Protein engineering of transaminase facilitating enzyme cascade reaction for the biosynthesis of azasugars

Summary Azasugars, such as 1-deoxynojirimycin (1-DNJ), exhibit unique physiological functions and hold promising applications in medicine and health fields. However, the biosynthesis of 1-DNJ is hindered by the low activity and thermostability of the transaminase. In this study, the transaminase from Mycobacterium vanbaalenii (MvTA) with activity toward d-fructose was engineered through semi-rational design and high-throughput screening method. The final mutant M9-1 demonstrated a remarkable 31.2-fold increase in specific activity and an impressive 200-fold improvement in thermostability compared to the wild-type enzyme. Molecular dynamics (MD) simulations revealed that the mutation sites of H69R and K145R in M9-1 played crucial roles in the binding of the amino acceptor and donor, leading to the stable conformation of substrates within the active pocket. An enzyme cascade reaction was developed using M9-1 and the dehydrogenase from Paenibacillus polymyxa (GutB1) for the production of mannojirimycin (MJ), which provided a new idea for the in vitro biosynthesis of 1-DNJ.


INTRODUCTION
Azasugars are sugar analogues in which one or more oxygen atoms in the sugar ring are replaced by nitrogen.Due to their structural similarity to common sugars, these compounds and their derivatives exhibit potent inhibitory activity against glycosidases, thus impeding carbohydrates decomposition and influencing the modification of oligosaccharide chains in glycolipid and glycoprotein biosynthesis. 1][4] One particular azasugar of significant interest is 1-deoxynojirimycin (1-DNJ).As a glucose analogue, it can competitively bind to glucosidase, inhibiting its activity and preventing the breakdown of polysaccharides into monosaccharides. 5Therefore, 1-DNJ and its derivatives demonstrate considerable potential for the prevention and treatment of diabetes. 6Miglitol, a 1-DNJ derivative, is a commonly used anti-diabetic drug. 7[10] Chemical synthesis methods for 1-DNJ tend to be cumbersome, inefficient, and environmentally unfriendly.However, certain organisms have the inherent capacity to synthesize 1-DNJ.For instance, the abundant presence of 1-DNJ in mulberry leaves suggested the existence of a biosynthetic pathway for 1-DNJ in mulberry. 11,124][15] Clark et al. identified key genes involved in the synthesis of 1-DNJ in B. amyloliquefaciens, confirming that the gabT1, yktC1 and gutB1 genes encode transaminase, phosphatase, and dehydrogenase, respectively. 16In the microbial synthetic pathway of 1-DNJ, fructose 6-phosphate acts as the amino receptor, and its transamination product undergoes a series of catalytic reactions to yield 1-DNJ as the final product.Evidently, the transaminase is essential for 1-DNJ biosynthesis.
The transaminase utilized for 1-DNJ biosynthesis is an u-transaminase widely employed as a biocatalyst for chiral amines synthesis. 17evertheless, the wild-type transaminase exhibits limitations in terms of substrate specificity, stability, and catalytic efficiency, necessitating enzyme engineering to meet industrial requirements.For instance, the specific activity of GabT1 from Paenibacillus polymyxa SC2 is merely 4.9 nmol/min/mg, resulting in extremely low efficiency in producing 1-DNJ with this strain. 15][20] In this study, we focus on the u-transaminase from Mycobacterium vanbaalenii (MvTA) which exhibits activity toward D-fructose rather than fructose 6-phosphate.Unlike GabT1, MvTA offers the advantage of producing transamination product that do not necessitate dephosphorylation by a phosphatase, thereby streamlining the synthesis pathway of 1-DNJ.Considering the low enzymatic activity of MvTA, a rapid and sensitive high-throughput screening strategy based on colorimetric assay was developed for enzyme engineering.A mutant M9 with significantly enhanced activity and thermostability has been successfully obtained and crystallized for structural resolution.Further improvement in specific activity was achieved through the utilization of high-resolution structure and computer-aided rational design approach.Molecular docking and molecular dynamics (MD) simulations have been conducted to elucidate the mechanism underlying the improved activity.Considering the enzyme's ability to directly transfer amino group to D-fructose, an in vitro enzyme cascade reaction has been established for the production of azasugars.The best MvTA mutant, in conjunction with the dehydrogenase from Paenibacillus polymyxa (GutB1), has been employed as biocatalysts to produce mannojirimycin (MJ), the precursor of 1-DNJ, using D-fructose as the substrate.Leveraging the efficient catalytic activity of MvTA mutant on various hexaketoses, a series of azasugars can be synthesized and utilized in the preparation of end products with desirable biological functions.

Development of high-throughput screening method
The transaminase MvTA demonstrates transaminate activity toward D-fructose 1, and the resulting 2-amino-2-deoxy-mannitol (ADM) 2 undergoes dehydrogenation under the enzyme GutB1 and spontaneously cyclizes to produce the valuable azasugar mannojirimycin (MJ), which serves as a precursor for 1-DNJ.However, the application performance of MvTA is hindered by its low activity and poor thermostability.To address these limitations, enzyme engineering is required to enhance its performance.Previous reports indicated that MvTA could catalyze the transamination reaction using D-fructose as the amino acceptor and 2-(4-nitrophenyl)ethan-1-amine 3 as the amino donor. 21In the reaction, 3 was converted into the corresponding aldehyde 4, which subsequently formed an imine 5 with 3.After tautomerization, a red precipitate 6 is produced (Figure 1A).The transamination reaction was conducted at different concentrations of MvTA, and the absorbance of the product was measured at 500 nm.A clear linear relationship was observed between the absorbance of the red precipitate product (OD 500 ) and the amount of enzyme added (Figure 1B), indicating that OD 500 can serve as a measure of enzyme activity.Based on the principle, a highthroughput screening method for mutants was established.Single clones from mutant libraries were cultivated in 96-deep-well plates.Cells were disrupted using freeze-thawing method to obtain supernatants, which were transferred to the 96-well plates for the transamination reaction.The product was tested using a colorimetric assay, and mutants exhibited higher OD 500 values (indicating a dark red color) were rapidly and sensitively screened (Figure 1C).

Enzyme engineering to improve activity and thermostability
The properties of enzyme MvTA were characterized, revealing that its optimal temperature and pH were 50 C and 6.0, respectively.The enzyme activity was independent of metal ions.However, MvTA displayed low activity toward D-fructose, with an activity level of 2.4 G 0.3 mU/mg under optimal conditions.Furthermore, MvTA exhibited extremely poor thermostability, with more than 80% activity loss after incubation for 1 h at 50 C (Figure S1).To gain a better understanding of the enzyme and facilitate its engineering, the three-dimensional structure of MvTA was obtained through homology modeling using Swiss-Model server.The transaminase from Arthrobacter sp.KNK168 (3WWI) was used as the template, 22 which showed 50.66% similarity with MvTA.The predicted structure of MvTA revealed a homodimeric arrangement with PLP (pyridoxal phosphate) as the coenzyme.The active pocket was situated at the interface of two chains of the homodimer.
Three strategies were employed for MvTA engineering.Firstly, based on the predicted structure of MvTA, 34 residues within a 5 A ˚range of catalytic site (Lys195) and PLP were selected for site saturation mutagenesis (Figure 2A).The second strategy involved the selection of mutation sites using the consensus analysis approach. 23,24A total of 250 transaminases exhibiting high similarity to MvTA were retrieved from the Uniprot database for multiple sequence alignment.Based on the alignment results, 24 candidate residues were identified for site saturation mutagenesis (Figure 2B).In total, site saturation mutagenesis libraries were constructed for 58 residues.Additionally, a random mutagenesis library was constructed using random PCR approach.The activities of individual clones from mutagenesis libraries were assessed using the high-throughput screening method, and the mutants exhibited increased activity were further confirmed using HPLC.
As a result, a total of 12 site-directed mutants with increased activity were identified from more than 6,000 individual clones (Figure 2C).It was observed that the residues located within the substrate binding pocket, such as L162, H69, and K145, significantly influenced the activity of MvTA.The mutant K142P suggested that the residues within the substrate channel might be involved in enzyme activity.Additionally, six mutants (A168E, A168N, H152N, R215G, S121A, and S121M) obtained through Strategy 2 showed increased activity compared to the wild type (WT).Furthermore, the random mutagenesis approach yielded the S105P mutant, which displayed a significant increase in activity.
The relative thermostability of the single mutants were evaluated.The results suggested that the mutations in residues surrounding the substrate binding pocket (L162I, H69K, K145R and K142P) had minimal impact on thermostability, while the mutations in residues positioned on the surface (S105P, S121M, H152N, A168E, and R215G) notably enhanced thermostability (Figure S2).The residues H152, A168 and R215 were mutated to the highly conserved amino acids based on sequence consensus analysis, indicating the significant contribution of natural evolution to the thermostability of the enzyme. 25he mutation sites within the substrate binding pocket and the substrate channel were combined to generate M2, M3, and M4, which exhibited progressively increased activity.Based on the level of thermostability improvement, the additional five mutation sites (S105P, S121M, H152N, A168E and R215G) were sequentially incorporated into M4, resulting in the generation of mutant M9 (Figure 3A; Table 1).The activity of M9 was measured as 39.4 G 1.5 mU/mg, demonstrating a remarkable 16.4-fold increase compared to that of WT.Notably, M9 exhibited significant enhancement in thermostability compared to WT (Figure 3B).The half-life (t 1/2 ) of M9 at 50 C represented a 208-fold improvement compared to WT (Table 2).Additionally, the results indicated that the optimal pH of M9 remained unchanged compared to WT, while the optimum temperature was elevated from 50 C to 55 C (Figure S3).
The structure of 8ISC and 8IVP displayed complete overall conformations except for the N-loop, suggesting that the mutation sites in M9 contributed to stabilizing the overall conformation of the enzyme and facilitating the acquirement of the complex structure.In all three structures, the asymmetric unit comprised two polypeptide chains forming a homodimer.The dimeric enzyme exhibited the characteristic structural architecture of class IV PLP-dependent enzymes, consisting of two subunits (Figure 4A).The interface between the two subunits constituted the binding pocket for the amino donor and acceptor (Figure 4B).Coenzyme PLP formed a Schiff base linkage with the catalytic residue Lys195 in the complex structures of 8ISC and 8IVP, consistent with previously research. 27B-factor analysis of 8ISC revealed pronounced flexibility in the loop region (Y138ÀT151) (Figure 4C).Upon comparing the B-factor values between 8ISC and 8IVP using the B-FITTER software, 28 it was observed that the overall B-factor values of 8IVP were lower than those of 8ISC (Figure S5).Particularly, Arg145 in 8IVP exhibited a significantly lower B-factor value than in 8ISC.The results suggested that fructose binding played a role in stabilizing the protein conformation and hinted at a potential interaction between fructose and the residue Arg145.

Molecular basis of enhanced activity
Molecular docking was performed to generate the crystal structure of M9-PLP-MBA and WT-PLP-MBA.Molecular dynamics (MD) simulations of these two complex structures were carried out using GROMACS.During the 100 ns trajectory, the distances between the N atom of (R)-MBA and O atom of aldehyde group on PLP (d MBA-PLP ) were measured (Figure S6A).It was observed that the value of d MBA-PLP reached a stable level within the last 80 ns trajectory of M9 complex trajectory, whereas it exhibited fluctuated throughout the 100 ns trajectory of WT complex.Additionally, the average value of d MBA-PLP for M9 was significantly smaller than that for WT.Trajectory analysis suggested that (R)-MBA displayed substantial swinging motion within the substrate binding pocket of WT.Hydrogen bond analysis revealed frequent formation of hydrogen bonds between (R)-MBA and H69 during the 100 ns trajectory of WT (Figure S6B), whereas hydrogen bonds between (R)-MBA and K69 in M9 were scarcely observed (Figure S6C).Representative structures from 90 to 100 ns trajectory were selected for further analysis, revealing the occurrence of a p-cation interaction between the N atom of the guanidine group of R145 and the benzene ring of (R)-MBA in M9 complex (Figure 5A).Furthermore, due to the mutation of H69K, the longer sidechain of lysine rotated away from (R)-MBA due to electrostatic repulsion by R145, thereby preventing the formation of a hydrogen bond with (R)-MBA (Figures 5A and 5B).Further analysis of the binding pocket demonstrated that the mutation of L162I stablized the benzene ring on (R)-MBA within the pocket (Figures 5C and 5D).Through the combined effects of the mutation sites K69, R145 and I162, (R)-MBA approached PLP in a stable conformation, thereby promoting the efficiency of the first half-reaction of transamination.
To elucidate the second half-reaction of transamination, molecular docking of the ligand pyridoxamine phosphate (PMP) and acetophenone (APO) were performed in both WT and M9 to generate the complex structures WT-PMP-APO and M9-PMP-APO, respectively, for MD simulation.Trajectory analysis and examination of representative structures from 90 to 100 ns trajectory revealed that hydrogen bonds were frequently formed between K142 and E56 in WT.However, due to the mutation of K142P in M9, the sidechain of proline was unable to form any hydrogen bonds with E56 (Figures 5E and 5F).As a result, a channel was opened for the release of APO and the entry of D-fructose,  potentially further enhancing the catalytic efficiency of MvTA.Likewise, in the M9-PLP-MBA system, there was no formation of hydrogen bond between P142 and E56, as indicated by an average distance of 8.39 A ˚between these residues.Moreover, the mutation of K142P notably heightened the rigidity of the loop region (137-146).MD simulations of the M9-PLP-MBA complex highlighted lower RMSF values for this loop region in M9 compared to WT (Figure S7).The increased rigidity in the loop region significantly contributed to the stabilization of R145, consequently enhancing the stability of (R)-MBA within the substrate pocket.Many studies reported the role of proline in enhancing stability within loop regions. 29,30The S105P mutation exhibited a similar situation.In both the first and second half-reactions, the RMSF values for the loop region (102-108) and its adjacent a-helices in M9 were notably lower compared to WT (Figure S7).These findings indicated that the proline substitution at position 105 could enhance the rigidity of the loop region, consequently reducing the flexibility of its neighboring a-helices.This increased rigidity significantly contributed to the thermostability of the enzyme, thereby resulting in enhanced activity.

Further enhancement of activity through rational design
To further enhance the activity of M9, a computational design approach was employed based on the high-resolution crystal structure (8IVP).Six specific residues (K69, Y74, V76, F129, R145, and T289) were selected as targets for mutation using the Cartesian_ddg application.The results revealed that residues Y74, V76, F129, and T289 played crucial roles in enzyme activity, as mutations at these positions completely abolished the activity.In contrast, residues K69 and R145 exhibited some variability, with their mutations leading to both increased and decreased activities.Notably, the mutant K69R (M9-1) demonstrated an impressive nearly 2-fold improvement in activity compared to M9 (Figure 6A).The half-lives (t 1/2 ) of M9-1 at 30, 40, and 50 C were 172.8, 122.5, and 44.3 h, respectively, showing no significant change compared to those of M9.
The kinetic parameters of WT, M9, and M9-1 toward (R)-MBA and D-fructose were determined and presented in Table 3.The kinetic parameters of WT toward D-fructose could not be reliably determined because of its extremely low activity.Notably, the K m value toward (F) WT: the hydrogen bond interaction between K142 and E56 may close the channel for substrate entry and exit.D-fructose of M9-1 was found to be smaller than that of M9, suggesting a higher affinity of protein for the amino acceptor.The significantly increased number of hydrogen bonds between D-fructose and M9-1 compared to WT also indicated the enhancement in fructose affinity.(Figure S8).Furthermore, The gmx_MMPBSA approach was employed to calculate the binding free energy between the amino acceptor and the protein.The results revealed that both the moving average (Mov.Av.) and average values of binding free energy between D-fructose and M9-1 were consistently lower compared to those between D-fructose and M9 throughout 500 ns trajectory (Figures 6B and 6C).The finding confirmed an enhanced affinity of the amino acceptor in the mutant M9-1.As a result, the distances between the N atom on amino  group of PMP and the O atom on ketone group of D-fructose were gradually reduced to less than 3 A ˚(Figure 6D), which facilitated the occurrence of transamination reaction.The hydrogen bonds were frequently formed between the sidechain of R69 and D-fructose (Figure 6E), suggesting the advantage of the replacement of lysine by arginine.
Conversely, the affinity for the amino donor in both M9 and M9-1 was weakened as a result of the protein engineering (Table 3), consistent with the observed reduction in the number of hydrogen bonds (Figure S8).However, the amino donor in the mutant showed a closer proximity to PLP due to a p-cation reaction between R145 and benzene ring of (R)-MBA.Significantly, the introduction of H69R and K145R mutations in M9-1 achieved a remarkable balance in the binding of the amino donor and acceptor, which provides insights into the protein engineering of enzymes with ping-pong mechanism.

In vitro enzyme cascade reaction for the preparation of azasugar
The maximum conversion rates of MvTA WT and M9-1 to four hexaketoses (D-fructose, D-allulose, D-tagatose, D-sorbose) were determined.The results demonstrated a significant enhancement in the maximum conversion rates of M9-1 compared to WT (Table 4).Notably, M9-1 exhibited a maximum conversion rate of over 50% for D-fructose.Mass spectrometric analysis confirmed the transamination products of the four hexaketoses, with molecular weights of 182.1 Da [M + H] + , providing evidence of the mutant M9-1's capability to catalyze the transamination reaction of these hexaketoses (Figure S6).
An in vitro enzyme cascade method was developed for the preparation of azasugar, employing M9-1 and the dehydrogenase GutB1 derived from Paenibacillus polymyxa (Figure 7A).When D-fructose was utilized as the substrate, the product MJ was quantified through the inhibition rate of glucosidase.Optimization of enzyme amounts and the M9-1-to-GutB1 ratio revealed that the transaminase played a crucial role as the rate-limiting enzyme, requiring a high concentration (1.0 mg/mL).Increasing the GutB1 concentration did not yield significant improvements in MJ yield, thus establishing an M9-1-to-GutB1 ratio of 10:1 (Figure 7B).As the reaction time extended, the inhibition rate of the reaction product gradually increased, indicating the accumulation of MJ.After 9 h of reaction, the increase in product inhibition rate showed a diminished trend.(Figure 7C).Subsequently, MJ could be further converted into 1-DNJ through epimerization, dehydration, and  reduction reactions.In the microbial biosynthetic pathway of 1-DNJ, 13 the transaminase GabT1 catalyzed the transamination reaction of fructose 6-phosphate (6-P-Fru), requiring subsequent dephosphorylation by YktC1 before dehydrogenation to generate ADM.By utilizing M9-1 for D-fructose transamination, the dephosphorylation step could be circumvented, and the cost of D-fructose is significantly lower compared to 6-P-Fru, rendering the in vitro enzyme cascade reaction more economical and efficient.

Conclusion
The MvTA was a suitable enzyme for the in vitro production of 1-DNJ through the initial transamination reaction of D-fructose.However, due to its low activity and thermostability, protein engineering was employed to enhance its performance.The obtained mutant M9, incorporating nine mutation sites, exhibited a remarkable 16.4-fold increase in specific activity and an impressive 208-fold improvement in thermostability compared to WT.Through further rational design using the crystal structure of M9, the mutant M9-1 with an additional 1.9-fold improvement in specific activity was obtained.MD simulations analysis indicated that the mutation sites H69R and K145R in the flexible loop of M9-1 were found to play crucial roles in the binding of the amino acceptor and donor, respectively.Together with other mutant sites such as L162I and K142P, the substrates achieved a more stable conformation within the active pocket and better access channels, facilitating the two-stage ping-pong reaction of transamination.The M9-1 mutant was combined with the dehydrogenase GutB1 to develop an enzyme cascade reaction for the in vitro production of mannojirimycin (MJ), a precursor of 1-DNJ.Furthermore, given the improved catalytic activity of M9-1 toward various hexaketoses, the synthesis of a range of azasugars might be explored, enabling the preparation of end products with desirable biological functions.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Escherichia coli DH5a strain was used to prepare the plasmids and construct mutants.The cells were incubated in LuriaÀBertani (LB) medium containing 100 mg/mL ampicillin at 37 C. E. coli BL21(DE3) strain was used to express the target protein.The cells were grown in LB medium containing 100 mg/mL ampicillin at 37 C until the OD 600 reached 0.6-0.8.Protein expression was induced by adding 0.5 mM IPTG at 25 C for 20 h.

METHOD DETAILS Materials
The D-fructose, D-allulose, D-sorbose and D-tagatose were purchased from Sigma-Aldrich (St Louis, MO).All other reagents were of analytical grade and obtained from Aladdin (Shanghai, China).The wild-type genes of MvTA and GutB1 were synthesized by GENEWIZ (Suzhou, China).

High-throughput screening method
The single clones from mutant libraries were cultivated in 96-deep-well plates with LuriaÀBertani medium supplemented with 100 mg/mL ampicillin at 37 C. Protein expression was induced by adding 0.5 mM IPTG to the medium.After 24 h of incubation, cells were centrifuged at 5000 rpm for 20 min and resuspended in phosphate buffered saline (PBS; 100 mM, pH 7.0) containing 1 mg/mL lysozyme.The suspension was incubated at 37 C for 1 h, followed by cell disruption using a freeze-thaw method.Supernatants were obtained by centrifugation at 5000 rpm for 30 min and transferred to 96-well plates for the transamination reaction.The reaction system included 25 mM 4-nitrophenethylamine, 10 mM D-fructose, and 0.1 mM pyridoxal phosphate (PLP).After incubating at 30 C for 12 h, the absorbance at 500 nm (OD 500 ) was measured using a BioTek Synergy LX Multi-Mode Reader (Agilent, US).Clones with higher OD 500 values than the wild type were selected for further analysis.

Figure 1 .
Figure 1.Establishment of a high-throughput screening method (A) Colorimetric measurement principle of transaminase activity using D-fructose as the amino acceptor.(B) Standard curve correlating absorbance values at 500 nm with enzyme activity.(C) Workflow of the high-throughput screening process.

Figure 2 .
Figure 2. Strategies for constructing site-saturation mutagenesis libraries and generating single-site mutants with enhanced activity (A) Strategy 1: saturation mutagenesis targeting amino acid sites within 5 A ˚of the catalytic site (K195) and PLP.(B) Strategy 2: saturation mutagenesis of conserved amino acid sites identified through consensus analysis.(C) Single-site mutants with improved activity.S1, S2, and S3 represented for Strategy 1, Strategy 2, and Strategy 3, respectively.

Figure 3 .
Figure 3. Ehancement of specific activity and thermostability in combinatorial mutants (A) Specific activities of the combinatorial mutants.(B) Residual activity of M9 (solid line) and WT (dashed line) after incubated at 30 (blue), 40 (green), and 50 C (orange).

Figure 4 .
Figure 4. Structure analysis of the M9 mutant (A) Overall structure of M9-PLP (PDB ID: 8ISC) complex highlighting the covalent bond between K195 and PLP.(B) Overall structure of M9-PLP-Fru (PDB ID: 8IVP) complex illustrating the binding both PLP and d-fructose, indicating the presence of a binding pocket at the interface of two subunits.(C) B-factor analysis of the M9 structure, providing insights into the flexibility of the protein.

Figure 5 .
Figure 5.The molecular basis of enhanced enzyme activity by mutation sites Structure comparison between M9 (A, C and E) and WT (B, D and F) illustrating the molecular basis of enhanced activity.(A) M9: p-cation interaction between R145 and the benzene ring of (R)-MBA, absence of hydrogen bond between K69 and (R)-MBA.(B) WT: hydrogen bond between H69 and (R)-MBA, resulting in greater distance between the amino donor and PLP.(C) M9: tight binding of (R)-MBA within the active pocket due to the synergistic effect of I162 and R145 in the complex structure.(D) WT: (R)-MBA potentially swings within the active pocket due to a larger active pocket.(E) M9: absence of hydrogen bonds between the sidechain of P142 and E56, facilitating the exit of APO and entry of D-fructose.(F)WT: the hydrogen bond interaction between K142 and E56 may close the channel for substrate entry and exit.

Figure 6 .
Figure 6.Further enhancement in MvTA activity through rational design (A) Relative activities of mutations exhibiting DDG < À1.0, caculated using the Cartesian_ddg application.(B) Moving average values (Mov.Av.) of binding free energy between D-fructose and M9 or M9-1.(C) Average values of binding free energy between D-fructose and M9 or M9-1.(D) Distances between the N atom on amino group of PMP and the O atom on ketone group of D-fructose during the 500 ns trajectory.(E) Hydrogen bond between R69 and D-fructose in M9-1.

Figure 7 .
Figure 7. Establishment of an enzyme cascade reaction for biosynthesis of azasugars (A) Biosynthesis pathway for azasugars, such as MJ, 1-DMJ and 1-DNJ.(B) Optimization of enzyme dosage of transaminase M9 and dehydrogenase GutB1 for MJ preparation using D-fructose as the substrate.(C) Process of the enzyme cascade reaction.

Table 1 .
Specific activities of MvTA mutants relative to WT

Table 2 .
Half-lives of purified WT and mutant M9 at different temperature

Table 4 .
Maximum conversion rate of WT and mutant M9-1 toward D-hexaketoses