Coenzyme Engineering of a Hyperthermophilic 6-Phosphogluconate Dehydrogenase from NADP+ to NAD+ with Its Application to Biobatteries

Engineering the coenzyme specificity of redox enzymes plays an important role in metabolic engineering, synthetic biology, and biocatalysis, but it has rarely been applied to bioelectrochemistry. Here we develop a rational design strategy to change the coenzyme specificity of 6-phosphogluconate dehydrogenase (6PGDH) from a hyperthermophilic bacterium Thermotoga maritima from its natural coenzyme NADP+ to NAD+. Through amino acid-sequence alignment of NADP+- and NAD+-preferred 6PGDH enzymes and computer-aided substrate-coenzyme docking, the key amino acid residues responsible for binding the phosphate group of NADP+ were identified. Four mutants were obtained via site-directed mutagenesis. The best mutant N32E/R33I/T34I exhibited a ~6.4 × 104-fold reversal of the coenzyme selectivity from NADP+ to NAD+. The maximum power density and current density of the biobattery catalyzed by the mutant were 0.135 mW cm−2 and 0.255 mA cm−2, ~25% higher than those obtained from the wide-type 6PGDH-based biobattery at the room temperature. By using this 6PGDH mutant, the optimal temperature of running the biobattery was as high as 65 °C, leading to a high power density of 1.75 mW cm−2. This study demonstrates coenzyme engineering of a hyperthermophilic 6PGDH and its application to high-temperature biobatteries.

Scientific RepoRts | 6:36311 | DOI: 10.1038/srep36311 pentose phosphate pathway, converts 6-phophogluconate and NADP + to ribulose 5-phosphate, NADPH and CO 2 . The thermostable 6PGDH from T. maritima is very stable with a half-life time of 140 h at 80 °C, and can be easily purified by heat precipitation 10 . However, this enzyme prefers NADP + to NAD + so that it has not been tested in biobatteries that should work on NAD + as the coenzyme.
Rational design is one of the protein engineering approaches to engineer enzymes based on knowledge of their structures and catalytic mechanisms 11,12 . It may offer a rapid solution to tailor enzymes with desired properties. In bioelectrocatalysis, enzyme kinetics, stability, and electron transfer efficiency can be enhanced by rational design methods, including mutating several key amino acids, trimming off nonessential protein fragments, shuffling domains, and modifying protein surface 13,14 . Coenzyme engineering that changes coenzyme selectivity (i.e., NAD + or NADP + ) of oxidoreductases is one of the important tools in metabolic engineering, synthetic biology, and biocatalysis. Changing the coenzyme selectivity of dehydrogenases from NADP + to NAD + is highly desirable because (1) NAD + is less costly than NADP + 15,16 and (2) NADH is more stable than NADPH [17][18][19] . For biobatteries, changing coenzyme preference to NAD + must be important because NADP + cannot be efficiently used to generate electricity 8,20 . Intensive studies have been conducted to change or broaden coenzyme selectivity of redox enzymes from NADP + to NAD + 21-24 . However, there are few studies pertaining to the coenzyme engineering of thermostable dehydrogenases with the application to biobatteries 25 .
In this study, we demonstrated to switch coenzyme preference of T. maritima 6PDGH (Tm6PGDH) from NADP + to NAD + by rational design. The best Tm6PGDH mutant was applied in the biobattery, exhibiting an improved performance and a high power density at the elevated temperature.
Scientific RepoRts | 6:36311 | DOI: 10.1038/srep36311 hand, NAD + -preferred 6PGDHs are rare, including those from Haloferax volcanii (D4GST8), Gluconobacter oxydans (G5EBD7) and engineered M. thermoacetica 6PGDH. The alignment of the loop region shows that three amino acids (positions 32, 33, and 34) in NADP + -preferred 6PGDHs are highly conservative (Fig. 1b), which are asparagine, arginine and threonine (Asn32, Arg33 and Thr34) in Tm6PGDH. Via NADP + docking, these residues are found in close contact with the 2′ -phosphate of the adenosine ribose via hydrogen bonds (Fig. 1b). Additionally, Arg33 is able to form a cation-phosphate interaction with the NADP + adenosine ring, further stabilizing the NADP-enzyme complex. However, NAD + -preferred 6PGDH enzymes have the acidic aspartate residue in the N-terminal end of loop region, which is also very conservative and completely different with that of NADP + preferred-6PGDH. The other two residues exhibit three different combinations. According to the above information, we hypothesized that the changes in Asn32, Arg33, and Thr34 could enable Tm6PGDH to work on NAD + . Consequently, three different mutants (i.e., N32D, N32D/R33I/T34I and N32D/R33L/T34S) were firstly designed as mutation candidates.
Structural basis of coenzyme preference. To identify key amino acids pertaining to coenzyme preference, a detailed analysis of the coenzyme-6PGDH complexes was performed based on the auto-docking program. The mutant N32D has the replacement of the alkaline asparagine with the acidic aspartate, deleting the former hydrogen bond between the 2′ -phosphate of NADP + and the asparagine residue of the wide type Tm6PGDH (Fig. 2a) and therefore resulting in a decrease in the binding affinity of the mutant N32D towards NADP + . To further enhance coenzyme preference of N32D, the amino acids with large and hydrophobic side chain (leucine and isoleucine) were introduced to the position 33 and 34 to construct the mutant N32D/R33L/T34S, N32D/ R33I/T34I. The newly-introduced residues can further disrupt the other two original hydrogen bonds with 2′ -phosphate of NADP + . Moreover, their large and hydrophobic side chains display much stronger steric exclusion effect and compress the space used to accommodate the 2′ -phosphate group of adenosine monophosphate moiety of NADP + (Fig. 2b,c). The combined effect of the deletion of original hydrogen bond and the enhanced steric exclusion should lead to a significant decrease in binding with NADP + for the enzyme.
To better bind NAD + , it is also important to form productive hydrogen bonds. In our case, the distance between the hydrogen atom in 2′ -hydroxyl group of NAD + and the oxygen atom in carboxyl group of Asp32 in the mutant N32D and N32D/R33I/T34I is calculated as 3.15 Å (Fig. 2e,g), which equals to the threshold value of the hydrogen bond formation 26 . Therefore, their ability to bind NAD + could be improved slightly. In the mutant N32D/R33L/T34S (Fig. 2f), however, such distance increases to 3.33 Å, which is too long to form a productive hydrogen bond. Consequently, this mutant may have no positive effect on NAD + binding. Instead, when the residue aspartate at position 32 is mutated to a similar one, glutamate, this distance can be decreased to 2.10 Å (Fig. 2h). In this new mutant N32E/R33I/T34I, a new hydrogen bond is formed, increasing the binding with NAD + .

Enzyme kinetics analysis.
According to the alignment result and structure analysis, the plasmids encoding mutants N32D, N32D/R33L/T34S, N32D/R33I/T34I and N32E/R33I/T34I were constructed by site-directed mutagenesis. The protein mutants were overexpressed in E. coli BL21 (DE3) and purified to homogeneity based on SDS-PAGE (data not shown). The kinetic constants on NADP + and NAD + of the wild type and mutant 6PGDHs are summarized in Table 1. The mutant N32D has a 460-time higher K m value of 2.3 mM on NADP + and a slightly decreased K m value of 4.0 mM on NAD + compared with those of the wide type. The triple-site mutant N32D/R33I/T34I exhibits a far higher K m value of 69.5 mM on NADP + and a slightly decreased K m value Hydrogen bonding between residues and cofactor were shown as red line. The distance between the two atoms were shown as yellow line. The figures were made using PyMol. of 3.9 mM on NAD + . Another mutant N32D/R33L/T34S has a K m value on NADP + of more than 100 mM. Its K m value on NAD + is also increased to 7.5 mM. However, these three mutants do not show significant differences of k cat on NADP + with that of the wild-type.
The mutant N32E/R33I/T34I was found to be the best one. As shown in Table 1 and Fig. 3, an almost 2-fold declined K m (from 4.3 mM to 2.5 mM) and a 2-fold increased k cat on NAD + (from 23.3 to 47.9 s −1 ) compared to that of wide type can be observed. The catalytic efficiency (k cat /K m ) towards NADP + decreases from 3,520 to 0.2 s −1 , while the catalytic efficiency towards NAD + increases from 5.4 to 19.2 s −1 . Overall, the coenzyme catalytic coefficient ratio of NAD + to NADP + increases about 6.4 × 10 4 times for this mutant.
Biobattery tests. The effect of the best mutant (N32E/R33I/T34I) versus the wild type 6PGDH was evaluated electrochemically in an anodic reaction system containing two enzymes: 6PGDH and DI, a coenzyme (NADP + or NAD + ), an electron mediator AQDS, and a 6-phosphogluconate substrate (6PG). Cyclic voltammetry results clearly show that both 6PGDHs produce significant oxidation current peaks at − 0.3 V versus Ag/AgCl. The mutant N32E/R33I/T34I exhibits a current density 25% higher than that generated by the wild type (Fig. 4a). The blank control samples including the one without the fuel, without the NAD + , or with denatured 6PGDH did not generate any oxidation current peak. The current densities of 6PG oxidation at various NAD + concentrations were further evaluated, by plotting peak currents at − 0.3 V versus Ag/AgCl against NAD + concentrations varying from 0 to 2 mM (Fig. 4b). The current density increases along with the increase of the NAD + concentration, while the mutant generates currents 20-30% higher than those by the wild-type. Both figures suggest that the mutant N32E/R33I/T34I is more effective than the wild-type on using NAD + as the coenzyme in the electrochemical reaction system.
Moreover, both wide-type enzyme and mutant N32E/R33I/T34I were compared in biobatteries in terms of power output using linear sweep voltammetry. The whole-cell biobattery consisted of a carbon nanotube-coated carbon felt anode and a commercial membrane electrode assemble stacked in pile. An increased loading of 6PG, NAD + , AQDS, and 6PGDH was applied in order to get a high power output of the biobattery. The two polarization curves represent that the maximum power density and current density of the biobattery catalyzed by the mutant 6PGDH are 0.135 mW cm −2 and 0.255 mA cm −2 respectively at the room temperature, while for the biobattery catalyzed by the wild type 6PGDH, they are 0.122 mW cm −2 and 0.200 mA cm −2 (Fig. 4c). At elevated temperatures, the mutant 6PGDH-powered biobattery significantly increased power outputs of 0.61 mW cm −2 at 37 °C, 1.58 mW cm −2 at 50 °C, 1.75 mW cm −2 at 65 °C, and 1.58 mW cm −2 at 75 °C (Fig. 4d). It is noted that the biobattery has an optimal power density at 65 °C, probably due to the optimal enzyme kinetics and the increased mass transfer at this temperature. Too high temperature caused the deactivation of diaphorase, resulting in decreased power outputs in the biobattery.

Discussion
Most NAD(P) + -dependent dehydrogenases have the Rossmann fold, whose loop region contains three conservative amino acids. Alkaline amino acids, such as asparagine and arginine, in the first two positions play a critical role in NADP + -preferred 6PGDHs through the formation of stable hydrogen bonds with 2′ -phosphate of NADP + 27-29 . In contrast, NAD + -preferred 6PGDH enzymes have acidic aspartate residues in the N-terminal end of loop region (Fig. 1b). The substrate docking result (Fig. 2) indicates that the formation of a new hydrogen bond with NAD + through the replacement of alkaline amino acid by acidic one is favorable to increase the enzyme activity on NAD + , and even reverse the coenzyme preference from NADP + to NAD + . Moreover, our previous research on the directed evolution of Moorella thermoacetica 6PGDH also demonstrated that the introduction of an acidic amino acid, aspartate, in the loop of the Rossmann fold to form a new hydrogen bond with 2′ -hydroxyl group of NAD + resulted in a 4,300-fold reversal of coenzyme selectivity from NADP + to NAD + 30 . Reports on many other dehydrogenases also confirmed the positive effect of the formation of new hydrogen bond in reversing the coenzyme preference 21,24,30-34 .
Our results of kinetic constants of the wild-type enzyme and the mutant N32E/R33I/T43I are in good agreement with the structure analysis. For the mutant N32E/R33I/T43I, the coenzyme binding ability of 6PGDH on NADP + is decreased greatly, because the mutant loses its original three hydrogen bonds with the 2′ -phosphate of NADP + and suffers from a strong steric exclusion effect from the hydrophobic side chain of two newly introduced isoleucines (Fig. 2d). Those significant structural changes result in a 15,800-fold increase in K m on NADP + and a 17,600-fold decline in catalytic efficiency (k cat /K m ) on NADP + (Table 1). While on NAD + , the introduced acidic residue, glutamate, improves the binding of coenzyme due to the formation of the new productive hydrogen bond between Glu32 and 2′ -hydroxyl group of NAD + . A 2-fold decline in K m and a 4-fold increase in catalytic efficiency on NAD + compared to wide type Tm6PGDH are observed ( Table 1). As a result, the overall coenzyme catalytic coefficient ratio of NAD + to NADP + of this mutant increases about 6.4 × 10 4 times.
As a thermostable enzyme with a half-life time of 140 h at 80 °C, 6PGDH from T. maritima has great potential in biobattaries. Firstly, it has a prolonged lifetime of biobattaries due to high stability of the enzymes 35 . Secondly, it allows to run biobattaries at elevated temperatures or tolerate harsh conditions, such as in deserts or extremely hot conditions. Third, high-temperature biobattaries would further generate high power density because of increased enzyme activity 8,10 and mass transfer 36 . To our limited knowledge, this present biobattery is the one running on the second highest temperature among all the EFCs reported. The highest temperature of EFC was 80 °C by using a thermostable alcohol dehydrogenase but its current density was only 0.21 mA cm −2 (5% of this study) 37 .
In conclusion, this work reports a coenzyme specificity change of Tm6PGDH through rational design with its application to biobatteries. Via enzyme sequence alignment and structure analysis of the coenzyme-enzyme complexes, the key amino acid residues that are responsible for binding coenzymes were identified. The coenzyme preference was reversed from NADP + to NAD + through site-directed mutagenesis. The optimal temperature of the biobattary with the mutant was as high as 65 °C, resulting in a high power density of 1.75 mW cm −2 .

Methods and Methods
Chemicals, Bacterial Strains and Medium. All chemicals were reagent grade or higher, purchased from Sigma-Aldrich (St. Louis, MO) or Fisher Scientific (Pittsburgh, PA), unless otherwise noted. E. coli Top10 was used for general molecular cloning and E. coli BL21 (DE3) was used for recombinant protein expression. All enzymes for molecular biology experiments were purchased from New England Biolabs (NEB, Ipswich, MA). Strains, plasmids, and oligonucleotides primers used in this study are listed in Table 2.
Site directed mutagenesis. The Tm6PGDH gene was amplified from plasmid 10 by a primer pair of 6PGDH_ IF and 6PGDH_IR. pET20b vector backbone was amplified with a primer pair of 6PGDH_VF and 6PGDH_VR. Plasmid pET20b-6pgdh based on the two DNA fragments was obtained using a Simple Cloning method 38 .
The QuickChange TM site-directed mutagenesis method (Stratagene, La Jolla, CA) was used to introduce point mutations into the Tm6PGDH sequence according to the protocol of the NEB Phusion site-directed mutagenesis kit. The mutant N32D, N32D/R33I/T34I, N32D/R33L/T34S, and N32E/R33I/T34I were prepared by their respective primer pairs (Table 2). PCR reaction solution (50 μ L) containing 1 ng of plasmid template (pET20b-6pgdh) was conducted as follows: 98 °C denaturation for 1 min; 20 cycles of 98 °C denaturation for 30 s, 60 °C annealing for 30 s and 72 °C extension for 2.5 min; and 72 °C extension for 5 min. The PCR product was digested by DpnI followed by the purification of gel electrophoresis using a Zymoclean ™ Gel DNA Recovery Kit (Zymo Research, Irvine, CA) and then transformed into E. coli BL21 (DE3).
Overexpression and purification of wide-type Tm6PGDH and mutants. The transformants were grown overnight at 37 °C in Luria-Bertani medium plates containing 100 μ g mL −1 ampicillin. For liquid culture, once A 600 of the cultured cell reached ~0.8, 0.2 mM IPTG was added to induce protein expression at 37 °C for 6 h. Cell pellets were harvested by centrifugation and then were re-suspended in a 20 mM sodium phosphate and 0.3 M NaCl buffer (pH 7.5) containing 10 mM imidazole. After sonication and centrifugation, the His-tagged protein in the supernatant was loaded onto the column packed with HisPur Ni-NTA Resin (Fisher Scientific, Pittsburgh, PA)  and eluted with 20 mM sodium phosphate buffer (pH 7.5) containing 300 mM NaCl and 250 mM imidazole. Mass concentration of protein was determined by the Bradford assay using bovine serum albumin (BSA) as the standard.
Tm6PGDH activity assay. The activities of wide type Tm6PGDH and mutants were measured in 100 mM HEPES buffer (pH 7.5) with the final concentration of 2 mM 6-phosphogluconate (6PG), 2 mM NAD(P) + , 5 mM MgCl 2 and 0.5 mM MnCl 2 at 70 °C for 5 min. The formation of NAD(P)H was measured at 340 nm by a UV/visible spectrophotometer (Beckman Coulter, Fullerton, CA). The enzyme unit was defined as one μ mole of NAD(P) H produced per min. To determine enzyme kinetic parameters on coenzymes, the enzyme activity was measured in the same buffer above except changing the concentration of NAD(P) + (2 μ M to 10 mM). The result was regressed by GraphPad Prism 5 (Graphpad Software Inc., La Jolla, CA) and the apparent K m and k cat for NAD(P) + of wide type and mutant Tm6PGDH were given based on Michaelis-Menten nonlinear regression. All the reactions contained three independent triplicates and the data were fitted within the linear range.
Electrochemical measurement. All electrochemical measurements were conducted using a 600D Potentiostat from CH Instruments Inc. (Austin, TX) interfaced to a PC. The reaction solution was flushed by argon gas for 20 minutes before each measurement. Each measurement was conducted in triplicate.
Cyclic voltammetry was performed in the anodic compartment at a scan rate of 20 mV s −1 using a 3-electrode system with a GC working electrode, an Ag/AgCl reference electrode and a Pt wire counter electrode. The anolyte contained up to 2 mM NAD + , 5 mM 6PG, 2.5 mM AQDS, 100 mM HEPES (pH 7.3), 10 mM MgCl 2 , 0.5 mM MnCl 2 , 100 mM NaCl, 0.0133 mg mL −1 of mutated or wild type Tm6PGDH, and 0.015 mg mL −1 of Geobacillus stearothermophilus diaphorase (DI).
Linear sweep voltammetry was performed at a scan rate of 5 mV s −1 in a stacked fuel cell configuration. A 1-cm 2 carbon felt (CF) was coated with 0.25 mL of CNTs solution (CNTs: CTAB: H 2 O = 1 mg: 1 mL: 1 mL) and 30 μ L of GA, dried at the room temperature, and used as the anode. An air-breathing carbon cathode was coated with 0.5 mg cm −2 Pt. A Nafion 212 membrane was used to separate two electrodes. The anolyte contained increased loadings of 8 mM NAD + , 40 mM 6PG, 16.7 mM AQDS, 0.05 mg mL −1 of mutated or wild type Tm6PGDH, and 0.06 mg mL −1 of DI. The performance of the biobattery with mutated Tm6PGDH was further evaluated at increased temperatures up to 75 °C.
Sequence alignment. Manually annotated and reviewed sequence data for 6PGDH were retrieved from the National Center for Biotechnology Information and Uniprot Database 39 . Clustal Omega 40,41 was used to perform a multiple sequence alignment. The sequence logo plot 42 was created with the WebLogo 3.3 interface 43 .
Structure analysis. The three-dimensional (3D) structure modeling of the wild-type Tm6PGDH and mutants were built by SWISS-MODEL based on the Lactococcus lactis 6PGDH (PDB: 2IYP) as a template (sequence identity: 46.5%). The final model was validated using the PROCHECK program 44 . The structures of NADP + and NAD + were built by using Chemdraw (PerkinElmer, Waltham, MA). The conformation space of the corresponding coenzyme binding area was analyzed using the Autodock program (Scripps Research Institute, La Jolla, CA).