The detailed source of materials and chemicals were given in the Supplementary Material. E. coli DH5α and E. coli BL21 (DE3) strains were purchased from Thermo Fisher Scientific (MA, United States) and New England Biolabs (MA, United States), respectively, and used as hosts for plasmid transformation and expression. The A. flavus NRRL3357 strain was obtained from the ARS Culture Collection (NRRL). A. flavus NRRL3357 was grown at 28°C in YM broth containing yeast extract 3 g L⁻¹, malt extract 3 g L⁻¹, and peptone 5 g L⁻¹, and dextrose 10 g L⁻¹ with shaking at 200 rpm. E. coli DH5α and E. coli BL21 (DE3) strains were used for plasmid transformation and expression, respectively.

The A. flavus NRRL3357 genomic sequence and XDH protein sequence were accessed from the NCBI site (www.ncbi.nlm.nih.gov). A. flavus NRRL3357 was cultured and genomic DNA was obtained using the DNA Purification Kit (Promega). The xdh gene was amplified from A. flavus NRRL3357 genomic DNA using the two primers, 5′–GGG​ATC​CAT​GGC​TAC​AGA​TAC​TCA​TCC​C–3′ and 5′–GAG​CTC​CTA​TAC​TGA​AGC​TTG​CTT​TGC–3′, by polymerase chain reaction (PCR).

The amplified xdh gene with restriction sites BamHI and SacI was cloned into pGEM-T vector and transformed into the E. coli DH5α strain. Plasmid DNA and the expression vector pQE80L were digested with BamHI and SacI to release the xdh gene and create a nick in the pQE80L, respectively. The xdh gene was released from the pGEM-T vector and was ligated with the pQE80L to generate a recombinant plasmid, pQE80L-xdh. The pQE80L-xdh expressed XDH with a 6xHis tag. The pQE80L-xdh was transformed into E. coli BL21 (DE3) and the XDH was expressed using isopropyl-β-d-thiogalactopyranoside (IPTG, 0.5 mM) at 25°C. The induced cells were then harvested at 4°C by centrifugation for 20 min at 4,000 × g and rinsed with lysis buffer (pH 8.0, 50 mM NaH₂PO₄, 10 mM imidazole, 300 mM NaCl). The spnox gene from S. pyogenes was cloned into pET28a, expressed in E. coli BL21 (DE3), and purified as described previously (Gao et al., 2012; Li et al., 2018). Detailed expression and purification results of SpNOX were given in the (Supplementary Figure S1).

The recombinant 6xHis-tagged AfXDH was purified as described previously for other dehydrogenases (Tiwari et al., 2010). The activity of AfXDH was assayed by monitoring the change in A₃₄₀ upon oxidation of NADH at room temperature. The assay mixture contained 2 mM NAD⁺, 200 mM of xylitol, and AfXDH in 20 mM Tris-glycine buffer (pH 9.5). The reaction was initiated by adding xylitol. Enzyme assays were performed in triplicate using the purified AfXDH.

The optimum pH of the recombinant AfXDH was determined using 20 mM sodium phosphate (pH 6.0–7.0), 20 mM Tris-HCl (pH 7.5–8.5), and 20 mM Tris-glycine (pH 9.0–10.0) buffers. The optimum temperature for the AfXDH activity was determined at different temperatures ranging from 20 to 60°C. The enzyme stability was assayed by incubating the purified AfXDH in 20 mM Tris-glycine buffer (pH 9.5) containing 0.1 mM Zn²⁺. At specific timepoints, samples were withdrawn, and remaining activity was determined.

The substrate specificity of the purified AfXDH was determined under optimum conditions using various substrates such as sorbitol, xylitol, erythritol, glycerol, ribitol (adonitol), mannitol, galactitol, maltitol and d/l-arabinitol (200 mM).

The purified AfXDH was dialyzed at 4°C against 20 mM Tris-glycine buffer (pH 9.5) containing 10 mM ethylenediamine-tetraacetic acid (EDTA) for 12 h, and was dialyzed against 20 mM Tris-glycine-NaOH (pH 9.5). Then, the dialyzed enzyme was assayed using 200 mM xylitol in the presence of MnCl₂, CoCl₂, CaCl₂, HgCl₂, ZnCl₂, MgCl₂, BaCl₂, CuSO₄, and KCl at concentrations ranging from 0.05 to 5 mM.

The kinetic parameters of AfXDH were determined from triplicate measurements in 20 mM Tris-glycine buffer (pH 9.5), 0.1 mM Zn²⁺, 5–300 mM xylitol, and 0.1–3 mM NAD⁺. The K _m and V_max values were determined via the Michaelis–Menten equation using Prism 5 (GraphPad, Inc., CA, United States).

To evaluate the reaction products obtained after the conversion of sorbitol and xylitol, we analyzed the reaction products from in vitro reactions using purified AfXDH by high-performance liquid chromatography (HPLC) (UltiMate 3000, Dionex, Idstein, Germany). A Shodex Sugar SP column (8.0 μm × 300 mm, Showa Denko, Tokyo, Japan) was used and maintained at 85°C. The eluent was H₂O at a flow rate of 1.0 ml min⁻¹. Peaks were detected using an evaporative light scattering detector (ELSD) detector.

The homology models were generated using the Prime module in Schrödinger suite 2019 (Schrondinger, NY, United States) (Schrödinger Release 2020-2, 2020). The multiple sequence alignment of XDH sequences was performed using Schrödinger Maestro (Schrödinger, NY, 2020). For comparative analysis, structure information was obtained from the Protein Data Bank (PBD) (Berman et al., 2000). The modeled structures were further analyzed by SAVES server (Koivistoinen et al., 2012) for Ramachandran plot. The docking studies were performed by Extra Precision (XP) glide (Bowers et al., 2006) in Schrödinger suite 2019.

Molecular dynamics of AfXDH and ReXDH in complex with xylitol was performed by Desmond (Harder et al., 2016) Molecular Dynamics package of Schrodinger 2019. The OPLS3 (Jorgensen et al., 1996) force field was used both for protein atoms and xylitol to generate topologies for molecular dynamics (MD) simulation. All the systems were solvated in orthorhombic box and neutralized by Na⁺ ions as system set up method. After system was built, each system was undergone into 8 stages of MD simulation inclusive of minimization, equilibration, and final step of production run. All the production runs were carried out for 50 ns for each system, AfXDH and ReXDH. The snapshots were saved at regular interval of 1 ps.

The xdh gene encoding XDH from A. flavus (GenBank accession number: XM_00238625) was 1,110 nucleotides long with an open reading frame of 369 amino acids. It possessed the NAD-binding Rossmann fold domain [(GXGPXG)] (position 182–187; where X is any amino acid), the catalytic Zn²⁺-binding site (Cys47, His72, Glu73, and Glu158), and the structural Zn²⁺-binding site (Cys102, Cys105, Cys108, and Cys116) highly conserved among MDRs (Figure 1). According to these bioinformatics studies, we considered AfXDH to be an XDH (a member of MDR) with two metal binding sites. The Afxdh gene was cloned into the pQE80L plasmid between the SacI and BamHI sites and expressed in E. coli BL21 (DE3) with a 6xHis affinity tag at C-terminal. Gel filtration chromatography of the purified 6xHis-tagged AfXDH protein showed a symmetrical peak between myoglobin and ovalbumin, corresponding to a Mr of 39 kDa (Figure 2A). SDS-PAGE of the purified AfXDH exhibited a single band at ∼39 kDa (Figure 2B), suggesting that the native AfXDH is a monomer.

The optimum pH for the dehydrogenation of xylitol by purified AfXDH was 9.5 with 43% (28.4 U/mg-protein), 60% (39.6 U/mg-protein), 83% (54.8 U/mg-protein), and 92% (60.7 U/mg-protein) maximum activity at pH 8, 8.5, 9, and 10, respectively (Figure 3A). Maximal AfXDH activity at an alkaline pH optimum is a common feature of similar dehydrogenases/reductases obtained from various microorganisms. The optimum dehydrogenation temperature was found to be 50°C (Figure 3B). The AfXDH retained more than 95% of the maximum activity when analyzed at 55°C. In the presence of Zn²⁺, the half-life (t_1/2) values of AfXDH were 15, 7.4, and 3.2 h at 30, 40 and 50°C, respectively. However, the AfXDH stability decreased significantly at 55°C with a t_1/2 value of 30 min (Figure 3C). The t_1/2 values of AfXDH (200 and 120 min) at 40 and 50°C are the longest t_1/2 values among all characterized XDHs until now. The half denaturation temperature (T_1/2) of AfXDH was determined to be 45°C by analyzing the remaining activities after heat treatment at various temperatures varying from 25 to 55°C.

Xylitol, sorbitol, ribitol, galactitol, mannitol, meso-erythritol, glycerol, maltitol, and d/l-arabinitol (200 mM) were used as substrates to estimate the substrate specificity of AfXDH. The enzyme had a high preference for xylitol followed by sorbitol and showed low activity with other polyols. The coenzyme specificity of AfXDH was also examined; it exhibited coenzyme preference for NAD(H) over NADP(H). With xylitol as the substrate, the activity with NADP⁺ was 0.7% of the activity observed with NAD⁺, when 1.5 mM NAD⁺ or NADP⁺ was used as the coenzyme. AfXDH was exclusively an NAD⁺-dependent enzyme showing almost no activity with NADP⁺.

The activity of the AfXDH (1 mM) was thoroughly inhibited by Hg²⁺ and Cu²⁺ and strongly inhibited by Co²⁺ and K⁺. Slight inhibition was observed in the presence of Mg²⁺, Ba²⁺, and Co²⁺. AfXDH activity was stimulated by Mn²⁺. However, the highest activity was obtained using Zn²⁺ (0.1 mM), which enhanced AfXDH activity by 2.73-fold compared to the control in the absence of supplementary metal ions. The AfXDH activity was found to be dependent on divalent metal ions (Table 1).

ICP-MS analysis was performed to analyze the presence of metals in the purified AfXDH, we subjected the recombinant protein to. On the basis of the molecular mass of 38.9 kDa, 1 μmol of the AfXDH was found to contain 2 μmol of Zn²⁺. The sequence alignment (Figure 1) and ICP-MS analysis suggest that AfXDH contains two zinc-binding sites per monomer of AfXDH: a catalytic Zn²⁺ and a second structural Zn²⁺. According to multiple sequence alignment analysis of AfXDH with other XDHs, catalytic Zn²⁺ can bind to CYS47, HIS72, GLU73, and GLU158 (Koivistoinen et al., 2012). In addition, AfXDH activity was completely lost upon treatment with 10 mM EDTA and restored by the addition of 0.1 mM Mn²⁺ or Zn²⁺.

Initial velocities were determined at pH 9.5 in the standard assay mixture. The kinetic parameters for AfXDH activity were analyzed using substrate concentrations varying from 5 to 300 mM (Figure 4). Maximum AfXDH activity was observed with a substrate concentration of 150 mM under the assay conditions. The K _m, _xylitol value of AfXDH was found to be 16.2 mM (Table 2), which was lower than those of other XDHs from Fusarium oxysporum (Panagiotou et al., 2002) and Pichia stipites (Watanabe et al., 2005) but higher than XDHs from Pachysolen tanophilus (Morimoto et al., 1986) and Bacillus pallidus (Takata et al., 2010). The k _cat /K _m for xylitol was 2.88 s⁻¹ mM⁻¹. When the NAD⁺ concentration increased from 0.1 to 3 mM, Michaelis–Menten type of kinetics were observed for AfXDH activity. The K _m, _NAD+ value was 0.15 mM. The properties of various XDHs have been compared in Table 2. AfXDH shared similar characteristics with other reported XDHs, such as M_r (∼39 kDa), and had high affinity toward NAD⁺. However, notably, AfXDH had the highest k _cat value (48.7 s⁻¹) and thermostability (t_1/2 = 200 min at 50°C) among all previously characterized XDHs (Tiwari et al., 2010; Tiwari et al., 2012).

The multiple sequence alignment of AfXDH, Rhizobium etli XDH (ReXDH), Bemisia argentifolli XDH (PDB ID:1E3J), and Homo sapiens XDH (PDB ID: 1PL6) is shown in Supplementary Figure S2 (Banfield et al., 2001; Pauly et al., 2003). AfXDH was found to have sequence identities of 45%, 43.1%, and 38.0% with 1PL6, 1E3J, and ReXDH, respectively. The highly conserved active site residues (SER, GLU, and ARG) of XDHs have been reported (Lunzer et al., 1998; Filling et al., 2002; Ehrensberger et al., 2006). The active site residues, SER49, GLU158, and ARG304 of AfXDH correspond to SER42, GLU152, and ARG293 of ReXDH, and well conserved in all the XDHs. The coordinating residues (CYS44, HIS69, and GLU70) with catalytic zinc were also completely conserved (Supplementary Figure S2). The homology model of AfXDH (Supplementary Figure S3) obtained using IPL6 and 1E3J as templates was further validated by Ramachandran plot (Supplementary Figure S4) with 92.1% of residues in favored region, 6.2% in allowed regions, and 1.7% in disallowed regions. AfXDH has higher catalytic activity towards xylitol than other XDHs including ReXDH shown in Table 2. Homology model of ReXDH was generated using the same protocol used for AfXDH modeling. The final model structure of ReXDH was validated by Ramachandran plot (Supplementary Figure S5) where 86.9% of residues in favored regions, 5.6% in allowed regions, and 7.6% in disallowed regions.

To gain insight into the higher enzyme activity of AfXDH than other XDHs, we docked xylitol into both AfXDH and ReXDH. The docking energy of xylitol obtained for AfXDH and ReXDH from Glide XP method was found to be −6.077 (kcal/mol) and −4.857 (kcal/mol), respectively. The active site residues of AfXDH (SER49, GLU158, and ARG304) formed four hydrogen bonds with xylitol (Figure 5A). However, in the case of ReXDH, only two hydrogen bonds were formed between xylitol and GLU152 and SER42 (Figure 5B). The active site residues conserved in both the XDHs have similar structural fold in binding the substrate (Figure 5C). The greater number of hydrogen bonds formed by AfXDH with xylitol as compared to ReXDH is likely to contribute to its higher activity. The distances between donor and acceptor atoms in hydrogen bond for AfXDH and ReXDH are given in Supplementary Table S1.

The root-mean-square deviation (rmsd) of backbone atoms of AfXDH and ReXDH in complex with xylitol is shown (Figure 5D). The backbone atoms of AfXDH initially started to shoot up from 0.00 to 0.20 nm during 0–10 ns time. After 10 ns, the structure is found to be stable with no fluctuations at all in backbone atoms till 16 ns with rmsd in the range of 0.18–0.19 nm. AfXDH showed highest fluctuation at 28 ns with rmsd of 0.28 nm. In subsequent time, the backbone atoms stabilized more and maintained the rmsd of 0.20 nm until the end of the simulation. In case of ReXDH, the stability of backbone atoms was found to be less throughout the simulation with highest fluctuation in rmsd of 0.32 nm at 35 ns and overall rmsd in the range of 0.25–0.30 nm. The rmsd analysis clearly signifies the binding of xylitol in the active site of AfXDH is stable than ReXDH.

There is considerable interest in studies related to polyol dehydrogenases owing to their enormous potential for practical application in pharmaceutical and industrial processes. The production of rare sugars using microorganisms and their enzymes, particularly dehydrogenases, has been the prime focus of many reports (Morimoto et al., 1986; Panagiotou et al., 2002; Watanabe et al., 2005). In this study, we move a step further by introducing a highly efficient AfXDH in the conversion of a cheap polyol, xylitol (∼$5/kg), to an expensive rare sugar, l-xylulose (∼$12,600/g). The product of xylitol conversion by AfXDH was determined to be l-xylulose by HPLC. The product exhibited an indistinguishable retention time from that of authentic l-xylulose (Figure 6). It was confirmed by CD spectrometric methods and compared with d- and l-xylulose standards; finally, it was determined to be l-xylulose. The specific optical rotation of the purified product at 3% in H₂O at 25°C was +29°, whereas those of standard d- and l-xylulose are −33° and +31°, respectively, suggesting that the product has the l-configuration. However, the conversion yield obtained from 30 mM xylitol using purified AfXDH was only 5.8% in the presence of 5 mM NAD⁺, probably owing to NADH product inhibition (K _{i, NADH} = 0.25 mM). To achieve a higher product yield in a sustainable manner, a coupling enzymatic system was generated by introducing SpNOX, an NAD⁺ regeneration enzyme (Figure 6).

AfXDH exhibited a half-life of 15.3 h at 45°C, and SpNOX did not lose any activity below 50°C for up to 6 h. Therefore, both AfXDH and SpNOX can be considered stable for the entire course of the coupling enzymatic reaction (6 h). In consideration of the long lifespan of both AfXDH and SpNOX, the coupling enzymatic reaction also could be reasonably expected to be stable. NAD⁺ used in l-xylulose production was recycled by SpNOX. Using 0.5 mM NAD⁺ as the initial coenzyme, AfXDH catalyzed the oxidation of xylitol, forming l-xylulose; SpNOX catalyzed the reduction of H₂O₂ and subsequently regenerated the oxidized coenzyme NAD⁺; this consequently maintained low levels of NADH resulting in the acceleration of xylitol oxidation by AfXDH. The reaction parameters were optimized to obtain the highest yield of xylulose by coupling with SpNOX. The optimal pH was changed from 9.5 to 8. The optimal reaction temperature and reaction time were 30°C and 9 h, respectively. The effect of the ratio of SpNOX to AfXDH on l-xylulose production was also investigated by varying the amount of SpNOX at a fixed concentration of AfXDH (10 U ml⁻¹). The optimal ratio of SpNOX and AfXDH for the coupling reaction was 10 (Supplementary Figure S6). Therefore, the detailed conditions for the conversion of xylitol to l-xylulose were as follows: 10 U/ml AfXDH, 100 U/ml SpNOX, 0.5 mM NAD⁺, 0.1 M Tris-glycine buffer (pH 8.0), 25°C. In the presence of 0.5 mM NAD⁺, the yield of l-xylulose obtained from 15 g L⁻¹ (100 mM) of xylitol using purified AfXDH without coupling with SpNOX was merely 0.6% (0.09 g L⁻¹ of l-xylulose). The yield was remarkably enhanced to 87% [13.1 g L⁻¹ of l-xylulose (87 mM)] by coupling AfXDH with SpNOX (Figure 6C). The productivity of 2.18 g L⁻¹ h⁻¹ was obtained using only 0.5 mM initial NAD⁺. Khan et al. (1991) reported 0.5 g L⁻¹ h⁻¹ and 80% of l-xylulose productivity and yield, respectively, using resting recombinant cells when Pantoea ananatis ATCC 43072 xdh gene was overexpressed in E. coli. Usvalamp et al. (2009) reported 1.5 g L⁻¹ h⁻¹ of l-xylulose productivity using E. coli BPT228 resting recombinant cells. The productivity of the purified AfXDH was higher than that of the whole-cell systems reported. However, using resting cells for l-xylulose production has several advantages over the use of purified enzymes: regeneration of cofactors, stability of intracellular enzymes, and easy separation of l-xylulose (Usvalampi et al., 2009; Meng et al., 2016; Tesfay et al., 2021). The process using resting recombinant cells overexpressing AfXDH is currently under investigation in our laboratory.