Inositol dehydrogenase from Bacillus subtilis (BsIDH) is a NAD+-dependent enzyme that catalyses the oxidation of the axial hydroxy group of myo-inositol to form scyllo-inosose. We have determined the crystal structures of wild-type BsIDH and of the inactive K97V mutant in apo-, holo- and ternary complexes with inositol and inosose. BsIDH is a tetramer, with a novel arrangement consisting of two long continuous β-sheets, formed from all four monomers, in which the two central strands are crossed over to form the core of the tetramer. Each subunit in the tetramer consists of two domains: an N-terminal Rossmann fold domain containing the cofactor-binding site, and a C-terminal domain containing the inositol-binding site. Structural analysis allowed us to determine residues important in cofactor and substrate binding. Lys97, Asp172 and His176 are the catalytic triad involved in the catalytic mechanism of BsIDH, similar to what has been proposed for related enzymes and short-chain dehydrogenases. Furthermore, a conformational change in the nicotinamide ring was observed in some ternary complexes, suggesting hydride transfer to the si-face of NAD+. Finally, comparison of the structure and sequence of BsIDH with other putative inositol dehydrogenases allowed us to differentiate these enzymes into four subfamilies based on six consensus sequence motifs defining the cofactor- and substrate-binding sites.
myo-Inositol is ubiquitous in soil and plants. Inositol derivatives and isomers are found in bacteria, plants, animals and fungi. Many soil and plant bacteria such as Bacillus subtilis, Corynebacterium glutamicum, Aerobacter (Klebsiella) aerogenes, Rhizobium leguminosarum bv. viciae, Sinorhizobium meliloti, Sinorhizobium fredii, Salmonella enterica serovar Typhimurium and Lactobacillus casei strain BL23 [1–7] have been reported to use myo-inositol as the sole carbon source. The molecular genetics and biochemistry of inositol catabolism in bacteria have been described in greatest detail for B. subtilis [1,8,9]. The iol divergon, containing the operons iolABCDEFGHIJ, iolRS and the gene iolT, was shown to be responsible for myo-inositol degradation [1,8,10]. Iol gene clusters have been reported in a number of bacterial species [2,4,11,12].
IDH (inositol dehydrogenase) (EC 22.214.171.124) is responsible for the first step in myo-inositol degradation. IDH is encoded by iolG in B. subtilis, and idhA in S. meliloti and S. fredii. Multiple genes considered to encode myo-IDH have been found in, for instance, L. casei strain BL23 (iolG1 and iolG2), Lactobacillus plantarum strain WCFS1 (iolG1–iolG4), S. enterica serovar Typhimurium (iolG1 and iolG2) and C. glutamicum (iolG1–iolG3) [4,11]. It is not known whether these multiple iolG genes have redundant functions or possess different substrate affinities towards inositol isomers or related molecules.
Only one iolG gene has been identified and characterized for B. subtilis. IDH from B. subtilis (BsIDH) is an NAD+-dependent enzyme that catalyses the oxidation of the axial hydroxy group of myo-inositol to form scyllo-inosose, as shown in 1 . The enzyme shows a broad substrate spectrum while remaining highly stereoselective [13,14]. BsIDH is able to oxidize the monosaccharides α-D-glucose and α-D-xylose, but not β-D-glucose, D-mannose and D-galactose. Also, scyllo-inositol, the all-equatorial stereoisomer of myo-inositol, is not a substrate nor an inhibitor for BsIDH . Recently, it was shown that BsIDH can also act on D-chiro-inositol . Pinitol (3-O-methyl-D-chiro-inositol), an inositol-related compound found in soya beans, appeared to be an alternative substrate . BsIDH is a member of the GFO (glucose–fructose oxidoreductase)/IDH/MocA family, a group of homologous dehydrogenases. BsIDH has a molecular mass of 39.17 kDa. The enzyme is active as a tetramer in solution and has a apparent molecular mass of 160 kDa . Enzyme kinetics revealed that the enzymatic reaction follows a Bi Bi mechanism, with NAD+ binding first, then inositol, with subsequent sequential product release (inosose followed by NADH) [15,18]. In order to understand the molecular details of substrate recognition and catalysis, we have initiated a structural study with and without substrates and products.
Reaction catalysed by IDH
The present paper describes the first structural characterization of a myo-IDH. The work described allowed us to identify residues important for substrate binding and catalysis. Furthermore, implications of catalytic mechanism and subfamily classification are discussed.
MATERIALS AND METHODS
Apo-BsIDH and apo-K97V BsIDH were purified and crystallized as described previously . Briefly, 4 μl drops, containing 2 μl of protein solution (10 mg/ml in 25 mM Tris/HCl, pH 8.0) were mixed with 2 μl of crystallization solution and overlaid with paraffin oil to prevent evaporation of water from the drops. Crystals were flash-cooled in liquid nitrogen after cryoprotection with 25% glycerol.
Crystals of BsIDH with NAD(H) were obtained by the microbatch method from crystallization solutions containing salt and ammonium sulfate. Before crystallization, BsIDH (10 mg/ml in 25 mM Tris/HCl, pH 8.0) was incubated with 5 mM cofactor. The best crystals were obtained with 0.1–0.2 M trisodium citrate (pH 5.4) and 1.6–2.9 M ammonium sulfate. Cubic crystals appeared within 1 week. These crystals are typically 0.15 mm×0.15 mm×0.15 mm and belong to space group I213 with two monomers per asymmetric unit. Crystals of the K97V BsIDH mutant with NAD+ were obtained in the same way. These holo-BsIDH crystals were flash-cooled in liquid nitrogen after cryoprotection with 25% ethylene glycol.
BsIDH ternary complex crystallization
Holo-BsIDH crystals (IDH with cofactor bound) were soaked in mother liquor containing 0.1 M trisodium citrate (pH 5.4), 2.6 M ammonium sulfate and either inositol or inosose (4 mg/100 μl of mother liquid). Before flash-cooling, the crystals were transferred to mother liquid containing 25% ethylene glycol and either inositol or inosose.
Data collection and processing
Diffraction data were collected at 100 K from single crystals. Native BsIDH, SeMet (selenomethionine)-BsIDH and complex BsIDH datasets were collected at beamline 08-ID-1 on a MAR CCD225 detector at the CLS (Canadian Light Source, Saskatoon, SK, Canada). For phasing, three-wavelength Se-MAD (multi-wavelength anomalous detection) data to 1.8 Å (1 Å=0.1 nm) were collected from a single SeMet-BsIDH crystal as described previously . A high-resolution 1.5 Å apo-BsIDH dataset was collected from a single SeMet-BsIDH crystal. All diffraction data were processed and scaled with D*TREK . Data collection statistics are shown in Supplementary Table S1 at http://www.BiochemJ.org/bj/432/bj4320237add.htm.
Model building and refinement
The 1.8 Å apo-BsIDH structure was solved using experimental-phase information derived from MAD data collected at the Se-edge. The electron-density map obtained from the experimental phases was of excellent quality and initial building was carried out automatically using ARP/Warp . Model corrections were carried out in Coot . The model was refined using Refmac5  to 1.8 Å resolution, R/Rfree were 17.6 and 21.9% respectively. This model was refined further against the 1.5 Å resolution diffraction data using Refmac5 . The final apo-BsIDH model has an R value of 16.3% and an Rfree value of 19.5% and consists of residues 1–337 and contains three chlorides, two triethylene glycol, one magnesium, one glycerol and water and shows good model geometry.
The structure of apo-BsIDH was used as the starting model for the refinement of the other structures. First, all solvent molecules and ions were removed from the apo-BsIDH model. Molrep  was used to find molecular replacement solutions, using the 1.5 Å apo-BsIDH structure as the search model. The refinement of these structures was done with Phenix . First, rigid-body refinement was carried out to fine-tune the position of the molecule(s) in the asymmetric unit, followed by simulated annealing using Cartesian dynamics at 5000 K to remove model bias. Clear positive density was present in the holo and complex Fo−Fc electron-difference maps contoured at the 3σ level for cofactor and bound carbohydrate molecules. The model was refined further using restrained refinement. NCS (Non-Crystallographic Symmetry) restraints were used throughout the refinement for the holo and the complex models. Model corrections and manual placement of cofactor and carbohydrates were performed using Coot . Libraries for cofactor, inosose and inositol were generated with ELBOW in Phenix . The refinement progress was monitored by following Rfree and inspecting the electron-density maps. When Rfree fell below 30%, water molecules were added using the water picking refinement in Phenix , and their positions were checked manually using Coot. Final refinement statistics are shown in Supplementary Table S2 at http://www.BiochemJ.org/bj/432/bj4320237add.htm. The co-ordinates and structure factors for each structure have been deposited in the PDB and the codes are included in Supplementary Table S2.
The programs DALI , SCOP  and the sequence search option from RCSB against known structures in the PDB (http://deposit.pdb.org/validate/) were used to search for structurally similar and functionally related proteins. Superposition of structurally related proteins and of all solved IDH models was carried out in Coot  based on the C-terminal domain (residues 124–324). Multiple sequence alignment was performed using ClustalW. Structure-based sequence alignment was performed with SEQUOIA . All models were validated using MOLPROBITY  and the ADIT validation server from RCSB (http://deposit.pdb.org/validate/). Figures were prepared using PyMOL (http://www.pymol.org).
RESULTS AND DISCUSSION
Crystallization and crystal packing analysis
Apo-BsIDH crystallizes with one molecule per asymmetric unit. These crystals belong to space group I222. Together with symmetry-related molecules, they form a tetramer with perfect 222 symmetry. Co-crystallization or soaking of these crystals with cofactor or substrate was unsuccessful. Analysis of the crystal packing revealed that part of the cofactor-binding site was occupied by residues from symmetry-related molecules, thus preventing binding of the cofactor. Holo-BsIDH crystallizes under different conditions. These crystals belong to space group I213 with two monomers per asymmetric unit. Together with symmetry-related molecules, they too form a tetramer with almost perfect 222 symmetry.
Overall ternary and quaternary structure of BsIDH
Monomer BsIDH consists of two structural domains (Figure 1a). The N-terminal domain (residues 1–124) has the typical α/β nucleotide-binding (Rossmann fold) motif , comprising two βαβαβ motifs stacked together to form a single six-stranded parallel β-sheet, flanked on both side by α-helices. Helix α10 of the C-terminal domain completes the Rossmann fold. The C-terminal domain (residues 125–337) shows structural similarity to members of the glyceraldehyde 3-phosphate superfamily in SCOP [27,31]. The domain contains a mixed parallel/antiparallel six-stranded β-sheet (β7–β12) with strand order 234156, a ‘crossed-over’ two-stranded antiparallel β-sheet (strands β13 and β14) and six α-helices (α6–α11). α-Helix 6 interacts with β-strand 12 (Figure 1a). The first β-strand, β7, which is found in the centre of the β-sheet forms a ‘ψ-structure motif’ with strands β10 and β11. ψ-Structure motifs are characterized by two antiparallel β-strands connected by a loop (ψ-loop) with a third strand inserted between them, thus resembling the Greek letter ψ. Main-chain oxygen atoms of Asn157 and Pro158 of the N-terminus of strand β7 form hydrogen bonds with main-chain amides of the N-terminus of strand β10 and the ψ-loop, and anchor β7 to the ψ-loop. The β-sheet topology in BsIDH differs when compared with members of the glyceraldehyde 3-phosphate superfamily in that the last two strands (β13 and β14) are not part of the main C-terminal β-sheet, but instead cross-over to form the tetramer core (Figure 1). The C-terminal domain is ‘open-faced’  in that the helices and loops are all on one side of the β-sheet. The inside of the β-sheet is mainly hydrophobic, whereas on the outside there are mainly hydrophilic and charged residues.
Crystal structure of BsIDH
The biologically active form of BsIDH is a tetramer . BsIDH also crystallizes in a tetramer arrangement (Figure 1b), with dimensions of approx. 106 Å×76 Å×40 Å. Extensive interface contacts exist between the four subunits in the tetramer. The main interface contacts are constituted by the C-terminal β-sheets of the four subunits forming two extended 16-stranded β-sheets. The β-sheet has a very pronounced right-handed twist and turns almost 180 ° over the whole length of the molecule. The two β-sheets pack against each other in a face-to-face fashion, creating an extensive interface along the entire core of the tetramer. The dimer contacts between subunits A and C and subunits B and D are mainly hydrogen bonds and salt bridges from the C-terminus of β8 and the subsequent loop, the ψ-structure motif and the loop connecting β11 and β12, resulting in a buried surface area of 13645 Å2. Interactions between subunits A and B and subunits C and D are formed between the C-terminus of helix α1 and the loop connecting helix α1 and β-strand β2 (residues 21–26) of the N-terminal domain, and between β-strands β12 and β13 and the loop connecting β12 and β13 of the C-terminal domains, with a buried surface area of 990 Å2. The small ‘crossed-over’ β-sheet (β13 and β14) packs against β-strand β12 of the other subunit, completing the classical glyceraldehyde 3-phosphate superfamily β-sheet topology. Additional interface interactions are made between the crossed-over β-strands β13 and β14 and the ψ-loop of subunits A and D and subunits B and C, burying 720 Å2 of surface area. Thus the loop connecting β-strands β12 and β13 and the crossed-over β-strands β13 and β14 each contact three subunits in the centre of the tetramer.
Structural and functional comparison with GFO/IDH/MocA oxidoreductases
A search using the DALI server revealed that the closest structurally related enzymes all belong to the GFO/IDH/MocA oxidoreductase family  (Table 1). According to their PFAM  classification, all enzymes belonging to the GFO/IDH/MocA oxidoreductase family contain two domains: an N-terminal nucleotide-binding domain [GFOR_IDH_MocA (PF01408)] and a C-terminal domain [GFOR_IDH_MocA_C (PF02894)], and are classified in the GFO/IDH/MocA family in UniProtKB/Swiss Prot . The closest ‘hits’ are annotated as IDH enzymes, followed by other members of the glyceraldehyde-3-phosphate dehydrogenase-like family. More distantly related enzymes are members of the SDR (short-chain dehydrogenase/reductase) family . This latter family of enzymes, also containing the Rossmann fold but otherwise structurally distinct from BsIDH, catalyse a diverse set of enzymatic reactions using a conserved nucleotide-binding domain and a SYKN (Ser-Tyr-Lys-Asn) catalytic tetrad [37,38]. All enzymes use either NAD+ or NADP+ as a cofactor. Table 1 shows the top six DALI hits, the putative myo-IDHs (EC 126.96.36.199) from Salmonella Typhimurium LT2 (StIDH), Thermotoga maritima (TmIDH), L. plantarum (LpIolG1) and C. glutamicum (CgIDH). These structures have been deposited by structural genomics consortia with little annotation. The next two DALI solutions are AFR (1,5-anhydro-D-fructose reductase) from Sinorhizobium morelense  and GFOR (glucose–fructose oxidoreductase) from Zymomonas mobilis . AFR catalyses the NADPH-dependent reduction of 1,5-anhydro-D-fructose to 1,5-anhydro-D-mannitol [39,40]. GFOR is a NADP(H)-dependent homotetrameric enzyme which catalyses the oxidation of glucose to gluconolactone and the reduction of fructose to sorbitol . The reaction catalysed by AFR and the first half-reaction of GFOR are very similar to the reaction catalysed by BsIDH.
|PDB code||Z-score||RMSD*||SID (%)†|
|PDB code||Z-score||RMSD*||SID (%)†|
RMSD of superimposed Cα atoms (Å).
Percentage of sequence identity over equivalent positions.
A structure-based sequence alignment of these enzymes shows that their secondary-structural elements are all well conserved (Supplementary Figure S1 at http://www.BiochemJ.org/bj/432/bj4320237add.htm). They all contain an N-terminal nucleotide-binding domain and a C-terminal domain, with a mixed parallel/antiparallel β-sheet. A structural comparison of BsIDH with these structurally related enzymes reveals that BsIDH is most similar to the putative myo-IDHs and in particular to the putative myo-IDHs from Salmonella Typhimurium (StIDH; PDB code 3EC7) (Table 1). BsIDH shows 50% sequence identity with StIDH, and 336 out of 337 Cα atoms overlap with an RMSD (root mean square deviation) of 1.0. The overall fold topology of the N- and C-terminal domains of BsIDH are identical with that of StIDH (see Figures 4a and 4b). The C-terminal domains of BsIDH and StIDH differ slightly from the other three IDH-related enzymes. LpIolG1 (PDB code 3CEA), TmIDH (PDB code 3EZY) and CgIDH (PDB code 3EUW) each contain an additional α-helix after strands β7 and β8 (see Supplementary Figure S1 and Figure 4). Furthermore, the loop connecting strands β8 and β9 in BsIDH and StIDH has a different orientation compared with that of LpIolG1, TmIDH and CgIDH. All of the enzymes crystallize as tetramers. Structural comparison shows the same association of the subunits in the tetramer in all of these IDH-related enzymes.
Although AFR (PDB code 2GLX) and GFOR (PDB code 1OFG) share almost the same overall fold topology as that of BsIDH (Figures 1a–1c) and catalyse a similar reaction, they differ in nucleotide recognition, and, in the C-terminal domain, they differ in their substrate-binding pocket and relative orientation of strands β13 and β14 and the loop connecting them. BsIDH and closely related enzymes use NAD(H) as cofactor, whereas AFR and GFOR use NADP(H). All contain an active-site pocket in the same location, at the ψ-structure (see below). Differences in the geometry of the binding pockets of AFR, GFOR and BsIDH are due to insertions and side-chain alterations in AFR and GFOR and are presumably due to differences in their substrate preferences and specificity. In the IDH-annotated enzymes, β-strands β13 and β14 are not part of the main β-sheet as compared with the same strands in AFR and GFOR (Figures 1c and 1d). The contacts within the oligomers of these enzymes are formed by the C-terminal β-sheet. AFR exists as a dimer , whereas GFOR exists as a tetramer . Although GFOR has the same overall tetramer arrangement as BsIDH, it lacks the ‘crossed-over’ β-sheet (β13 and β14) found in BsIDH (Figures 1c and 1d). In addition to this, GFOR contains an N-terminal arm (residues 1–31), which is absent from BsIDH and AFR, that forms a tight association with an adjacent subunit and probably helps to establish oligomer stability and tetramer formation [41–43].
The crystal structure of holo-BsIDH was determined to 2.3 Å resolution by molecular replacement using the apo-BsIDH structure as the template. Holo-BsIDH contains two molecules in the asymmetric unit and forms a tetramer with symmetry-related molecules by applying a crystallographic 2-fold axis along the longest dimension of the tetramer. Both molecules have bound cofactor. The electron density allows us to determine unambiguously that NAD+ is bound in molecule A, whereas in molecule B NADH is bound. Superposition of molecule A on molecule B reveals that their N-terminal nucleotide-binding domains are slightly rotated, whereas the C-terminal domains superimpose well (Figure 2a). Molecule A is in a slightly more open conformation than molecule B. This difference is even more pronounced relative to apo-BsIDH (Figure 2a). Upon cofactor binding, the N-terminal domain rotates to allow tighter interactions with the cofactor. Details of the residues involved in cofactor binding are shown in Figure 4(b).
Stereo representation of cofactor binding to BsIDH
The cofactor is clearly defined in the electron-density map and is bound in an elongated fashion with the adenine and nicotinamide group anti to the ribose and pyrophosphate group located at the N-terminus of helix α1 (Figure 2b). On the N-terminus of helix α1, Ile13 partially blocks the cofactor-binding site in apo-BsIDH, but is rotated when NAD(H) is bound. The side chain Cδ and Cγ of Ile13 stack with C5N and C6N of the nicotinamide ring. In holo-BsIDH, NAD+ and NADH form similar hydrogen bonds; however, there are fewer interactions between BsIDH and the nicotinamide group of NADH than between the protein and the nicotinamide of NAD+. The overall binding conformation of NAD(H) to BsIDH is similar to that in StIDH (PDB code 3EC7) and LpIolG1 (PDB code 3CEA) and the binding is similar to how NADP+ binds to GFOR (PDB code 1OFG) and to AFR (PDB code 2GLX). The slight differences between the adenine ribose moiety binding to GFOR and AFR compared with BsIDH are due to the additional phosphate in NADP+.
In BsIDH, the adenine ribose binding in NAD+ and NADH is mediated by hydrogen-bonding to Asp35. The side chain of Asp35 is held in position by bonds formed with main-chain amides from Asn37 and Thr10. A similar interaction with the ribose is made by Asp35 in StIDH. However, in the case of LpIolG1, these interactions are different because of the lack of aspartate at this position (Ala40 is found instead). In LpIolG1, the loop connecting β2 and α2, which is part of the adenine ribose-binding pocket, is slightly shifted and the main-chain amide of Leu41 and the side-chain Oϵ1 of Gln45 are hydrogen-bonded to the ribose. In StIDH, an additional hydrogen bond is made by Arg40 in a manner similar to Gln45 in LpIolG1.
The diphosphate is hydrogen-bonded to main-chain amides of the glycine-rich region between β1 and α1 which contains the GXGXXG consensus sequence motif for NAD(H)-binding proteins [44,45]. One of the β-phosphate oxygens is hydrogen-bonded to a structurally conserved water molecule found in the GXGXXG motif . The negative charge of the diphosphate moiety is stabilized by the positively charged end of the dipole through helix α1. Ribose binding is mediated by hydrogen-bonding to a loop between β5 and α5. This loop contains the functional C95EKP motif , which is part of a conserved motif found in sugar dehydrogenases [40,46]. The 3′-hydroxy group of the nicotinamide ribose is hydrogen-bonded to Nϵ2 of His79. The side chain of His79 is held in position by the main-chain carbonyls of Ser74 and Pro98. Unlike in BsIDH, in StIDH and LpIolG1, the 3′-hydroxy group of the nicotinamide ribose forms an additional hydrogen bond with the side chain of Asn76 and Thr80 respectively.
There are a number of differences in binding observed between the nicotinamide moiety of NADH and the nicotinamide moiety of NAD+. First, there are fewer interactions with the NADH nicotinamide group than with the NAD+ group. The side-chain carboxy Oϵ1 of Glu96 makes a hydrogen bond to the carboxamide group of the nicotinamide ring of NAD+ and is in close contact (2.9 Å) with the carbon atom C2 of the nicotinamide ring, thereby making a CH-O hydrogen bond possible. This CH-O interaction has been suggested previously to be important for cofactor binding [39,47,48]. Additionally, Glu96 is hydrogen-bonded to His17 directly, and via a conserved water molecule to the side-chain carboxylate of Glu283. Glu96 does not interact with the NADH nicotinamide group, instead Glu96 (Oϵ1) now makes a hydrogen bond to the side-chain carboxy group of Glu283 via the conserved water molecule. Secondly, the main-chain carbonyl group of Lys97, of the functional motif CEKP, forms hydrogen bonds to the 2′-hydroxy and 3′-hydroxy groups of the NAD+ nicotinamide ribose, and is close (3.1 Å) to the C2 atom of the nicotinamide. When NADH is bound, Lys97 is hydrogen-bonded to His176. Lastly, the position of Tyr280, which forms a hydrogen bond (through the hydroxy group) to the carboxamide oxygen group of the nicotinamide ring in both subunits, is different.
The ternary complex structures with inositol and with inosose were obtained by soaking holo-BsIDH crystals with solutions of inositol and inosose, and have been refined to 2.5 and 2.9 Å resolution respectively. The models have R-factors of 23.4 and 24.1% and Rfree values of 28.0 and 29.1% respectively (Supplementary Table S2). There are no differences with the protein backbone between the ternary complex structures and holo-BsIDH. The average RMSD is 0.3 Å for all Cα atoms. The substrate-binding pocket is located between the N-terminal domain and the C-terminal domain. Electron density for both inosose and inositol is clearly present in difference maps (contoured at 3σ) (Figures 3a and 3b). The substrate-binding site is formed by the nicotinamide group of NAD(H), the C95EKP motif of the N-terminal domain and by the small α-helix η2 that is part of the consensus sequence motif G124F(M/N)RR(Y/F)D (Supplementary Figure S1). The bottom of the substrate-binding site is formed by the ψ-structure and by the loop connecting β7 and α7. The walls are formed by helices α7 and α8 on one side and by the small helix α9 and the N-terminus of helix α10 on the other side.
Substrate-binding site of BsIDH
The G124F(M/N)RR(Y/F)D consensus motif makes numerous hydrogen bonds with conserved residues. In BsIDH, Gly124 is located on a sharp turn between β6 and η2. Phe125 is situated underneath Lys97 and may contribute, together with Val175, to orientate Lys97 for proper interaction with cofactor and substrate. Met126 points into the active site, whereas Arg127 points into the substrate-binding site and is hydrogen-bonded to Glu177 of α8 and a structurally conserved water molecule. Glu177 forms a hydrogen bond with His155, and Arg128 holds the small helix η2 in position via hydrogen bonds to Asp179 from α8 and Asp304 from α10. Asp304 is also hydrogen-bonded to His182 and Arg194. Asp130 is located at the beginning of α6 and forms a hydrogen bond to Arg275 of α9, anchoring the helices.
In the NADH–inositol complex structure, inositol is located between Trp272 from α9 and Thr173 (Figure 3a) and forms multiple hydrogen bonds with the protein (Figure 3a). The hydrogen bond between the C1 hydroxy group of inositol and Arg127 is through a water molecule, whereas the side chain of Asn157 is rotated 180 ° compared with the unliganded structures to form hydrogen bonds with the C4 and C5 hydroxy groups of inositol. Both Lys97 and His176 form hydrogen bonds with the C2 hydroxy group of inositol. Other residues within 5 Å of the substrate are Tyr164, Met126 and Tyr235. Tyr164 is found on the loop between β7 and α7. Tyr235 is part of the Y233GY motif of the ψ-loop. This motif is likely to be involved in hydrophobic stacking interactions with the substrate ; however, only Tyr235 points into the substrate-binding site. Tyr233 is found in a hydrophobic pocket that is formed by its own ψ-loop, the N-terminus of strand β12 and its connecting loop and by the C-terminus of β13–loop–β14 of a another subunit. This residue probably has a structural role in stabilizing the dimer interface and in stabilizing the ψ-loop via hydrogen bonding to Arg156 from strand β7 and Asp236 from strand β10 of the ‘ψ-structure’. The nicotinamide group of the reduced cofactor has its re-face exposed to the substrate-binding site; however, it is not in position to promote hydride transfer from the C2 hydroxy group of inositol.
In the ternary complex with inosose, inosose makes similar interactions to those of inositol (Figure 3b). The C2-carbonyl is positioned between His176 and Lys97. The nicotinamide group of NADH in molecule B has been rotated about the glycosidic bond, by ~180 ° compared with the nicotinamide group of NAD+ in molecule A (Figure 3C). As a consequence of this rotation, the interactions with the nicotinamide group that are present when NAD+ is found in the binary and tertiary complexes are all lacking. The nicotinamide group now points into the substrate-binding site. The side chain of Trp272 forms a stacking interaction with the re-face of the nicotinamide group, shielding it from the solution. NAD+ is bound in a non-productive binding mode, similar to the NADH conformation in the inositol ternary complex structure, whereas the NADH in molecule B has its si-face exposed to the substrate-binding pocket. Similar cofactor orientation has been observed in other NAD(P)-dependent enzymes such as mannitol 2-dehydrogenase, 6-phosphogluconate dehydrogenase and NAD-malic enzyme, and is suggested to induce conformational changes or cofactor–substrate positioning for catalysis to take place [50–53]. However, in all of the other structures of the GFO/IDH/MOCA family deposited in the PDB, including all other IDHs, the nicotinamide ring is oriented as observed in the absence of substrate.
Lys97 is part of the highly conserved C95EKP motif. The K97V mutation abolishes all dehydrogenase activity, consistent with an important role in the catalytic mechanism (see below). The structure of K97V-BsIDH has been solved with and without cofactor and inositol and has been refined to 1.9 Å (apo-K97VBsIDH), 2.6 Å (holo-K97V-BsIDH) and 2.65 Å (ternary K97V-BsIDH complex) resolution. Superposition of apo-K97V-BsIDH on apo-BsIDH reveals that the structures superpose with an RMSD of 0.2 Å. The peptide bond between Val97 and Pro98 still adopts the cis conformation as observed between Lys97 and Pro98. The electron density for the cofactor, for both the holo- and ternary K97V-BsIDH complexes, allows us to determine unambiguously that NAD+ is bound in both molecule A and molecule B. Because co-crystals of wild-type BsIDH and K97V-BsIDH mutant were obtained under identical conditions, we believe that the reduction of the cofactor as observed above is caused by the enzyme and is not due to the crystallization conditions. Upon binding of the cofactor to K97V-BsIDH, the same N-terminal domain movement is observed as in wild-type BsIHD. NAD+ makes similar interactions with the mutant enzyme as in wild-type BsIDH; however, the ribose and nicotinamide group are slightly tilted compared with NAD+ binding in the wild-type enzyme (Figure 3d). Comparison of the substrate-binding sites between wild-type and K97V-BsIDH reveal few changes in side chain positions and substrate binding. Tyr280 is closer to Trp272 in the mutant, and Trp272 has moved inwards towards the active site, thereby narrowing the substrate-binding site. Also the position of inositol differs (Figure 3d), with the C2-hydroxy group now positioned between His155 and Thr173. The Asn157 side chain does not form the hydrogen bond with the substrate as observed in the ternary wild-type BsIDH–inositol complex and remains in the same orientation as observed in apo-BsIDH. Thus Lys97 probably functions to position both the nicotinamide ring and the substrate in the proper orientation for the enzymatic activity.
BsIDH contains an apparent catalytic triad Lys97, Asp172 and His176 (which aligns with Tyr217 in GFOR) similar to what has been proposed for AFR and other members of the GFO/IDH/MocA family [34,39,40,42]. Active-site mutants H176A, proposed to be the catalytic base that abstracts the hydrogen from the C2 hydroxy group, and D172N showed a marked loss in activity . Additionally, Lys97 and Asp172 may be involved in a proton relay system similar to what has been proposed for ‘classical’ SDRs [37,38,54]. Structural comparison of BsIDH with SDR enzymes reveal that Asp172, Lys97 and His176 are located in a similar spatial positions to asparagine, lysine and the catalytic acid/base tyrosine in SDR [38,53,55]. Abstraction of a proton by His176 would facilitate equatorial hydride transfer from C2 of inositol to C4 of the nicotinamide group of NAD+. His176 is close to Lys97 (3.7 Å) to facilitate proton transfer from Lys97 via a water molecule to Asp172. Lys97 is also found at hydrogen-bonding distance from the C2-hydroxy group of inositol, and may help to positioning the substrate for catalysis. However C4 of the nicotinamide group of NAD(H) is not aligned with the C2 hydrogen to facilitate hydride transfer in the inositol ternary complex or in molecule A of the inosose ternary complex (NAD+). In molecule B of the inosose complex structure, the pro-S hydrogen of the NADH is now close to the C2 group of inosose to allow hydride transfer, suggesting that this orientation represents the productive binding mode (Figures 3b and 3c). This would suggest that BsIDH is a B-specific (pro-S) enzyme, similar to SDR proteins [53,55,56], in contrast with the reported IDH from Klebsiella pneumoniae . In both ternary complex structures, the cofactor and substrate are highly solvent-exposed. This would imply that, in BsIDH, the cofactor, substrate and products can easily access/leave the enzyme. This is consistent with the Bi Bi mechanism determined by enzyme kinetics [15,18]: NAD+ binding precedes inositol binding, followed by sequential product release (inosose followed by NADH).
The present paper is the first report describing the high-resolution structure of a myo-IDH. Our results allow a detailed characterization of the NAD(H)- and substrate-binding site. The N-terminal domain contains the NAD(H)-binding site and the C-terminal domain contains the active site and residues involved in substrate binding and catalysis. As in AFR and GFOR, the substrate-binding pocket of BsIDH is located at the ψ-structure. We expect that other putative myo-IDHs would have active sites at the same location and bind the substrate in a similar way to BsIDH. Also, we expect the geometry of the binding pocket to be the same and substrate-binding residues to be conserved. Alterations in the geometry of the binding pocket due to insertions or deletions and side-chain alterations are likely to be related to differences in substrate preferences, specificity and mechanism.
The binding pocket of BsIDH and StIDH and residues involved in substrate binding and catalysis are highly conserved. This supports the annotation of StIDH as a myo-IDH. Conversely, TmIDH, CgIDH and LpIolG1 all differ significantly from BsIDH with respect to their binding pocket and residues involved in substrate binding. A structure-based sequence alignment of BsIDH with IDH-related enzymes (TmIDH, CgIDH and LpIolG1) reveals that there is not much overall sequence conservation in the C-terminal domain (Figure 3a). Except for His176 (catalytic base) and the GFXRRXD motif, the highly conserved residues are important in stabilizing the overall fold. An additional α-helix after β7 in LpIolG1, TmIDH and CgIDH alters the shape of the active site compared with BsIDH, forming a lid on the putative binding pocket (Figure 4). The binding pockets of LpIolG1, TmIDH and CgIDH are less solvent-exposed, and the residues pointing in the binding site would generate different interactions with substrates. Furthermore, the ‘catalytic’ sequence motif, from helices α7 and α8, differs between BsIDH/StIDH and the three other putative IDHs. In BsIDH and StIDH, the active-site consensus sequence motif is Y164XTX2AX2(D/E)TLXHEIDX2H. However, in LpIolG1, TmIDH and CgIDH, this motif is different (SGGIFXDMXIHDXDX2R), showing more similarity to the active-site consensus sequence motif GGX3DX3(Y/H) found in dimeric dihydrodiol dehydrogenase, AFR, GFOR, IDHA and rhizopine catabolic protein MocA [3,58].
Known crystal structures of the IDH subfamily
A BLAST search with these two sequence motifs followed by a structure-based sequence alignment allows the identification of six consensus sequence motifs that differentiate four IDH subgroups (Figure 4e and Supplementary Figures S2–S5 at http://www.BiochemJ.org/bj/432/bj4320237add.htm). Consensus sequence motifs I and II are involved in NAD(H) binding, whereas the other four consensus sequence motifs define the substrate-binding pocket (Figure 4e). L. plantarum contains four putative myo-IDHs (IolG1–IolG4) , and each of them represents a member of one of the IDH subgroups. The first group contains BsIDH. Residues involved in substrate binding are highly conserved among this family and suggest that they are probably myo-IDHs (Figures 4a and 4b). This group also contains LpIolG2. The second subgroup contains members that have sequence identity with LpIolG1 (Figure 4C). Subgroup 3 contains TmIDH and CgIDH (Figure 4D), which show more overall sequence identity with each other than with the other IDH-related enzymes (Supplementary Figures S1 and S6 at http://www.BiochemJ.org/bj/432/bj4320237add.htm), which also includes LpIolG3 and IDHA from various Rhizobium and Brucella species, and MocA from various Brucella species. Members of this subgroup all contain a proline-rich loop after β7 that forms part of the substrate-binding pocket. This loop forms a left-handed polyproline helix (PP-II helix) . This type of helix is often surface-exposed and is flexible, containing no internal hydrogen-bonding . In members of subgroup 3, this PP-II helix, which is approximately five residues long, forms a connection to the additional α-helix after β7. This PP-II helix probably functions to reposition the additional α-helix upon cofactor/substrate binding, thereby shielding the substrate from solvent. The last subgroup contains members with conserved sequence motifs most resembling LpIolG4, although no structure for this subgroup is known. Although the consensus sequence motifs I, II, III, V and VI show similarity to subgroup 3, the active-site consensus sequence motif IV is different (Figure 4E). For instance, at the position of Thr173 (BsIDH), this group of enzymes has a phenylalanine residue that appears to be positioned to form stacking interactions with the substrate. The diversity in active-site geometry and the differences in the conserved residues in the proposed active site would suggest that these IDH-related enzymes have different substrate specificities from BsIDH.
Bacillus subtilis IDH
Corynebacterium glutamicum IDH
Lactobacillus plantarum IDH
multi-wavelength anomalous detection
root mean square deviation
Salmonella enterica serovar Typhimurium LT2 IDH
Thermotoga maritima IDH
Karin van Straaten carried out the majority of the experiments and wrote the paper. Hongyan Zheng carried out the molecular biology aspects of this project and produced pure protein. David Palmer and David Sanders provided conceptual input, critical advice and helped to write the paper.
This work was supported by Natural Sciences and Engineering Research Council Discovery Grants to D.A.R.S. and D.R.J.P.; K.E.V. was supported by a Saskatchewan Health Research Foundation postdoctoral fellowship. Research was performed at the Canadian Light Source, which is supported by the Natural Sciences and Engineering Research Council of Canada, the National Research Council Canada, the Canadian Institutes of Health Research, the Province of Saskatchewan, Western Economic Diversification Canada and the University of Saskatchewan.
The structural co-ordinates reported for the inositol dehydrogenases used in the present study have been deposited in the PDB under accession codes 3MZ0, 3NT2, 3NT4, 3NT5, 3NTO, 3NTQ and 3NTR.