Structural basis of L-glucose oxidation by scyllo-inositol dehydrogenase: Implications for a novel enzyme subfamily classification

For about 70 years, L-glucose had been considered non-metabolizable by either mammalian or bacterial cells. Recently, however, an L-glucose catabolic pathway has been discovered in Paracoccus laeviglucosivorans, and the genes responsible cloned. Scyllo-inositol dehydrogenase is involved in the first step in the pathway that oxidizes L-glucose to produce L-glucono-1,5-lactone with concomitant reduction of NAD+ dependent manner. Here, we report the crystal structure of the ternary complex of scyllo-inositol dehydrogenase with NAD+ and L-glucono-1,5-lactone at 1.8 Å resolution. The enzyme adopts a homo-tetrameric structure, similar to those of the inositol dehydrogenase family, and the electron densities of the bound sugar was clearly observed, allowing identification of the residues responsible for interaction with the substrate in the catalytic site. In addition to the conserved catalytic residues (Lys106, Asp191, and His195), another residue, His318, located in the loop region of the adjacent subunit, is involved in substrate recognition. Site-directed mutagenesis confirmed the role of these residues in catalytic activity. We also report the complex structures of the enzyme with myo-inositol and scyllo-inosose. The Arg178 residue located in the flexible loop at the entrance of the catalytic site is also involved in substrate recognition, and plays an important role in accepting both L-glucose and inositols as substrates. On the basis of these structural features, which have not been identified in the known inositol dehydrogenases, and a phylogenetic analysis of IDH family enzymes, we suggest a novel subfamily of the GFO/IDH/MocA family. Since many enzymes in this family have not biochemically characterized, our results could promote to find their activities with various substrates.


Introduction
Homochirality, in the form of L-amino acids and D-sugars, exists in all living organisms In the case of sugars, it had been a long held belief, first reported in 1940 by Rudney, that L-glucose cannot be metabolized by either mammalian or bacterial cells [1]. Subsequently, Sasajima et al. purified D-threo-aldose dehydrogenase from Pseudomonas caryophylli, which was capable of a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Tris-HCl, pH 7.5, 10% glycerol] containing 20 mM imidazole, and the recombinant protein was eluted with buffer A containing 500 mM imidazole. The eluted protein was then dialyzed against buffer A. The collected fraction was concentrated to 10 mg/mL with an Amicon Ultra 10K prior to crystallization. To obtain the selenomethionine-substituted protein, E. coli B834 (DE3) was used as the host strain with the LeMaster medium [10,11] supplemented with thiamine (40 mg/L), and buffer A containing 1 mM and 10 mM DTT was used for the Ni-column purification and protein concentration, respectively.

Crystallization and structure determination
Crystals were grown at 20˚C by the hanging-drop vapor-diffusion method, with a drop consisting of 2.0 μL of protein solution and 2.0 μL of a crystallization reservoir solution containing 0.1 M sodium acetate pH 4. 8-5.4, and 12-35% polyethylene glycol (PEG) 3350. Under these conditions, both native and Se-Met crystals were obtained, and these belonged to the space group P2 1 2 1 2 1 . Prior to data collection, the crystal was transferred for 1 min into a cryoprotectant solution [35%(w/v) PEG3350 in the reservoir solution], and flash-cooled in a liquidnitrogen gas stream at 95 K. Diffraction data were collected on beamline 5A equipped with a Quantum 315 CCD detector and beamline AR BL-NW12A equipped with a Quantum 210 CCD detector (Area Detector Systems Corp., CA) at the Photon Factory, High Energy Accelerator Research Organization, Tsukuba, Japan. The diffraction data were processed using the HKL2000 package [12].
The initial phase was determined by the Se-SAD method and the structure of Se-Met Plscyllo-IDH was solved using AutoSol in Phenix [13]. The native apo-structure was then solved by the molecular replacement method, using the MORLEP software [14] in the CCP4 suite [15] with the Se-Met-structure as the search model. The native apo-structure was then used as a search model for other Pl-scyllo-IDH structures in complexes with substrates. Iterative model refinements were carried out using Coot [16] and REFMAC5 [17]. The coordinates of L-glucono-1,5-lactone were prepared using Discovery Studio (BIOVIA), and the topology file was produced by the PRODRG server [18]. The data collection and refinement statistics are presented in Table 1.
For determination of Pl-scyllo-IDH structures in complexes with substrates, crystals were soaked for one minute with 25 mM of the oxidized form of nicotinamide adenine dinucleotide, (NAD + ) (Sigma-Aldrich Co., MO) and 100 mM substrate in the reservoir solution. L-Glucose and myo-inositol were purchased from Sigma-Aldrich, and scyllo-inosose was from the Hokko Chemical Industry Co., Ltd. (Tokyo, Japan), respectively. Figures for all protein structures were generated using PyMOL (Schrödinger, Inc., MA). The atomic coordinates and structure factors for the apo, L-glucono-1,5-lactone/NADH, myo-inositol/NAD + and scylloinosose/NAD + complexes have been deposited in the Protein Data Bank under the accession codes 5YAB, 5YAP, 5YA8 and 5YAQ, respectively.

Mutant preparation
Mutants of Pl-scyllo-IDH containing single amino acid substitutions with alanine (K106A, D191A, H195A, R178A, and H318A) were prepared using a PrimeSTAR mutagenesis kit (Takara Bio, Shiga, Japan) according to the manufacturer's protocol, with pET21a(+)-lgdA as the template. The primers used for construction of each mutant genes are shown in Table 2. Mutant enzymes were expressed and purified as described for the wild-type enzyme.

Enzyme assay of Pl-scyllo-IDH
Purified proteins were assayed for enzyme activity. The reaction mixture comprised of 100 mM Tris-HCl, pH 9.0, 20 mM NAD + , and a substrate in a total volume of 900 μL. The concentrations of substrate were 12.5, 25, 50, and 100 mM. Fifty microliter of 0.2 mg mL -1 enzyme solution was added, and the activities were measured by monitoring absorption at 340 nm using a V-750 spectrophotometer (Jasco corp., Tokyo, Japan). Kinetic parameters were calculated by the Michaelis-Menten equation using the software supplied with the spectrophotometer. Kinetic parameter of Pl-scyllo-IDH for L-epi-2-inosose (Hokko Chemical Industry, Co., Ltd.) was measured in 190 μL reaction mixture comprised 100 mM MES-KOH, pH6.0, 200 mM NADH, and various concentration of L-epi-2-inosose: 1.25, 2.5, 5, 10, 20, and 40 mM at 25˚C. Reactions were started by addition of 10 μL of 42 μg mL -1 enzyme solution, and the activities were measured by monitoring the decrease in absorption at 340 nm by using DU800 spectrophotometer (Beckman Coulter, Inc., CA). Data obtained were fit to the Michaelis-Menten equation using the non-linear regression tool of Origin 6.0 (Light Stone, Tokyo, Japan).

Phylogenetic analysis
The data set comprised of 44 sequences from the PDB, annotated by Pfam as being GFO/IDH/ MocA family members. The tree was constructed by the neighbor-joining method with the MEGA7 programs [19]. Bootstrap values were calculated from 1000 repeats.

Overall structure of LGDH
We have solved four different Pl-scyllo-IDH structures, which are the apo-form, a complex with L-glucono-1,5-lactone (described below), a complex with myo-inositol, and a complex with scyllo-inosose. All the crystals belonged to the space group P2 1 2 1 2 1 with unit cell parameters of a = 91.0, b = 127.5, c = 137.5 Å for the apo-structure and a = 86.6, b = 121.2, c = 140.9 Å for the complex structures. The apo-Pl-scyllo-IDH structure was refined at 1.75 Å, and the other complex structures ranged from 1.8-2.3 Å, depending on the substrate.
Pl-scyllo-IDH formed a homo-tetrameric structure in the asymmetric unit (Fig 2A), and this structure was considered the biologically relevant form, since a similar unit has been observed for myo-IDH from B. subtilis [8] and GFOR from Zymmomonas mobilis [6].
As was observed for the other GFO/IDH/MocA structures, one Pl-scyllo-IDH subunit contained two domains ( Fig 2B). The N-terminal domain adopted a Rossmann fold motif for the The lower case letters, f and r, in primers names denote forward and reverse, respectively. Nucleotide sequences are presented as the 5' to 3' direction. https://doi.org/10.1371/journal.pone.0198010.t002 binding of NAD + , and the C-terminal domain bound substrate. The C-terminal domain contained a large β-sheet composed of 8 parallel and anti-parallel β-strands.
A long loop extending from G303 -G321 was found here to be a characteristic feature of Pl-scyllo-IDH (Fig 2A), which has not been found in other known myo-IDHs. This loop is involved in subunit interactions, as well as substrate recognition in the adjacent subunit, as described below. Other interactions involved in oligomer formation were mainly mediated by the 8-stranded β-sheet in the C-terminal domain.
A 3D structure comparison using the Dali server [20] revealed that the top hit in PDB was 4FB5 (Z = 52.0, rmsd = 1.1 Å) followed by 4H3V, 4GQA, 3V5N, 3DTY, and 2GLX, as well as others. The top five structures are biochemically uncharacterized, but commonly have loops extending to the active site in the adjacent subunit, as was seen here in Pl-scyllo-IDH. On the other hand, the hit structures listed below of 3DTY do not possess such a loop structure. 2GLX is 1,5-anhydro-D-fructose reductase (1,5-AFR) from Sinorhizobium morelense [9], followed by other IDHs mainly annotated by amino acid sequence homology. GFOR from Zymmomonas mobilis [6] was listed at the 13 th place in the non-redundant list with a Z-score and rmsd of 35.6 and 2 Å, respectively. myo-IDH from B. subtilis, whose activity has been biochemically confirmed [21], was listed at the 43 rd place with a Z-score and rmsd of 30.4 and 2.7 Å, respectively.

L-glucose recognition by Pl-scyllo-IDH
To elucidate the L-glucose recognition mechanism, the crystal was soaked with L-glucose and the resulting structure was determined. The structure was refined at 1.8 Å, and the electron density of the bound sugar molecule was clearly observed. Among the four subunits, apparent electron density due to the substrate was observed in only three of the subunits. This The overall structure of Pl-scyllo-IDH is shown as ribbon model. The structure adopts a homotetrameric conformation. In the structure, L-glucono-1,5-lactone and NADH as shown as sphere and stick models, respectively. A loop region colored in red in the yellow colored subunit protrudes to the active site in the next grey colored subunit, circled with a blue dotted line, described in detail in the text. (B) Subunit structure of Pl-scyllo-IDH shown as a ribbon model. Colors are assigned based on the secondary structure. NADHbinding and substrate-binding domains are colored yellow with red, and magenta with cyan, respectively. L-glucono-1,5-lactone and NADH are shown as stick models colored in blue. The loop region extending to the next subunit is marked with a blue dotted circle, which also corresponds to the circle in panel A.
https://doi.org/10.1371/journal.pone.0198010.g002 observation was also seen in the two complex structures containing inositols, as described in the next section. The molecule in the active site fit in the density with a correct orientation, since L-glucose has a C6-hydroxyl group, making it easy to locate the positions of each hydroxyl group. As a result, the C1-hydroxyl group was seen to come into close contact (2.9 Å distance) to the Nz atom of K106, which is likely to be one of the catalytic residues. This is in accordance with the result obtained by NMR analysis, which revealed L-gluconate, which was considered to be a compound spontaneously hydrolyzed from L-glucono-1,5-lactone [4].
Furthermore, based on the Fo-Fc omit map of the bound molecule, its C1, O1, C2 and O5 atoms were located on an identical plane, and the C1-hydroxyl group adopted neither an equatorial nor an axial conformation. We therefore concluded that the bound molecule in the active site was in fact the reaction product of L-glucose oxidation, namely L-glucono-1,5-lactone, despite the fact that the crystal had been soaked with L-glucose for data collection ( Fig  3A). Accordingly, although we added NAD + for crystallization, the bound cofactor molecule might therefore be NADH. The plane of the nicotinamide ring was parallel to the plane of the lactone ring, and C4 atom of the nicotinamide ring was located just above the C1 atom of the lactone at a distance of 3.4 Å.
As for the binding mode of the lactone in the active site, hydrogen bonds were formed from O1 through O4 with K106 (O1 at 2.9 Å; distance to a counter-part oxygen), Y135 (O2 at 2.8 Å), Y163 (O3 at 3.1 Å), E165 (O3 at 2.5 and O4 at 2.7 Å), H195 (O1 at 2.9 Å), and H318 (O3 at 2.6 Å) (Fig 3A). The substrate was also sandwiched between F17 and L192 through hydrophobic interactions. Among these residues, H318 is located in the loop region of the adjacent subunit, which is a characteristic of the Pl-scyllo-IDH structure compared to other IDHs.
As Pl-scyllo-IDH is a member of GFO/IDH/MocA family, it contains a conserved EKP motif (Fig 4). The residue K106 is in the middle of this motif, and forms a hydrogen bond to a carbonyl oxygen in the lactone ring. Among the seven residues described above, K106 and D191 are conserved in the GFO/IDH/MocA family, and H195 is conserved in the IDH family (Fig 4).

Inositol derivative complex
Since Pl-scyllo-IDH also oxidizes scyllo-inositol and myo-inositol, the complex structures of Pl-scyllo-IDH with scyllo-inosose and myo-inositol were also determined. Both inositols bound to the catalytic site in essentially the same manner as each other, where residues K106, Y135, Y163, E165, R178, D191, and H318 formed hydrogen bonds with the hydroxyl oxygen of the sugar. With the exception of R178 and D191, these residues were also involved in the binding of L-glucono-1,5-lactone. All six inositol hydroxyl oxygens were involved in hydrogen bonds, including bipartite and tripartite bonds making for a total of eleven bonds ( Fig  3B and 3C). The mode of substrate recognition by R178 was remarkably different between the lactone and the inositols. The loop region, A169 to G185, which is not conserved in the other known myo-IDH structures (Fig 4), showed structural flexibility, and accordingly, R178 in the loop altered its position depending on the substrate. In case of L-glucono-1,5-lactone, R178 was located outside of the active site to avoid the protruding C6-hydroxyl group of the lactone ring. On the other hand, in the case of the inositols, R178 was positioned inside the active site in order to be able to interact with the substrates (Fig 5). Therefore, the flexibility of the position of R178 appears to contribute to the broad substrate specificity of Pl-scyllo-IDH.
In the case of myo-inositol binding, based on the shape of the electron density map, the 4'-OH formed a hydrogen bond with K106 (Fig 3B), which suggests that the reaction product is L-epi-2-inosose (Fig 1). myo-IDH from B. subtilis is known to oxidize 2´-OH of myo-inositol resulting in scyllo-inosose production (Fig 1) [21]. Therefore, Pl-scyllo-IDH may have a different regioselectivity with myo-IDH toward myo-inositol. Then we performed a reverse reaction assay of Pl-scyllo-IDH using L-epi-2-inosose as a substrate. As shown in Table 3, Pl-scyllo-IDH exhibited NADH-dependent reductase activity for L-epi-2-inosose. These values are comparable to other substrates studied previously with Pl-scyllo-IDH (Table 3). Therefore, L-epi-2-inosose can be a substrate for Pl-scyllo-IDH, and the enzyme is highly possible to be the first enzyme catalyzing reversible oxidation of myo-inositol to L-epi-2-inosose.
In the case of scyllo-inosose bound to Pl-scyllo-IDH, the position of the carbonyl oxygen was determined based on the electron density of the Fo-Fc omit map (Fig 3C), and the carbonyl oxygen located between K106 and H195 formed hydrogen bonds, with distances of 3.2 Å and 2.9 Å, to K106 and H195, respectively (Fig 3C). The complex structure with scylloinosose has been reported for myo-IDH from B. subtilis. In this complex, D172 (corresponding to D192 in Pl-scyllo-IDH) is not involved in the binding of scyllo-inosose, since its side chain is parallel to the plane of the sugar ring, as has been observed in the Pl-scyllo-IDH structure complexed with the lactone. On the other hand, in the Pl-scyllo-IDH structure complexed with scyllo-inosose, D192 formed hydrogen bonds to the hydroxyl groups of scyllo-inosose, as was observed in the case of the myo-inositol complex of Pl-scyllo-IDH (Fig 3C).
Pl-scyllo-IDH has a lower K m value for scyllo-inositol than that for myo-inositol [4], although the k cat values are almost same. All six of the hydroxyl groups in scyllo-inositol adopted an equatorial conformation; on the other hand, one hydroxyl group in myo-inositol adopted an axial conformation (Fig 1). Therefore, scyllo-inositol can be accepted in the catalytic site of Pl-scyllo-IDH in any direction, and this may contribute to its lower K m .

Enzymatic activity
A catalytic triad consisting of K-D-H has been identified in the inositol dehydrogenase family [8,9,22,23]. Based on the crystal structure of Pl-scyllo-IDH complexed with inositols, side chains of D191 and H195 were found to be oriented to K106; on the other hand, in the structure complexed with L-glucono-1,5-lactone, the side chain of D191 was parallel to the sugar plane and extended outward from the active site. Therefore, to confirm whether these conserved residues in Pl-scyllo-IDH are important for the catalytic activity, we prepared single amino acid mutants of 3 residues substituted to alanines, which resulted in a loss of  Table 3. Kinetic parameters of Pl-scyllo-IDH with L-2-epi-inosose and substrates.

Substrate
K m (mM) k cat (min -1 ) k cat /K m (min -1 mM -1 ) enzyme activity (data not shown) as expected. Therefore, in the complex with lactone, a water molecule between K106 and D191 may be involved in a proton relay system that is part of the catalytic mechanism, where distances were 3.1 Å (K to the water molecule) and 3.5 Å (D to the water molecule) in chain A, 2.9 Å and 2.8 Å in chain B, and 3.3 Å and 3.0 Å in chain C, respectively. Based on the substrate recognition modes revealed by the crystal structures, in addition to the three proposed catalytic residues, R178 and H318 were unique residues compared to residues in other known IDHs. Therefore, we also generated the single-amino acid substituted mutants R178A and H318A. The results revealed that both mutations showed different effects (Table 4). R178A increased the Km value by 10-fold against wild type and decreased the kcat value for scyllo-inositol as a substrate. Since R178 bound to both scyllo-inositol and NADH (Fig 5), the alanine mutation lost these connections resulting in lower activity. Therefore, R178 is necessary for the full activity with scyllo-inositol as a substrate. On the other hand, when Lglucose was used as a substrate, Km value increased, and kcat value also increased by approximately 5-fold against wild type. Although R178 formed no interactions with either L-glucose or NADH on the basis of the crystal structure (Fig 5), the result may suggest some interactions exist between R178 and L-glucose and/or NAD + in solution. If R178 could interfere with the C6 group of L-glucose, the product release from the active site may faster in R178A mutant. In case of H318A mutant, the kcat value for scyllo-inositol decreased resulting in a very weak activity. Furthermore, when L-glucose was used as a substrate, the activity was so low that we were unable to obtain kinetics parameters. H318 is located in the loop from the next subunit and is involved in the substrate binding for both scyllo-inositol and L-glucose. Based on the crystal structures complexed with substrates, H318 appeared not to be critical in the substrate binding, as other two residues, Y163 and E165, formed hydrogen bonds with the same hydroxyl oxygen of substrates as H318 (Fig 3). Therefore, further analyses are needed to understand these results.
Interestingly, in the apo structure, an acetate molecule originating from the crystallization solution was also found to be bound in the active site. This molecule formed hydrogen bonds with Y135, Y165, E195, and H318, which were the same residues used for substrate binding, other than the three catalytic residues, which suggests that these 4 residues may only be required for substrate binding, although the role of H318 remains unclear as described above.
As for hydride transfer of the C1 hydrogen to NAD + , in the complex of Pl-scyllo-IDH with the lactone, the nicotinamide ring of NADH was found to be stacked on top of the lactone ring, where the C4 atom of the nicotinamide ring of NADH was aligned over the C1 atom of the lactone at a distance of 3.4 Å (Fig 6A). With this conformation, the C1 hydrogen atom at the axial position in β-L-glucose would be transferred to the R side of the C4 atom of NAD + , therefore, Pl-scyllo-IDH is considered an A-type enzyme. The corresponding distances in the myo-inositol and scyllo-inosose complexes were 2.8 Å and 3.1 Å, respectively, and the relative positions of NAD + and the inositols were the same as those seen in the lactone complex. Alizade et al. previously demonstrated, using an isotope labeling experiment, that myo-IDH from Klebsiella pneumoniae is an A-type enzyme [24], whose structure is unknown. On the other hand, van Straaten et al. reported that myo-IDH from B. subtilis is a B-type enzyme similar to SDR proteins, which use pro-S hydrogen, based on the crystal structure in complex with scyllo-inosose [8] (Fig 6B). Therefore, this may also be one of the characteristics that distinguish Pl-scyllo-IDH from other known myo-IDHs.

Phylogenetic study
When the amino acid sequence of Pl-scyllo-IDH was subjected to a BLAST search, we obtained a list of myo-IDHs, including one from B. subtilis. Therefore, we expected that the structure of Pl-scyllo-IDH would be similar to that of myo-IDH from B. subitlis, which is characterized in detail [23,26]. However, the crystal structure of Pl-scyllo-IDH and that of myo-IDH from B. subtilis mainly differ in the mode of subunit interactions (Fig 2). Furthermore, the overall structure of Pl-scyllo-IDH was rather similar to GFOR from Zymmomonas mobilis. Therefore, structures deposited in PDB, which were categorized as containing the Pfam_GFO/IDH/ MocA domain, were used to perform a phylogenetic analysis using the MEGA7 software. As shown in Fig 7, the Pl-scyllo-IDH and PDB entries 4FB5, 4GQA, 4H3V, 3DTY and 3V5N, which were the top hits from the Dali search, were found to be in the same clade. With the exception of 4GQA, all of these structures have the same swapped loop like Pl-scyllo-IDH. 4GQA also has a loop region similar to that seen in Pl-scyllo-IDH, but it does not loop out to interact with the active site in the adjacent subunit. On the other hand, myo-IDHs, including one (3NT2) from B. subtilis, were clustered into a clade different from that of Pl-scyllo-IDH.
Most of the structures listed in Fig 7 have been deposited in PDB without an accompanying publication, therefore, it is difficult to judge whether the phylogenetic tree corresponds to substrate specificity. However, based on our data examining Pl-scyllo-IDH substrate specificity, showing that it is a scyllo-inositol dehydrogenase with a broad substrate specificity, and the phylogenetic analysis, Pl-scyllo-IDH could represent a different subfamily distinct from that of known myo-IDHs. shown as stick models. β-L-glucose, colored in cyan, is superposed on the lactone molecule, and the H1 hydrogen in the axial conformation is colored white. The distance between NC4 and C1 is 3.4 Å, and the distance between H1 and NC4 is 2.0 Å shown as a red dotted line. From this structure, Pl-scyllo-IDH is concluded to be an A-type enzyme. (B) Structures of each of chain B in Pl-scyllo-IDH complexed with L-glucono-1,5-lactone colored in green and pale green, and myo-IDH from B. subtilis complexed with scyllo-inosose (PDB code: 3NT5) colored in orange and pale orange were superposed using the program Chimera [25]. The superposed active sites are shown with substrates, NADH, and the catalytic triad, K-D-H in stick models. The difference is observed in the orientations of the nicotinamide rings between Pl-scyllo-IDH and myo-IDH, which represent the A-type and B-type enzymes, respectively. https://doi.org/10.1371/journal.pone.0198010.g006