leifsonia alcohol dehydrogenase (LnADH) as a model. Our results demonstrate the effectiveness of this concept in improving protein thermostability.

RESULTS

Analysis and modeling of the LnADH structure.

SDRs are a large family of homomultimeric enzymes whose members have been widely used in industry and metabolite biosynthesis (13), which include alcohol dehydrogenases (ADHs), glucose dehydrogenases (GDHs), halohydrin dehalogenases (HHDHs), and ketoreductases (KRs). These proteins are mostly tetrameric and have two subunit interfaces (Fig. 1A; see also Fig. S1 in the supplemental material): interface A has buried C-terminal ends and outward N-terminal ends, and interface B is composed of two helixes in the middle. The overall arrangement of subunits forms a mirror image at interface B with two pairs of head-to-head dimers. The ADH from Leifsonia sp. strain S749 (LsADH) is one of very few dehydrogenases that is able to use NADH as a cofactor rather than the more expensive NADPH. More so, it can generate chiral alcohols from various ketones, with very high conversion yield and enantioselectivity (14, 15). These features make it advantageous over many other NADPH-dependent ADHs and have promise for industrial applications. Due to this potential and the availability of Leifsonia aquatica ATCC 14665, we chose the highly similar protein LnADH from this strain, which differs by three residues from LsADH, as a study model throughout this work (see Materials and Methods). The protein structure of LnADH is not currently available, but there are a few structures of ADHs that have been solved (16–18). These structures are identical to each other and bear the typical homotetrameric fold of SDRs. By referring to the structure with PDB identification (ID) 4URE (49% identity), (16), we deduced the structure of LnADH by molecular modeling (Fig. 1A). It was revealed that the C-terminal ends of interface A have many interactions, and the N-terminal ends in each pair are directed toward each other. In order to connect the four subunits, we introduced disulfide bonds at the N-terminal ends of interface A and the subunit interaction regions of interface B, converting the tetramer into a large monomer via intersubunit disulfide linkages.

FIG 1.

Improvement of the LnADH thermostability via intersubunit cyclization. (A) The modeled structure and cartoon illustration of the cyclized LnADH. Disulfide bond connections at terminal ends and internal interfaces are indicated by blue and dark red, respectively. (B) Standard SDS-PAGE (left) and nonreducing SDS-PAGE (right) analysis of LnADH (lane 1), LnADH-C_T (lane 2), LnADH-C_I (lane 3), and LnADH-C_TI (lane 4). Lane M, molecular weight marker (in kilodaltons). (C) Relative activity and residual activity (treated at 50°C for 15 min) of LnADH and mutant proteins connected in the N-terminal ends. (D) Relative activity and residual activity (treated at 42°C for 30 min) of LnADH and the mutant proteins connected in the internal interface. (E) MALDI-TOF mass spectrometry analysis of LnADH (measured [meas.], 27,088.6901; calculated [calcd.], 27,071.7734), LnADH-C_T (meas., 54,381.7068; calcd., 54,347.5507), LnADH-C_I (meas., 54,185.6197; calcd. 54,149.4022), and LnADH-C_TI (meas., 108,783.6098; calcd., 108,712.859). (F) The residual activity of LnADH, LnADH-C_T, LnADH-C_I, and LnADH-C_TI after treatment at 45°C (black), 50°C (blue), and 55°C (red) for 15 min.

Terminal ends connection to improve LnADH thermostability.

To facilitate protein expression and purification, LnADH was cloned into the NdeI and HindIII sites of pET28a, which expresses 20 additional residues, including a 6×His tag at the N-terminal end of LnADH. Right after the NdeI site, codons for cysteine and the peptide linker Gly-Gly-Ser-Gly were introduced for insertion of a CGGSG sequence in the N terminus of LnADH. The aim of introducing this peptide linker was to overcome the distance between the terminal ends. The expression level of the recombinant protein LnADH-C_TLinker in Escherichia coli is similar to that of the wild-type protein, suggesting that the cysteine in the terminal regions did not interfere with the protein folding. After purification, LnADH-C_TLinker was incubated at 4°C overnight to allow for the formation of the disulfide bonds by oxidation of the cysteine residues with air. The formation of the disulfide bonds at the terminal ends was confirmed by nonreducing SDS-PAGE analysis with and without treatment with dithiothreitol (DTT). LnADH-C_TLinker shows a 2-fold increase in molecular weight (54 kDa) compared to the wild-type protein (27 kDa) (Fig. S2B, lane 6) which can be reversed by treatment with DTT (Fig. S2D, lane 1). Activity measurements after treatment at 45°C for 15 min revealed increased thermostability of LnADH-C_TLinker compared to the wild-type protein, resulting in 89.2% and 13% activity toward the substrate ethyl 4-chloroacetocaetate (COBE), respectively (Fig. S3). When the temperature was elevated to 50°C, only marginal activity (<1%) was retained by LnADH-C_TLinker (Fig. S3), suggesting that the backbone rigidity of this protein was not strong enough to resist dissociation of the subunits.

To reduce the flexibility of the connected backbone, the peptide linker GGSG was removed from LnADH-C_TLinker, giving the protein construct LnADH-C_T (disulfide bonds connected in the terminal ends of interface A). SDS-PAGE and matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry analysis confirmed complete subunit connection in this protein (Fig. 1B, lane 2 and 1E), suggesting that the N-terminal ends are indeed highly flexible and are able to reach each other. This protein showed an obvious improvement upon treatment at 50°C (Fig. 1C and F), with an improvement of 5.6°C and 5.5°C for the melting temperature (T_m) and temperature at which enzyme activity is reduced to 50% after a 15-min heat treatment (T₅₀¹⁵), respectively (Fig. 2A and B) without an obvious reduction in catalytic activity (Fig. 1C and 3B). These results revealed that the loop length formed by the connected terminal ends acts as a pivot that increases protein stability. In order to test the possibility for further reduced flexibility, we trimmed the backbone length by individually deleting one to three residues (Ala, Ala-Gln and Ala-Gln-Tyr), which are located right after the introduced cysteine in the LnADH-C_T. The resulting LnADH-C_T-dA, LnADH-C_T-dAQ, and LnADH-C_T-dAQY mutants, respectively, were expressed, and their activities under heat treatment were assayed. All three proteins show reduced ability to resist heat denaturing compared with LnADH-C_T (Fig. 1C and S3), especially the LnADH-C_T-dAQY mutant, which lost almost all its acquired capability. We hypothesized that the LnADH-C_T-dAQY mutant may fail to form the disulfide connection due to the short length of its terminal ends. Based on SDS-PAGE analysis, it still mostly (69%) exists as a connected form (Fig. S2B, lane 5). These clues revealed that the reduced thermostability is not from the nonformation of the disulfide connection, but more likely the improper length of the connected loop. This was further supported by the homology model of the LnADH-C_T-dAQY mutant, which suggested an obvious alteration of the whole peptide backbone compared with the N-terminal loop changes of other mutants (Fig. S4). The orientation of the Cβ atoms is also able to influence the thermostability. The LnADH-C_T-dAQ mutant, which has parallel cysteine residues, showed an obvious elevated robustness compared to the LnADH-C_T-dA mutant (Fig. 1C and S3), which has antiparallel cysteines. Altogether, these results revealed that flexibility of the terminal ends in multimeric proteins is indeed significant for protein stability and that their connection is highly effective at improving protein robustness.

FIG 2.

The T_m, T₅₀¹⁵ value, and activity recovery of the cyclized LnADH mutants. (A) Evaluation of the T_m value of LnADH (44.4°C), LnADH-C_T (50°C), LnADH-C_I (56.3°C), and LnADH-C_TI (67.7°C). (B) Evaluation of the T₅₀¹⁵ value of LnADH (43°C), LnADH-C_T (48.5°C), LnADH-C_I (49°C), and LnADH-C_TI (61°C). (C) Recovery of activity of LnADH-C_TI. Residual activity was measured at an interval of 1 to 5 h after LnADH-C_TI was treated at 50°C (black), 60°C (red), and 70°C (blue) for 15 min.

FIG 3.

Kinetic analysis of LnADH and its mutants. (A) LnADH. (B) LnADH-C_T. (C) LnADH-C_I. (D) LnADH-C_TI.

Internal interface connection to enhance LnADH thermostability.

We next aimed to introduce disulfide connections into the internal interface. Based on the online program Design Site-Directed Mutants (19), a total of nine pairs of residues were found to be potentially suitable for mutation into cysteines to form the disulfide bonds (Table S1). By careful curation, Q169C and A250C, E221C and F236C, and A176C and A212C mutations were excluded due to their location at interface A. The six V101C A171C, T159C A163C, T159C L167C, H162C, S155C N170C and T100C E174C mutants (Fig. S5) were constructed and their thermostability measured (Fig. 1D). Interestingly, all of these mutants showed an improvement in stability, but most showed decreased catalytic activity (Fig. 1D). In LnADH-C_I, disulfide bonds connecting S155C and N170C around interface B (Fig. S5E) showed a remarkable improvement in stability (T₅₀¹⁵ and T_m increased by 6°C and 11.9°C, respectively) and retained catalytic activity (Fig. 1D and F, 2, and 3C). The formation of disulfide bonds was confirmed by mass spectrometry analysis (Fig. 1E). SDS-PAGE analysis revealed that this protein has 40% formation of dimer (Fig. 1B, lane 3). The moderate proportion of covalent connection is probably due to the steric hindrance of the internal interface, and this might be improved by increasing the incubation time or oxygen concentration.

Combining terminal ends and internal interface connection to improve LnADH thermostability.

Since the addition of the terminal and internal interface connections showed significant increases in thermostability, we were next interested in combining these connections. By introducing S155C and N170C into the LnADH-C_T mutant, the LnADH-C_TI double mutant (disulfide connected in both of the N-terminal ends and S155C N170C) was constructed. The expression level of this mutant in E. coli is similar to those of the other mutants and the wild-type protein. After incubation at 4°C overnight, the formation of intersubunit connections was measured. Given that LnADH-C_TI should form a circular tetramer, it will show a different migration rate from those of the linear dimers of LnADH-C_T and LnADH-C_I. In the SDS-PAGE, the LnADH-C_TI mutant shows a clearly different pattern from the LnADH-C_I mutant and wild-type proteins (Fig. 1B, lane 4), which was in agreement with our expectation. Although the migration pattern of the LnADH-C_TI mutant is similar to that of the LnADH-C_T mutant (Fig. 1B, lane 2), we assumed that it was the combined results of the molecular weight and protein conformation. This assumption was confirmed by mass spectrometry analysis (Fig. 1E) and stability assessment (Fig. 1F). The backbone-cyclized mutant showed a dramatic improvement in thermostability (Fig. 1F) without an obvious decrease in catalytic activity (Fig. 3D) and acquired an increase in T₅₀¹⁵ and T_m of 18°C and 23.3°C, respectively, compared to the wild type (Fig. 2A and B). Moreover, the LnADH-C_TI mutant is able to recover its catalytic activity after heating treatment. About 25% to 33% recovery of activity was observed after incubation at room temperature for 8 h following heat treatment at 50°C to 70°C (Fig. 2C). This refolding effect is probably driven by the potential energy of protein conformation which was stored by the cyclized protein backbone. The merit of increased thermal resilience makes this multimeric protein even more robust and suitable for practical application. Taken together, our results demonstrated that this intersubunit cyclization strategy is highly effective at improving the thermostability of multimeric proteins.

DISCUSSION

The inactivation of monomeric proteins is usually initiated by the disassembling of the flexible parts of the backbone. Terminal ends are often the most flexible and liable parts of a protein; thus, connecting these has been widely utilized to engineer monomeric proteins (9, 10). Unlike monomeric proteins, the inactivation of multimeric proteins starts from the dissociation of subunits (3). Hence, reinforcing the association of subunits via the introduction of stabilizing mutations into the internal interface is currently the major approach for improving stability (3, 11, 12). The different mechanisms in protein inactivation suggest that the importance of terminal ends in multimeric protein stability is largely underestimated. By careful analysis, we noticed that in many multimeric proteins, such as SDRs (13) and polyketide synthases (PKSs) (20), terminal ends are also located at the interface of subunits, albeit outside. By connecting the terminal ends in LnADH, we confirmed that they are as significant as the internal interfaces of subunits for improving protein stability. Unlike the introduction of mutations in the internal interface, which requires a precise analysis of the intersubunit interactions and screening of a large number of mutants, connection at terminal ends is much more straightforward. Protein terminal ends are flexible and could thus be connected at a diverse range of distances. By adjusting the length of the connection via the insertion or deletion of residues in the protein backbone, it is simple to achieve the most effective mutant. In addition, unlike the engineering of the internal interface, which resulted in only one of six mutants retaining activity, terminal end connection did not cause an impairment in catalytic activity (Fig. 1C and D). Thus, for some instances, terminal end connection could be an ideal choice for improving the stability of multimeric proteins.

By further combining this technique with the traditional internal interface connection, we confirmed that multimeric proteins can be converted into an even more robust form. The value of improvement (ΔT₅₀¹⁵ = 18°C, ΔT_m = 23.3°C) in the monomeric-like LnADH-C_TI protein is prominent and far better than the improvements seen in LnADH-C_T (ΔT₅₀¹⁵ = 5.5°C, ΔT_m = 4.9°C) and LnADH-C_I (ΔT₅₀¹⁵ = 6°C, ΔT_m = 11.9°C). This indicates that a synergistic effect on the stability could be achieved by cyclization of the entire scaffold, which probably results in the strengthening of both the subunit associations and backbone rigidity. It is worth noting that the improved T₅₀¹⁵ value achieved by this strategy (18°C) is superior to the value (2 to 15°C) that can be achieved via traditional protein engineering (21). These factors together confirm that our approach is highly effective at improving the stability of multimeric proteins.

The family of SDR proteins, which includes ADHs, KRs, GDHs and HHDHs, is very useful in industry and synthetic biology engineering (22, 23). In the pharmaceutical industry, hydroxyl groups are considered one of the most important, and approximately 30% of the reported industrial biotransformations employ ADHs and KRs to produce chiral alcohols (1, 23–25). In synthetic biology, KRs and ADHs are pivotal machineries for many natural products and biofuel biosynthesis (23). The overall structure of these SDRs is identical to that of LnADH, and it is feasible to apply our results to these important members for stability improvement. Besides the SDR family, a large number of other multimeric protein families of industrial interest, such as leucine dehydrogenases (26) and modular PKSs (20), also have an ordered organization of their subunits. Both their terminal ends and internal interfaces are positioned within a reachable range by covalent connection and thus are potentially suitable for engineering with this approach.

Disulfide bonds are widely used in protein engineering. The cytoplasm of the expression host Escherichia coli has strong glutathione reductase (Gor) and thioredoxin reductase (TrxB) which can impede disulfide bond connection, (27); thus, formation of the disulfide bond is often proceeded by in vitro air oxidation. Unlike biocatalysis, which mostly uses cell lysates to perform a reaction, synthetic biology requires their function in vivo. In this case, reductase-deleted strains, such as E. coli Origami, or coexpression with disulfide oxidoreductases DsbA and DsbC can overcome this problem (27). More so, other useful hosts, such as yeast and actinomycetes, can form disulfide bonds in vivo (28); thus, the application of our strategy has no barrier to in vivo systems. Moreover, in terms of the vulnerability of disulfide bonds in the reducing environment and elevated temperature (>75°C), more resilient connections, such as amide bonds, can be applied to cyclize the subunits; methods, such as native chemical ligation (29), cyclase (30–32), intein (33), transpeptidase (34), and assembly tags (35), are able to introduce this connection.

In summary, we have developed a novel intersubunit cyclization strategy for stabilization of multimeric proteins. With LnADH as a model, we confirmed that the connection of the N-terminal ends is as effective as the internal interface for stability improvement. By combining the terminal and internal interface connections, LnADH can be converted into a monomeric-like protein whose thermostability and resilience were significantly improved. Our results not only achieved a highly robust alcohol dehydrogenase but also provide an effective method for engineering members of the SDR family and other classes of multimeric proteins with close terminal ends.

MATERIALS AND METHODS

Strains, plasmids, and general methods.

Escherichia coli was cultivated and manipulated according to standard methods (36). The primers used in this study are listed in Table 1. The strains and plasmids used in this study are listed in Table 2. DNA isolation and manipulation in E. coli and Leifsonia aquatica ATCC 14665 were performed according to standard methods (36). Primer synthesis and DNA sequencing were performed at Genewiz Biotech Co., Ltd. (China). Restriction enzymes and DNA polymerases (Taq and PrimeSTAR) were purchased from TaKaRa Biotechnology Co., Ltd. (China). All chemicals and reagents were purchased from Santa Cruz Biotechnology, Inc. (USA) or Shanghai Sangon Biotech (China) Co., Ltd., unless noted otherwise.

TABLE 1.

Primers used in this study

^a Restriction sites are underlined.

TABLE 2.

Bacterial strains and plasmids

Strain or plasmid	Description	Source
Strains
E. coli DH5α	Host for general cloning	Invitrogen
E. coli BL21(DE3)	Host for protein expression	Stratagene
Leifsonia aquatica ATCC 14665	Used for amplification of LnADH	ATCC
Plasmids
pET28a	Protein expression vector in E. coli	Novagen
pWHU2449	pET28a derivative for LnADH expression	This study
pWHU2450	pWHU2449 derivative with an insertion of a CGGSG right after NdeI in the N terminus of LnADH	This study
pWHU2451	pWHU2449 derivative with an insertion of cysteine right after NdeI in the N terminus of LnADH	This study
pWHU2452	pWHU2451 derivative with a deletion of alanine right after the cysteine in the N terminus of LnADH	This study
pWHU2453	pWHU2451 derivative with a deletion of AQ right after the cysteine in the N terminus of LnADH	This study
pWHU2454	pWHU2451 derivative with a deletion of AQY right after the cysteine in the N terminus of LnADH	This study
pWHU2455	pWHU2449 derivative with V101C and A171C	This study
pWHU2456	pWHU2449 derivative with T159C and A163C	This study
pWHU2457	pWHU2449 derivative with T159C and L167C	This study
pWHU2458	pWHU2449 derivative with H162C	This study
pWHU2459	pWHU2449 derivative with S155C and N170C	This study
pWHU2460	pWHU2449 derivative with T100C and E174C	This study
pWHU2461	pWHU2459 derivative with an insertion of cysteine right after NdeI in the N terminus of LnADH	This study

Structure modeling and selection of residues for mutations.

All of the modeling work was performed by using Discovery Studio 4.0 (Biovia). The initial three-dimensional (3D) structural model of LnADH was constructed based on the crystal structure of 1-(4-hydroxyphenyl)-ethanol dehydrogenase from Aromatoleum aromaticum Ebn1 (PDB code 4URE) (16) by using the Build Homology Models module and further optimized by molecular dynamics. Then, the resultant model was uploaded to the online program DSDBASE (http://caps.ncbs.res.in/dsdbase//mainFrame.html) (19) to predict potential sites for mutation into the cysteine residues. The candidate sites for possible disulfide bond connection are shown in Table S3. The homology models of these candidates and terminus-connected mutants were refined by the Disulfide Bridges module of Discovery Studio 4.0. After that, energy minimizations of these models were performed using the Minimization module. The obtained structural models of the mutants and the wild type were compared using the Align Structures module to further exclude the mutants with obvious backbone shift.

Cloning, overexpression, and purification of recombinant proteins.

L. aquatica ATCC 14665 was cultivated in LB medium, and its genomic DNA was extracted according to the standard procedure with E. coli (36). LnADH (accession no. ERK72999.1) is different in only three residues from LsADH (G78A, T111K, and D202E). It was amplified by the LnADH-for and LnADH-rev primer pair from genomic DNA of L. aquatica and cloned into the NdeI and HindIII sites of pET28a to generate expression plasmid pWHU2449 (LnADH). Overlap PCR was used to obtain the rest of the mutants with the primers (Table 1), which were then transformed into E. coli BL21(DE3) for overexpression of N-terminal 6×His-tagged fusion proteins. The purified, desalted, and concentrated protein was stored at −80°C and used for enzymatic assays. Protein concentration was determined by the Bradford method (37).

In vitro enzymatic assays and kinetic analysis.

The activities of enzymes were assayed at 30°C by monitoring the decrease in the absorbance of NADH at 340 nm. The reaction mixture contained 2 mM ethyl 4-chloroacetocaetate (COBE), 0.1 mM NADH, 100 mM potassium phosphate buffer (KH₂PO₄-K₂HPO₄ [pH 7.0]), and an appropriate amount of enzyme in a total volume of 1 ml. One unit of enzyme activity was defined as the amount of enzyme that catalyzes the oxidation of 1 μmol NADH per minute under these conditions. Kinetic analysis of the enzymes was performed in potassium phosphate buffer (100 mM KH₂PO₄-K₂HPO₄ [pH 7.0]) containing 0.2 mM NADH and different concentrations of COBE (0.1 to 2.5 mM) at 30°C. The resulting enzymatic activities were fitted to the Michaelis-Menten equation by OriginPro 9.0 (OriginLab software, Northampton, MA) to obtain estimates for k_cat and K_m.

Measurement of thermal denaturation by circular dichroism.

The melting temperature (T_m) is the temperature of the transition midpoint. For thermal scans, the protein samples (300 μl, 1 mg/ml, 100 mM [pH 7.5] HEPES buffer) were heated from 20 to 95°C, with a heating rate of 1.50°C/min controlled by a Chirascan circular dichroism. The dichroic activity at 222 nm was continuously monitored every 1°C. The resulting data were fitted to the logistic equation by OriginPro 9.0 (OriginLab software, Northampton, MA) to obtain estimates for T_m.

Thermostability measurement, T₅₀¹⁵ values, and recovery of catalytic activity.

Thermostability was assessed by measuring the residual activity subsequent to the exposure to certain temperatures. In the literature, the T₅₀¹⁵ value (8) is often used to quantitatively characterize thermostability. It is the temperature required to reduce the initial enzymatic activity by 50% within 15 min. Thus, the T₅₀¹⁵ values of the purified enzymes were measured and displayed by the residual activity curves. We exposed LnADH and LnADH-C_TI to high temperatures (50°C, 60°C, and 70°C) for 15 min and measured the residual activity at 30°C at 1 to 5 h.