Heterologous expression, characterization and evolution prediction of a diaphorase from Geobacillus sp. Y4.1MC1

β-hydroxybutyric acid is the most sensitive indicator in ketoacidosis detection, and accounts for nearly 78% of the ketone bodies. Diaphorase is commonly used to detect the β-hydroxybutyric acid in clinical diagnosis. However, the extraction of diaphorase from animal myocardium is complex and low-yield, which is not convenient for large-scale production. In this study, a diaphorase from Geobacillus sp. Y4.1MC1 was efficiently heterologous expressed and purified in E. coli with a yield of 110 mg/L culture. The optimal temperature and pH of this recombinant diaphorase (rDIA) were 55 °C and 6.5, respectively. It was proved that rDIA was a dual acid- and thermo-stable enzyme, and which showed much more accurate detection of β-hydroxybutyric acid than the commercial enzyme. Additionally, we also investigated the molecular interaction of rDIA with the substrate, and the conformation transition in different pH values by using homology modeling and molecular dynamics simulation. The results showed that 141–161 domain of rDIA played important role in the structure changes and conformations transmission at different pH values. Moreover, it was predicted that F105W, F105R, and M186R mutants were able to improve the binding affinity of rDIA, and A2Y, P35F, Q36D, N210L, F211Y mutants were benefit for the stability of rDIA.


Introduction
Ketone bodies are a class of metabolites that produced by fatty acid, including β-hydroxybutyric acid (~ 78%), acetoacetic acid (~ 20%) and acetone (~ 2%), and excess ketone bodies will exceed the utilization ability of liver (Yue et al. 2014). The imbalance between ketone bodies and liver utilization plays a key role in the occurrence of ketoacidosis, such as acetonemia and acetonuria. In addition, type II diabetes mellitus complicated with ketosis or ketoacidosis also often occurs in clinical for the less insulin secretion in the body (Jerreat and Lynne 2009). β-hydroxybutyric acid is the sensitive indicator for monitoring the ketoacidosis due to the rich distribution in the blood (Goldstein et al. 1995). Therefore, it is with great signi cance to e ciently and accurately detect the β-hydroxybutyric acid in the early detection and prevention ketoacidosis and type II diabetes (Sheikh-Ali et al. 2008).
Traditional method for the β-hydroxybutyric acid detection was to test acetone, the oxidation product of βhydroxybutyric acid, by titration, speci c gravity and colorimetry (Deng et al. 2004;Kalapos 2003). But it is a semi quantitative method, and the amount of acetone is affected by the experimental conditions. The quantitative analysis of acetone by gas chromatography can only determine the concentration of total ketones in serum that lack of speci city (Deng et al. 2004). Nowadays, enzymatic method by diaphorase and β-hydroxybutyrate dehydrogenase has been developed (Daniel et al. 1992). Under the speci c oxidation of β-hydroxybutyric acid dehydrogenase and NAD + , β-hydroxybutyric acid produces acetoacetic acid and NADH. And in the presence of NADH and diaphorase, oxidized iodo nitro tetrazolium chloride blue (INT) can be catalyzed to reductive INT, which has absorbance at 505 nm (Huang et al. 2006;Cimen et al. 2010). There are lots of advantages of the enzymatic method, including high precision, high speed, low cost, wide linear range, and can be directly determined without pre-puri cation or chemical treatment (Parry et al. 2010).
Diaphorases are a class of avin-containing enzymes, which have been widely used in biosensor design, biotransformation and in vitro diagnostic tests (Kianmehr et al. 2018). Various diaphorases have been identi ed from bacterial (Ahmed et al. 1989), plan (Eftink and Bystrom 1986), and animal sources (Radi et al. 1993). Due to the di culties on high amounts puri cation of diaphorase from pig heart muscle, heterologous expression has become the main method to produce diaphorase. Some studies have used genetic engineering methods and subsequent expression and puri cation steps to obtain various types of recombinant diaphorase. For example, a diaphorase with molecular weight of 24 kDa was successfully expressed, puri ed and characterized from Clostridium kluyveri, and applied in the reduction of diphosphopyridine nucleotide and triphosphopyridine nucleotide Madan et al. 2010). Both ferrocenium compounds and hexacyanoferrate (III) of this kind diaphorase could be used as electron transfer mediators and co-immobilization with several dehydrogenases in a membrane electrode format yielded sensors (Antiochia et al. 2015). Lipoamide dehydrogenase (DLD) with diaphorase activity, was successfully heterologous expressed and puri ed from E. coli, and the suitable kinetic features of DLD could make this biocatalyst useful for application as a diagnostic enzyme (Kianmehr et al. 2017). However, there are still some disadvantages such as low yield (Klyachko et al. 2005), low activity ) and poor stability (Kianmehr et al. 2017).
In this study, a novel recombinant diaphorase (rDIA) from Geobacillus sp. Y4.1MC1 was e ciently heterologous expressed and characterized in E. coli BL21(DE3). The enzymatic properties including optimum pH and temperature, thermo-and acid-stability were studied. And we also probed the molecular interaction of rDIA with β-hydroxybutyric acid, and investigated the mechanism of pH stability related with the rDIA structure by using homology modeling and molecular dynamics (MD) simulation. Finally, mutation evolution on pH stability was also performed by the interaction energy analyzation.

Construction of the expression vector and engineered strains
The gene sequence of DIA (GenBank: CP002293.1) from Geobacillus sp. Y4.1MC1 was optimized according to the codon preference of E. coli, and the rDIA gene ( Fig SI 1) was cloned into the pUC57 vector (pUC57-rDIA). Then the rDIA was subcloned into the pET28a expression vector by NdeI and XhoI restriction enzymes under the T7 promoter. After identi ed by gene sequencing, the plasmid was named pET28a-rDIA. Finally, the vector pET28a-rDIA was transformed into the expression host E. coli BL21(DE3) by chemical competence method to heterologous expression of rDIA. After identi ed by colony PCR, the engineered strain was named BL21-rDIA.
Expression and puri cation of rDIA Single colony of engineered strain was picked and cultured in 20 mL LB medium at 37°C overnight to prepare the seeds, and transferred to 1000 mL fresh TB medium with 5% inoculation. After OD 600 of the inoculated broth reached to 0.6, the BL21-rDIA culture was incubated at 16°C for 18-22 h.
After centrifugation of BL21-rDIA culture at 8000 rpm for 20 min, the supernatant was removed and cell pellets were collected., The bacterial pellets were resuspended in the 0.1 M potassium phosphate (pH 7.5, 0.05 mol/L K 2 HPO 4 , 0.05 mol/L KH 2 PO 4 , 5 mg/L FAD) and sonicated by 9 pulses (30 sec sonication with the same intervals at 30% amplitude) (Damough et al. 2021), and the mixture was centrifuged at 12,000 rpm at 4°C for 40 min. The supernatant was loaded on the Ni-NTA column, which has been equilibrated using the binding buffer (0.1 M potassium phosphate, 5 mg/L FAD; pH 7.5) in the AKTA Pure chromatography system (GE Healthcare). After washed by the buffer I (0.1 M potassium phosphate, 5 mg/L FAD, 10 mM Imidazole; pH 7.5), The rDIA were eluted using the elution buffer (0.1 M potassium phosphate, 5 mg/L FAD, 500 mM Imidazole; pH 7.5) under 1 ml/min ow rate. The imidazole in the rDIA solution was removed by using PD-10 Desalting Columns (GE Healthcare). Nanodrop 2000 (Thermo Fisher Scienti c, Wilmington, USA) was used to measure protein concentration at 280 nm. At last, the product was lyophilized and stored at -80°C.

Enzyme activity measurement
One unit of rDIA activity is de ned as the amount of enzyme which oxidizes 1 µM of NADH to NAD + per minute at 37°C under the conditions speci ed in the assay procedure. Pipette accurately 1 mL of reaction mixture into a small test tube and preincubate it at 37°C. Reaction mixture: 0.5 mL KH 2 PO 4 -NaOH buffer After cooling on ice for 5 min, the relative residual activity was determined using the above method. For calculating its thermostability, the activity of un-treated enzyme was de ned as 100%.
Effect of pH on the activity and stability of rDIA In order to optimal pH, the enzymatic activity of the rDIA was measured in Acetate buffer (pH 5.0-6.0), Phosphate buffer (pH 6.0-8.5) and Glycine-NaOH buffer (pH 8.5-9.5) at 37°C. pH stability of rDIA was determined by assaying its catalytic activity after treatment of rDIA in Citrate buffer (pH 3.0-6.0), Phosphate buffer (pH 6.0-9.0) and Glycine-NaOH buffer (pH 9.0-12.0) at 4°C for 6 h, respectively. The maximum enzyme activity was set to 100%, and the relative enzyme activity was calculated. Each experiment was performed in triplicate to ensure the reproducibility.

Comparation of rDIA with commercial diaphorase
We compared the enzymatic property of the rDIA and commercial diaphorase (purchased from ToYoBo) by using the β-hydroxybutyric acid test kit (Abbexa, UK). In the presence of β-hydroxybutyric acid dehydrogenase and oxidized coenzyme I, β-hydroxybutyric acid is oxidized to acetoacetic acid and oxidized coenzyme I is reduced to reduced coenzyme I. And in the presence of diaphorase, reduced coenzyme I and INT (iodo nitro tetrazolium chloride blue) produce formazan. The absorbance of formazan was measured at 505 nm by Hitachi 7170A automatic biochemical analyzer (Hitachi, Japan), which was positively correlated with the enzyme activity.

Molecular dynamics (MD) simulations
We selected the crystal structure of Bacillus smithii FMN-dependent NADH-azoreductase (PDB entry 6JXS) as the template for homology modeling. MD simulations were performed using the program GROMACS 2018.4 with Amber99SB force eld parameters (Hornak et al. 2006). Firstly, the diaphorase was placed in a simulation box, lled with water molecules and equilibrium ions, at least 12Å distances from any atom of the enzyme. To remove possible steric stresses, system energy minimization was then performed for 50000 steps. The six polypyrrole-pyrethrin con gurations were energy minimized and subjected to 100 ps NVT (N for the number of atoms, V for volume, and T for temperature) equilibration at 300 K. Then, the six systems were respectively run for 50 ns of NPT (N for the number of atoms, P for pressure, and T for temperature) production. In this study, we carried out a series of all-atom MD simulations using the standard AMBER99SB force eld (Hornak et al. 2006;Gao et al. 2017) at pH 3.0 and 6.5. Under the two pH values of mimic environment, the side chains of the Asp55, Glu75 and Glu194 were modeled to take different charge states: neutral at pH 3 and negatively charged at pH 6.5 (Rostkowski et al. 2011). The same strategy is often applied in MD simulations to investigate the pHdependent behaviors of protein folding/aggregation (Khandogin and Brooks 2007;Li et al. 2013). A substrate receptor-binding model was proposed to identify key residues that bind to DIA. The simulation trajectories were analyzed using several auxiliary programs provided with the GROMACS 2018.4 package.
Secondary structure analyses were carried out using the dictionary secondary structure of proteins (DSSP) (Kabsch and Sander 1983) method and examined visually using VMD software (Humphrey et al. 1996).

Results And Discussion
Construction and identi cation of the expression vector According to the protein sequence of DIA (GenBank: CP002293.1) in NCBI, the gene sequence was codon optimized for the preference of E. coli. The optimized gene was synthesized, and cloned into the pUC57 vector. The rDIA fragment was double-digested using NdeI/XhoI and sub-cloned into the pET28a expression vector. Schematic of the construction of recombinant expression plasmid pET28a-rDIA was shown in Fig. S2A. The recombinant expression vector was identi ed by PCR using the primer pairs of T7-F (5'-TAATACGACTCACTATAGGG-3') and T7-R (5'-GCTAGTTATTGCTCAGCGG-3'). As shown in Fig. S2B, the target fragment of rDIA was shown to be approximately 1000 bp, which is almost identical to the theoretical value. We also identi ed the pET28a-rDIA plasmid by gene sequencing from GENEWIZ Inc.

Expression and puri cation of the rDIA
The ow chart of expression and puri cation process of rDIA was shown in Fig. 1A. As shown in Fig. 1B, the target protein band was found in both bacterial lysate of BL21-rDIA and supernatant in TB culture, which indicated that the rDIA was successfully soluble expressed. Figure 1B also indicated the molecular mass of recombinant rDIA consistent with deduced molecular mass of 29 kDa. Large scale protein expression was performed under the auto-induction fermentation medium. After puri ed by using HisTrap FF column, the protein samples were identi ed using SDS-PAGE (Fig. 1C). The rDIA was over 95% pure and with a yield of 110 mg/L culture. The molecule weight of rDIA was similar to the thermophilic alcohol dehydrogenase (htADH) (Liang et al. 2004), an intracellular diaphorase, which was isolated from a strain of Bacillus stearothemophilus with the molecular weight of 30 kDa.

Enzymatic characterization of rDIA
As the temperature have great in uence on the enzymatic reaction, the optimum temperature of puri ed rDIA was determined by measuring its activity at a series of temperatures of 20, 25, 30, 35, 37, 40, 45, 50, 55, 60, 65, 70, 75, and 80°C at pH 6.5. As shown in the Fig. 2A, rDIA displayed maximum activity when the temperature at 55°C, and maintained good catalytic activity at broad temperature range from 30°C to 80°C. Nearly 60% of the maximum activity was remained when the temperature increased to 80°C. We also measured the thermostability of rDIA by heating the enzyme at different temperatures for 40 min before the measurement. The rDIA showed good activity at temperatures of 30-60°C (Fig. 2B), and nearly 85% enzymatic activity was still remained after treated at 60°C. These results indicated that the optimum temperature of this rDIA was 55°C and possessed outstanding thermostability from 30-60°C. rDIA showed more thermal stability than DLD, which has been heterologous expressed and puri ed from E. coli with a sharp decrease in enzyme activity above 30°C and completely inactivated at 70°C (Kianmehr et al. 2017).
The effects of pH on the rDIA activity were evaluated after exposure to pH ranged from 5.0 to 9.5 at 37°C, and the results were shown in Fig. 2C. The enzyme activity in all pH conditions were all above 40%, and the optimum pH was 6.5. Besides, the stability of this enzyme at different pH was probed by treating the rDIA with different pH buffer ranging from 3 to 12 at 4°C for 6 h before the enzymatic measurement (Fig. 2D). The rDIA showed excellent pH stability with a broad range from 5.0 and 9.0, which remain over 85% activity. Even at pH 10.0, the enzyme maintained approximately 60% catalytic activity. These results proved that the optimum pH of rDIA was 6.5, and possessed good pH stability. In this study, the optimum pH and stability of the rDIA were similar to rBfmBC, encoded by Clostridium kluyveri rBfmBC gene . And the optimum pH of the recombinant enzyme rBfmBC was found to be about 7.0, while it showed high activity only in a narrow pH range of 6.0 to 8.0 in diaphorase assays ).
We also probed the effect of different mental ions and chemical compounds on the enzymatic activity of rDIA. The rDIA (5 U/mL) was dissolved in 0.1 M potassium phosphate (pH 6.5) was incubated with each chemical at 37 ℃ for 30 min, and the catalysis activity were measured. As can be seen in Table 1, all of the regents showed negligible effect on the enzyme activity, which was similar to the results of other diaphorases (Kianmehr et al. 2017;Dietrichs et al. 1990), which indicated that the rDIA presented promising stable enzymatic property. Moreover, to calculate the Michaelis constant, the substrate NADH was diluted with 2 times gradient to determine the activity of diaphorase at different concentrations of NADH. The kinetic curve of rDIA was drawn based on GraphPad prism 6 software. The results calculated by software of Michaelis constant showed Km = 0.09 mM and Vmax = 81.40 µM/min (Fig. 2E). Comparation of rDIA with commercial diaphorase We compared the enzymatic property of the rDIA with commercial diaphorase by using the βhydroxybutyric acid test kit (Abbexa, UK). The results showed that the enzymatic activity of rDIA was similar to the commercial diaphorase (Fig. 3A). Moreover, we evaluated the enzyme stability of rDIA and commercial diaphorase by storing at 4°C and 37°C for 14 days. As shown in Fig. 3B, the enzymatic activity of rDIA was similar to the commercial enzyme after stored at 4°C, while signi cantly higher than the other one at 37°C, which indicated the rDIA was more stable than commercial enzyme.
Structure changes and conformations transmission of rDIA at different pH were probed by MD simulations As the pH was important in the application of diaphorase, we investigated the structure and conformation transitions of rDIA at different pH solutions by using homologous modeling and MD simulations. Firstly, the Cα-RMSD and atom-atom contacts of the whole protein were used to represent the enzyme activity at two pH values of 3.0 and 6.5. Currently, Bacillus smithii's FMN-dependent NADH-azoreductase (PDB code: 6JXS, resolution: 1.95Å) has been crystallized, and has a sequence identity of 60.66% with rDIA (Fig. S3). The residue completeness and crystal structure resolution were also proved FMN-dependent NADHazoreductase could be considered as the homology model of rDIA in the following MD simulations.
The Cα-RMSD and atom-atom contacts were displayed as a function of time in Fig. 4. As shown in Fig. 4A, the Cα -RMSD value at pH 3.0 was higher than pH 6.5 during the last 25 ns. And atom-atom contacts at pH 6.5 was generally larger than pH 3.0 during the whole 50 ns (Fig. 4B). Therefore, the structure of rDIA was more stable at pH 6.5 than pH 3.0 based on the results of the MD simulation.
We also analyzed the changes of RMSF, which represented the exibilities of the individual residues in the protein (Sun et al. 2016), at pH 3.0 and 6.5. Time dependence of the RMSF values of Cα atoms of rDIA were calculated and displayed as shown in Fig. 4C. Two residue regions of 125-246 and 150-170 present higher RMSF values at pH 3.0 than 6.5, indicating the structure of these two regions were exible and easily in uenced by the pH changes. Of note, majority residues in 125-246 and 150-170 regions were negatively charged, we speculated that the negatively charged residues might contributed most in the rDIA structure changes at different pH.
The changes of secondary structures of rDIA at pH 3.0 and pH 6.5 was also investigated. The total contents of secondary structures of rDIA during the simulations at pH 3.0 and 6.5 were similar, but in the 141-161 domain, there were obvious differences in the secondary structure composition between these two pH ( Table 2). Real time simulations of the secondary structure changes 50 ns at pH 3.0 and 6.5 were also monitored in Fig. 5. Of note, in the 141-161 domain that labeled by black wire frame, the structure of helix changed into turns with the simulation time under the condition of pH 3.0 (Fig. 5A), while there were not signi cant changes in the same fragment at pH 6.5 (Fig. 5B). It was indicated that the secondary structures of rDIA easily changed at pH 3.0. Representative snapshots of rDIA at pH 3.0 and pH 6.5 values (Fig SI 4) also proved that pH can affect the structure of rDIA. Mutation prediction of rDIA by MD simulations The above results proved that pH can affect the structure of rDIA through different charged states of amino acids, especially of the 141-161 domain. Based on these results, we furtherly investigated the key residues on maintaining the enzymatic stability by using MD simulations, which was helpful for the direction evolution to improve the enzyme activity and stability. According to the homology modeling and MD simulations, the binding sites were formed from positions of 17-20, 52-60, 101-105, 146-150, 172, 186-187, which were related to enzyme catalysis pocket (Fig. 6). In order to improve the binding a nity, we did alanine-scanning mutagenesis for activity domain of rDIA and screened out the amino acid positions that had great energy changes, and then made site directed full mutation scanning of these amino acids. F105, M186 and W60 were found that there was a large energy change after alaninescanning mutagenesis. Then site directed full mutation scanning was carried out for F105, M186 and W60, and the energy of F105W, F105R and M186R was negative (Fig. 6A), which indicated that the energy was reduced. Due to the reduction in energy, these mutants were able to improve the binding a nity of rDIA (Fig. 6B). We did mutation-scanning far away from the activity domain to screen for stability mutations under the optimal pH 6.5. The results of mutation-scanning showed that the energy change of A2Y, P35F, Q36D, N210L, F211Y was negative, especially A2Y and Q36D (Fig. 6C). Due to the reduction in energy, these mutants were able to improve the stability of rDIA (Fig. 6D). In summary, F105W, F105R and M186R mutants were able to improve the binding a nity of rDIA, and A2Y, P35F, Q36D, N210L, F211Y mutants were bene t for the stability of rDIA.

Conclusion
In this work, we successfully expressed and puri ed a novel diaphorase, rDIA from Geobacillus sp. Y4.1MC1 in E. coli with a high purity over 95% and yield of 110 mg/L culture. Optimal temperature for the rDIA was ~ 55°C, and optimum pH was 6.5. The rDIA showed temperature stability at 30-60°C, and possessed high activity at a broad pH range from 5.0 to 9.0. MD simulations results showed the 141-161 domain of rDIA played important role in the structure changes and conformations transmission at different pH values. Moreover, we recommended that F105W, F105R and M186R mutants were able to improve the binding a nity of rDIA, and A2Y, P35F, Q36D, N210L, F211Y mutants were bene t for the stability of rDIA.