The crystal structure of coenzyme B12-dependent glycerol dehydratase in complex with cobalamin and propane-1,2-diol


T. Toraya, Department of Bioscience and Biotechnology, Faculty of Engineering, Okayama University, Tsushima-naka, Okayama 700–8530, Japan. Fax: + 81 86 2518264, E-mail:


Recombinant glycerol dehydratase of Klebsiella pneumoniae was purified to homogeneity. The subunit composition of the enzyme was most probably α2β2γ2. When (R)- and (S)-propane-1,2-diols were used independently as substrates, the rate with the (R)-enantiomer was 2.5 times faster than that with the (S)-isomer. In contrast to diol dehydratase, an isofunctional enzyme, the affinity of the enzyme for the (S)-isomer was essentially the same or only slightly higher than that for the (R)-isomer (Km(R)/Km(S) = 1.5). The crystal structure of glycerol dehydratase in complex with cyanocobalamin and propane-1,2-diol was determined at 2.1 Å resolution. The enzyme exists as a dimer of the αβγ heterotrimer. Cobalamin is bound at the interface between the α and β subunits in the so-called ‘base-on’ mode with 5,6-dimethylbenzimidazole of the nucleotide moiety coordinating to the cobalt atom. The electron density of the cyano group was almost unobservable, suggesting that the cyanocobalamin was reduced to cob(II)alamin by X-ray irradiation. The active site is in a (β/α)8 barrel that was formed by a central region of the α subunit. The substrate propane-1,2-diol and essential cofactor K+ are bound inside the (β/α)8 barrel above the corrin ring of cobalamin. K+ is hepta-coordinated by the two hydroxyls of the substrate and five oxygen atoms from the active-site residues. These structural features are quite similar to those of diol dehydratase. A closer contact between the α and β subunits in glycerol dehydratase may be reminiscent of the higher affinity of the enzyme for adenosylcobalamin than that of diol dehydratase. Although racemic propane-1,2-diol was used for crystallization, the substrate bound to glycerol dehydratase was assigned to the (R)-isomer. This is in clear contrast to diol dehydratase and accounts for the difference between the two enzymes in the susceptibility of suicide inactivation by glycerol.




isopropyl thio-β-d-galactoside.

Adenosylcobalamin is one of the most unique compounds in nature. It is a water-soluble organometallic compound possessing a Co–C σ bond and serves as a cofactor for enzymatic radical reactions including carbon skeleton rearrangements, heteroatom eliminations and intramolecular amino group migrations [1]. Diol dehydratase (EC of Klebsiella oxytoca is an adenosylcobalamin (AdoCbl1) dependent enzyme that catalyzes the conversions of 1,2-diols, such as propane-1,2-diol, glycerol, and 1,2-ethanediol, to the corresponding aldehydes [2,3] (Fig. 1). This enzyme has been studied intensively to establish the mechanism of action of AdoCbl [4–7]. The structure–function relationship of the coenzyme has also been investigated extensively with this enzyme [5–8]. Recently, we have reported the three-dimensional structures of its complexes with cyanocobalamin [9] and adeninylpentylcobalamin [10] and theoretical calculations of the entire energy profile along the reaction pathway with a simplified model [11–13]. In this sense, together with methylmalonyl-CoA mutase [14], glutamate mutase [15], and class II ribonucleotide reductase [16], diol dehydratase, is one of the most suitable systems with which to study the structure-based mechanisms of the AdoCbl-dependent enzymes [17,18].

Figure 1.

Conversion of 1,2-diols to the corresponding aldehydes by diol dehydratase.

Glycerol dehydratase (EC catalyzes the same reaction (Fig. 1) as diol dehydratase [19–21]. Although this enzyme is isofunctional with diol dehydratase, these two enzymes bear different physiological roles in the bacterial metabolisms [6,7]. Selected genera of Enterobacteriaceae, such as Klebsiella and Citrobacter, produce both glycerol and diol dehydratases, but the genes for them are independently regulated [22–25]: glycerol dehydratase is induced when Klebsiella pneumoniae grows in the glycerol medium, whereas diol dehydratase is fully induced when it grows in the propane-1,2-diol-containing medium, but only slightly in the glycerol medium. Glycerol dehydratase is a key enzyme for the dihydroxyacetone (DHA) pathway [23,26,27], and its genes are located in the DHA regulon [28,29]. On the other hand, diol dehydratase is a key enzyme for the anaerobic degradation of 1,2-diols [30,31], and its genes are located in the pdu operon [32–34]. Furthermore, although glycerol and diol enzymes are similar in their subunit structures, there are several distinct differences between them in the following properties: the rate of suicide inactivation by glycerol, substrate spectrum, monovalent cation requirement, affinity for cobalamins, and immunochemical cross-reactivity [6,7].

In this paper, we report the method of purifying recombinant apoglycerol dehydratase from overexpressing Escherichia coli cells and the crystal structure of glycerol dehydratase in complex with cyanocobalamin and propane-1,2-diol. We intended to explain the above-mentioned differences between two dehydratases by comparing the three-dimensional structure of this enzyme with that of diol dehydratase [9,10].

Experimental procedures


Crystalline AdoCbl was a gift from Eizai, Tokyo, Japan. DEAE-cellulose was purchased from Wako, Osaka, Japan. The other chemicals were analytical grade reagents.

Preparation of expression plasmids for His6-tagged glycerol dehydratase and its His6-tagged β subunit

DNA segments encoding carboxyl terminal region of the glycerol dehydratase α subunit was amplified by PCR using pUSI2E(GD) [28], pfu DNA polymerase (Stratagene) and pairs of primers 5′-TCTGAGTGCGGTGGAAGAGATGATGAAGCG-3′ and 5′-AGATCTTATTCAATGGTGTCGGGCTGAACC-3′ and digested with EcoRV and BglII. Resulting 210-bp fragment was ligated with the 1.5-kb HindIII-EcoRV fragment from pUSI2E(GD) and pUSI2E digested with HindIII and BglII to yield pUSI2E(αG). DNA segments encoding the β and γ subunits of glycerol dehydratase were amplified by PCR using pairs of primers 5′-CATATGCAACAGACAACCCAAATTCAGCCC-3′ and 5′-AGATCTTATCACTCCCTTACTAAGTCGATG-3′ for the β subunit and 5′-CATATGAGCGAGAAAACCATGCGCGTGCAG-3′ and 5′-AGATCTTAGCTTCCTTTACGCAGCTTATGC-3′ for the γ subunit. The segments were digested with NdeI and BglII and ligated with 3.5-kb ApaI-BglII fragment and 1.5-kb ApaI-NdeI fragment from pUSI2E(βD) [35] to yield pUSI2E(βG) and pUSI2E(γG), respectively. Plasmid pUSI2E(βG) was digested with NdeI and BglII. Resulting 600-bp NdeI-BglII fragment was inserted into the NdeI-BglII region of pET19b to produce pET19b(H6βG). A pair of synthetic oligonucleotides, 5′-TATGGGCAGCAGCCATCATCATCATCATCACAGCAGCGGCCTGGTGCCGCGCGGCAGCAC-3′ and 5′-TAGTGCTGCCGCGCGGCACCAGGCCGCTGCTGTGATGATGATGATGATGGCTGCTGCCCA-3′ were hybridized and inserted to the NdeI site of pUSI2E(γG) to produce pUSI2E(H6Gγ). pUSI2E(βG) was digested with BamHI and BglII. The resulting 0.7-kb DNA fragment was ligated with BglII-digested pUSI2E(αG) to produce pUSI2E(αGβG). The 0.5-kb BamHI–BglII fragment from pUSI2E(H6γG) was ligated with BglII-digested pUSI2E(H6γG) to produce plasmid pUSI2E(αGβGH6γG).

Purification of recombinant glycerol dehydratase

Glycerol dehydratase was purified from recombinant Escherichia coli by a conventional procedure (method 1) or Ni-nitrilotriacetate affinity chromatography (method 2). Substrate propane-1,2-diol was added to all the buffers used throughout the purification steps to minimize dissociation of the enzyme into components A and B [36]. All operations were carried out at 0–4 °C.

Method 1.  Recombinant E. coli JM109 harboring expression plasmid pUSI2E(GD) [28] was aerobically grown at 37 °C in Luria–Burtani (LB) medium containing propane-1,2-diol (0.1%) and ampicillin (50 µg·mL−1) to D600  0.9, induced with 1 mm isopropyl thio-β-d-galactoside (IPTG) for 5 h, and harvested by centrifugation. Harvested cells were resuspended in buffer A (0.05 m potassium phosphate buffer; pH 8) containing 2 mm phenylmethanesulfonyl fluoride and disrupted by sonication for 10 min, followed by centrifugation. Ammonium sulfate was added to the supernatant to a final concentration of 50% saturation. After 60 min at 4 °C, the suspension was centrifuged, and the precipitate was re-dissolved in buffer A. The solution was subjected to gel filtration on Sepharose 6B®, and peak fractions containing the enzyme were pooled, dialyzed for 12 h against 40 volumes of 1.5 mm potassium phosphate buffer (pH 8) containing 0.2% propane-1,2-diol with a buffer exchange, and loaded on to a hydroxyapatite column which had previously been equilibrated with 2 mm potassium phosphate buffer (pH 8) containing 0.2% propane-1,2-diol. After washing the column with 5 mm potassium phosphate buffer (pH 8) containing 0.2% propane-1,2-diol, the enzyme was eluted with 13 mm potassium phosphate buffer (pH 8) containing 2% propane-1,2-diol. The eluate was concentrated and loaded on to a Sephadex G-200® column which had previously been equilibrated with 20 mm potassium phosphate buffer (pH 8) containing 2% propane-1,2-diol. The enzyme was eluted with the same buffer, and peak fractions containing the enzyme were pooled.

Method 2.  Recombinant E. coli JM109 harboring pUSI2E(αGβGH6γG) was aerobically grown at 30 °C in terrific broth containing propane-1,2-diol (0.1%) and ampicillin (50 mg mL−1) to D600  0.9, induced with 1 mm IPTG for 7 h, and harvested by centrifugation. Harvested cells were resuspended in buffer A containing 2 mm phenylmethanesulfonyl fluoride and sonicated as described above. The extract containing His6-tagged enzyme was mixed with an equal volume of buffer A containing 20 mm imidazole and 600 mm KCl and loaded on to an Ni-nitrilotriacetate agarose gel (Qiagen GmbH, Germany) column which had previously been equilibrated with buffer A containing 10 mm imidazole and 300 mm KCl. After washing the column with buffer A containing 10 mm imidazole and 300 mm KCl, the enzyme was eluted with buffer A containing 50–100 mm imidazole and 300 mm KCl. After dialysis against 40 volumes of 50 mm TrisHCl buffer (pH 8) containing 2% propane-1, 2-diol, 150 mm KCl and 2.5 mm CaCl2, His6-tagged enzyme was digested with thrombin at 25 °C for 120 min and run through the Ni-nitrilotriacetate agarose column to remove the His6-tag peptide. Because a part of the enzyme had lost the β subunit, the enzyme solution was concentrated and supplied with purified β subunit by incubation at 30 °C for 30 min, followed by Sepharose 6B® gel filtration to remove unbound, excess β subunit. The β subunit was purified from E. coli BL21 (DE3) carrying pET19b(H6βG), as described above for glycerol dehydratase.

Enzyme and protein assays

Glycerol dehydratase activity was determined by a 3-methyl-2-benzothiazolinone hydrazone method [37] or an NADH–alcohol dehydrogenase coupled method [38] at 37 °C. Propane-1,2-diol was used as a substrate for routine assays because glycerol acts as both a good substrate and a potent suicide inactivator [39]. One unit of glycerol dehydratase is defined as the amount of enzyme activity that catalyzes the formation of 1 µmol of propionaldehyde per minute under the assay conditions. Protein concentration of crude enzyme was determined by the method of Lowry et al. [40] with crystalline bovine serum albumin as a standard. The concentration of purified enzyme was determined by measuring the absorbance at 280 nm. The molar absorption coefficient at 280 nm calculated by the method of Gill and von Hippel [41] for this enzyme is 112 100 m−1 cm−1.

Separation of the enzyme into components A and B

A purified preparation of the enzyme (80 units) was applied to a column (bed volume, 2.0 mL) of DEAE cellulose that had been equilibrated with 10 mm potassium phosphate buffer (pH 8) containing 10 mm propane-1,2-diol. After washing the column with 50 mL of 10 mm potassium phosphate buffer (pH 8), components A and B were eluted successively with 5 mL of 10 mm potassium phosphate buffer (pH 8) containing 40 mm KCl and then with 50 mL of 10 mm potassium phosphate buffer (pH 8) containing 300 mm KCl, respectively. Five-milliliter fractions were collected. Neither component alone was active, while the enzyme activity was restored upon addition of the other component. Therefore, components A and B were assayed by adding an excessive amount of one component and making the other rate-limiting.

PAGE and activity staining of glycerol dehydratase

PAGE was performed under nondenaturing conditions as described by Davis [42] in the presence of 0.1 m propane-1,2-diol [32], or under denaturing conditions as described by Laemmli [43]. Protein was stained with Coomassie brilliant blue G-250. Densitometry was carried out by Personal Scanning Imager PD110 (Molecular Dynamics). Activity staining for glycerol dehydratase was performed as described previously for diol dehydratase [32]. The apparent molecular weight of the enzyme was estimated by the nondenaturing PAGE on a Multigel 2–15% gradient gel (Daiichi Pure Chemicals, Tokyo, Japan) [44].

Kinetic analysis of the enzyme

Substrate-free enzyme used for measuring Km values for substrates was obtained by gel filtration on Sephadex G-25® or dialysis for 3 days against 500 volumes of buffer A with several changes. One-minute assay was employed for measurement of Km for glycerol, as glycerol induces suicide inactivation of the enzyme [39]. Apparent Km values for substrates and AdoCbl were determined at an AdoCbl concentration of 15 µm and at a fixed propane-1,2-diol concentration of 100 mm, respectively.

EPR measurements

The complex of glycerol dehydratase with adenosylcobinamide 3-imidazolylpropyl phosphate [45] was formed by incubating apoenzyme (≈100 units, 4.55 nmol) at 25 °C for 5 min with 50 nmol of the coenzyme analog in 0.65 mL of buffer A (pH 8) under a nitrogen atmosphere. Propane-1,2-diol (50 µmol) was then added. After the mixture had been incubated at 25 °C for an additional 30 min, the mixture was quickly frozen in an isopentane bath (cooled to ≈ −160 °C) and then in a liquid nitrogen bath. The sample was transferred to an EPR cavity and kept at −130 °C with a cold nitrogen gas flow controlled by a JEOL JES-VT3A temperature controller. EPR spectra were taken at −130 °C on JEOL JES-RE3X spectrometer modified with a Gunn diode X-band microwave unit under the same conditions as those described for diol dehydratase [46].

Crystallization and data collection

Purified glycerol dehydratase (64 mg·mL−1) in 20 mm potassium phosphate buffer (pH 8) containing 2% propane-1,2-diol was converted to the enzyme·cyanocobalamin·propane-1,2-diol complex by the same method as that for diol dehydratase [9] except that lauryldimethylamine oxide was not included. The complex was crystallized by the sandwich-drop vapor diffusion method at 4 °C. X-ray diffraction data were collected at 100 K using the Quantum-4R CCD detector (ADSC) on the BL40B2 beam line at SPring-8, Japan (Table 1). Reflection data were indexed, integrated and scaled using the programs Mosflm and SCALA in the CCP4 suite [47] with DPS [48].

Table 1. Statistics of data collection and structure determination. The values in parentheses are for the highest resolution shell.
  • R-factor=Σ||Fo| − |Fc||/Σ|Fo|. Rwork or the working R-factor is calculated on the 90% of the observed reflections used for the refinement.

  • Rfree or the free R-factor is calculated on the 10% of reflections excluded from the refinement.

Data collectionX-ray sourceSPring-8 BL40B2
DetectorADSC Quantum-4R
Wavelength (Å)0.816
Temperature (K)100
Space groupP21
Unit cell (Å) 
β (°)96.8
Resolution (Å)2.0
Measured reflections958 394
Unique reflections130 635
Completeness (%)99.6 (97.6)
Rmerge0.097 (0.38)
RefinementResolution range (Å)45.0–2.1
Used reflections113 453
Completeness (%)99.9

Structure determination and refinement

The structure of the enzyme was determined and refined using the program cns[49]. The models were built using Xfit of xtalview[50] and checked by procheck. No noncrystallographic symmetry (NCS) restraints were enforced during whole refinements. For adjusting the positions of atoms, a composite-omit map (2Fo − Fc) and an Fo − Fc map were used.

A data set obtained was up to 2.0 Å resolution. Crystallographic data are listed in Table 1. The α, β and γ subunits of glycerol dehydratase show substantially high homology to the corresponding subunits of diol dehydratase: their identities are 71, 58 and 54%, respectively, and their similarities 87, 78 and 73%, respectively [28]. We started the structure determination by the molecular replacement method with the αβγ heterotrimer unit of diol dehydratase from Protein Data Bank (PDB) entry 1DIO [9] as a reference structure. After a cross-rotation search, multiple translation searches were performed, and the monitor and the packing values were checked to determine the result from the candidates. We concluded that there isan(αβγ)2 dimer in an asymmetric unit of the cell. The calculated VMvalue was 3.03 Å3·Da−1. (VM = Vcell/Z·Mr, where Vcell and Z are the unit cell volume and the number of protein molecules per unit cell, respectively).

At this stage, the residues of diol dehydratase were replaced with the corresponding residues of glycerol dehydratase. After one set of rigid-body refinement and simulated annealing were applied, a composite-omit map (2Fo − Fc) was calculated. On this map, distinct electron densities were observed in the positions next to N- and C-ends of each chain. They could be assigned to certain amino-acid residues, because the C-terminal three residues of αD, the N-terminal 45 residues of βD, and the N-terminal 36 residues and the C-terminal three residues of γD were missing in the reported structure of diol dehydratase [9]. In addition, αG and βG are longer by one amino acid than αD in the N-terminal and by three than βD in the C-terminal, respectively. In the final structure, all residues of the α chain (Met1–Glu555), all but N-terminal 10 residues of the β chain (Phe11–Glu194), and all but N-terminal 3 residues of the γ chain (Lys4–Ser141) could be assigned to the electron density map. In glycerol dehydratase, the numbers of missing residues were smaller than those of diol dehydratase. We have not determined yet whether these missing residues are actually truncated by hydrolysis or not visible because of their high mobility.

After the successive repeats of modeling, energy-minimization and simulated annealing, about 900 water molecules were picked up, and B-factors for all the atoms were refined. The structure showed good stereochemistry with root-mean-square (rms) deviations of 0.006 Å from the ideal bond length and 1.30° from ideal bond angles. The resulting Rwork and Rfree were 0.208 and 0.248, respectively, in the resolution range of 45.0–2.1 Å.

Unless otherwise stated, structural figures were created with molscript[51] and raster3d[52].

Accession number

The atomic coordinates have been deposited in the Protein Data Bank with an accession code of 1IWP.

Results and discussion

Purification and characterization of recombinant glycerol dehydratase

Recombinant nontagged glycerol dehydratase was purified by a conventional method. As shown in Table 2, glycerol dehydratase overexpressed in E. coli was purified by ammonium sulfate fractionation and chromatography on Sepharose 6B®, hydroxyapatite, and Sephadex G-200® (method 1). The enzyme was purified 4.3-fold in a yield of 63%. Specific activity was about 65 units/mg. SDS/PAGE analysis showed that three bands with an Mr of 61 000 (α), 22 000 (β) and 16 000 (γ) (marked with an arrowhead) were overexpressed in E. coli carrying pUSI2E(GD) (Fig. 2A) and progressively enriched upon purification, and that only these subunits were found in the purified preparation of the enzyme. When the enzyme was electrophoresed under nondenaturing conditions in the presence of substrate (Fig. 2B), however, two bands were seen upon protein staining. The ratio of the upper protein band to the lower one was estimated to be approximately 2 by densitometric scanning. Activity staining of the enzyme indicated that the upper band reconstituted catalytically active holoenzyme with added AdoCbl, but the lower one did not (data not shown). The mobility of the upper band was identical with that of active glycerol dehydratase in the extract of K. pneumoniae ATCC 25955 (data not shown). Two-dimensional PAGE showed that the upper band consisted of the α, β and γ subunits in a molar ratio of 1.0 : 1.0 : 1.2. The lower band was composed of the α and γ subunits in a molar ratio of 1.0 : 0.9. When the purified enzyme was subjected to nondenaturing PAGE on a Multigel 2/15 gradient gel [44], two bands appeared upon protein staining, and only the upper one stained upon activity staining (data not shown). The Mr values for the upper and lower bands were 220 000 and 200 000, respectively. These data suggest that the subunit compositions of the proteins in the upper and lower bands are most likely α2β2γ2 (active apoenzyme, predicted molecular mass of 196 236 Da) and α2γ2 (component B, predicted molecular mass of 153 526 Da), respectively.

Table 2. Purification of recombinant glycerol dehydratase.
Purification stepTotal activity
Total protein
Specific activity
  • a

    Purification from 4.7 g of wet cells.

  • b

    b Purification from 13 g of wet cells.

Method 1a
 Crude extract12 40082315.11001
 (NH4)2SO4 fractionation11 50037630.6932.0
 Sepharose 6B®10 50018556.8853.8
 Sephadex G-200®778011965.4634.3
Method 2b
 Crude extract46 000192024.01001
 Ni-nitrilotriacetic24 100234100524.2
 Thrombin digested/Ni-nitrilotriacetic977010593213.9
 Sepharose 6B®845070.6120185.0
Figure 2.

Characterization of purified glycerol dehydratase by PAGE. SDS/PAGE (A, E) and nondenaturing PAGE (B, F) analyses of the enzyme at each purification step of method 1 (A, 13.5% gel; B, 7.5% gel) and method 2 (E, 12% gel; F, 7.5% gel). Resolved components A and B were also subjected to SDS/PAGE (C) and nondenaturing PAGE (D). Proteins were stained by Coomassie brilliant blue G. Molecular mass markers, Sigma SDS-7 L. Positions of the α, β and γ subunits and their complexes are indicated with arrowheads in the right. H62β2γ2 and H6-γ represent the His6-tagged α2β2γ2 complex and His6-tagged γ subunit, respectively.

In order to confirm this assignment for the lower band, we attempted to see what happens if component A is added to the purified preparation of the enzyme. We prepared components A and B by separation of purified enzyme upon DEAE–cellulose chromatography in the absence of substrate. Recoveries of activity of components A and B were 24% and 10%, respectively, although weak glycerol dehydratase activity was observed in the ‘component B’ fraction alone. SDS/PAGE analysis showed that components A and B contain the β subunit alone and a 1 : 1 mixture of the α and γ subunits, respectively (Fig. 2C). Thus, it was concluded that the inactive protein contaminated in the purified enzyme (lower band in Fig. 2B) is component B. When an excessive amount of component A was added to the purified enzyme, propane-1,2-diol-dehydrating activity increased by 59%. PAGE analysis under nondenaturing conditions showed that the catalytically inactive lower band seen in the purified enzyme was converted to the active upper band upon the addition of component A (Fig. 2D). Three bands were observed in the ‘component B’ fraction upon nondenaturing PAGE. Positions of the top and middle bands coincided well with the two bands observed with the purified enzyme. Thus, it was suggested that the middle and top minor bands of the ‘component B’ fraction correspond to component B (α2γ2) and a trace of contaminating active apoenzyme, α2β2γ2. The bottom major band that had newly appeared has not been identified yet.

For brevity, we developed a quick purification method (method 2) for His6-tagged component A and glycerol dehydratase. His6-tagged enzyme was overexpressed in E. coli and purified with an Ni-affinity column. After removal of His6 tag by digestion with thrombin, followed by passage through the Ni-nitrilotriacetate column, the enzyme obtained was incubated with excess component A and run through a Sepharose 6B® column. Highest specific activity of the enzyme (120 U·mg−1) was obtained by this simple procedure (Table 2). PAGE analysis under denaturing (Fig. 2E) and nondenaturing conditions (Fig. 2F) indicated that the purified enzyme was contaminated with component B, and that the contaminating component B recombined with the β subunit (component A) to form α2β2γ2 that resisted dissociation upon Sepharose 6B® column chromatography. As a result, the enzyme was purified 5.0-fold in a yield of 18%. This method was employed for crystallization of glycerol dehydratase.

Kinetic parameters and stereospecificity of recombinant glycerol dehydratase

Kinetic constants of the purified recombinant glycerol dehydratase for AdoCbl, propane-1,2-diol, and glycerol were in reasonable agreement with those reported previously for the enzyme from K. pneumoniae(Table 3), suggesting that the recombinant enzyme and the enzyme from K. pneumoniae are not distinguishable. When (R)- and (S)-propane-1,2-diols were used independently as substrates, the rate with the (R)-enantiomer was 2.5 times faster than that with the (S)-isomer (Table 3). The affinity of the enzyme for the (S)-isomer was essentially the same or only slightly higher than that for the (R)-isomer [Km(R)/Km(S) = 1.5]. This preference to the (S)-isomer is significantly less marked than that reported with diol dehydratase [53,54].

Table 3. Kinetic constants for the coenzyme and substrates.
Glycerol dehydrataseKm for AdoCbl (nm)Km (kcat) [mm (s−1)]
  1. a From [39]. bFrom [61]. cFrom [62]. dMean ± SD, n = 11–14.

Recombinant (method 1)81.20.32  
Recombinant (method 2)81.1 (860)0.32 ± 0.08d (392)0.40 ± 0.09 (604 ± 32)d0.27 ± 0.10 (244 ± 23)d
Klebsiella pneumoniae enzyme20a1.50b0.36c  

EPR spectroscopic evidence for the ‘base-on’ mode of cobalamin binding

To identify the Co-coordinating base, EPR spectra of the suicidally inactivated complexes of the enzyme with unlabeled and [imidazole-15N2]-labeled adenosylcobinamide 3-imidazolylpropyl phosphate were compared. The EPR spectra obtained with these analogs were exactly the same as those reported for diol dehydratase [46,55] (data not shown). With the unlabeled imidazolyl analog, each line of the hyperfine octet (coupling constant, 10.6 mT) showed superfine splitting into triplets (coupling constant, 1.9 mT). With the [imidazole-15N2]-labeled analog, on the other hand, the hyperfine lines (coupling constant, 10.7 mT) showed superhyperfine splitting into doublets (coupling constant, 2.7 mT). The ratio of the coupling constant with 14N (A14N) to that with 15N (A15N) was 0.704, which is in good agreement with the theoretical one that can be calculated as follows:


where γ is a gyromagnetic ratio. These lines of evidence indicated that the axial ligand to Co(II) is the imidazole of the coenzyme analog. Therefore, it is evident that, like diol dehydratase [46,56], glycerol dehydratase binds AdoCbl in the so-called ‘base-on’ mode. This conclusion is consistent with the finding of Poppe et al. that p-cresolylcobamide coenzyme is inactive and serves as an inhibitor for diol and glycerol dehydratases [57].

Overall structure of glycerol dehydratase in complex with cobalamin and propane-1,2-diol

The crystal structure of glycerol dehydratase in complex with cobalamin and propane-1,2-diol was determined at 2.1 Å resolution by the molecular replacement method. The schematic view of the overall structure is shown in Fig. 3A. The enzyme exists as a dimer of the αβγ heterotrimer. There is a noncrystallographic twofold axis around the center of Fig. 3A. The structure of an αβγ heterotrimer unit is shown in Fig. 3B. The central region of the α subunit constitutes the (β/α)8 barrel, the so-called TIM (triosephosphate isomerase) barrel. Propane-1,2-diol, a substrate, and K+, an essential cofactor, are bound inside the barrel. The active site-cavity is covered by the corrin ring of cobalamin that is bound on the interface of the α and β subunits. Two α subunits form dimer α2 to which two β and two γ subunits are bound separately. This structure is quite similar to that of diol dehydratase [9]. To compare the Cα trace between glycerol and diol dehydratases, the αβγ structure of glycerol dehydratase superimposed on the structure of diol dehydratase is shown in Fig. 3C with the rms deviation ranges differently colored. It is clear that deviations of atoms in the β and γ subunits are relatively large, although the rms deviation of Cα atoms in the α subunit was less than 1.0 Å. The Km values of glycerol dehydratase for AdoCbl is 40–100 times lower than that of diol dehydratase (Table 3). Such higher affinity of glycerol dehydratase for AdoCbl may be explained by the closer contact between the α and β subunits in which cobalamin sits.

Figure 3.

Structure of glycerol dehydratase in complex with cobalamin and propane-1,2-diol. (A) Overall structure. One αβγ heterotrimer unit of the two was drawn in a schematic model, and the other in a tube model. α, β and γ chains are colored in blue, yellow and orange, respectively, darkening continuously from the N- to C-terminal sides. Eight β strands constituting the (β/α)8 barrel are drawn in cartoon. Cobalamin, propane-1,2-diol and K+ are shown as CPK models colored in pink, green, and cyan, respectively. (B) Structure of the αβγ heterotrimer unit. (C) Stereoview of the Cα traces of the αβγ heterotrimer unit. The corresponding traces of diol dehydratase are drawn in gray. Root mean square deviation (Å) from the diol dehydratase structure: dark blue, < 0.5; light blue, 0.5–1.0; yellow, 1.0–1.5; orange, 1.5–2.0; red, 2.0–5; pink, > 5. Cobalamin, propane-1,2-diol and K+ are shown as ball-and-stick models. (D) Stereo drawing of the distribution of the conserved and different residues. Identical and different residues are shown in a blue ball-and-stick model and colored CPK models, respectively. Red, different; orange, weakly conserved; yellow, strongly conserved [63].

Glycerol dehydratase is isofunctional with diol dehydratase, and its amino acid sequences of the α, β and γ subunits are 71, 58 and 54% identical with those of diol dehydratase [28]. They are immunologically different or only slightly cross-reactive under nondenaturing conditions [22], but anti-(K. oxytoca diol dehydratase) antiserum cross-reacted with K. pneumoniae glycerol dehydratase to some extent under denaturing conditions (data not shown). As shown in Fig. 3D, most of the amino acid residues that are not conserved between these enzymes are located on the surface of the glycerol dehydratase molecule, whereas the conserved residues constitute the core part of the enzyme. This fact explains the above-mentioned very low cross-reactivity of glycerol dehydratase with anti-(diol dehydratase) antiserum under nondenaturing conditions and its low but distinct cross-reactivity under denaturing conditions.

Cobalamin-binding site and the conformation of bound cobalamin

Figure 4A depicts the structure of the active site in the (β/α)8 barrel. Substrate propane-1,2-diol and K+ are locked in the active-site cavity that is isolated from a bulk of water by the corrin ring of cobalamin. Figure 4B shows the structure around the enzyme-bound cobalamin. The cobalamin molecule is bound between the α and β subunits in the so-called ‘base-on’ mode − that is, with the 5,6-dimethylbenzimidazole moiety coordinating to the cobalt atom. Again, this binding mode is quite similar to that of diol dehydratase [9]. Crystallographic indication of the base-on mode of cobalamin binding in class II ribonucleotide reductase of Lactobacillus leichmannii has also been reported very recently by Drennan and coworkers [16]. The amino acid residues of diol dehydratase that are hydrogen-bonded to the peripheral amide side chains of the corrin ring [9] are all conserved in glycerol dehydratase as well. In addition to these conserved residues, the hydroxyl group of Serβ122 is hydrogen-bonded to the amide oxygen of the g-acetamide side chain of the corrin ring in glycerol dehydratase (red dotted line in Fig. 4B). In diol dehydratase, the corresponding residue is Proβ155 that cannot form the hydrogen bond. Furthermore, the lengths of the hydrogen bonds are shorter with five amino acid residues and longer with four residues in glycerol dehydratase than those in the diol enzyme. These may be reminiscent of the fact that the former enzyme binds AdoCbl much more tightly than the latter enzyme (Table 3).

Figure 4.

Structures of the active site and the cobalamin-binding site. (A) Stereo drawing of the active-site cavity viewed from the direction parallel to the plane of the corrin ring. Active-site residues interacting with the substrate (green) and K+ (cyan) are shown in ball-and-stick models. Cobalamin, pink. (B) Residues hydrogen-bonded to cobalamin. The residues interacting with cobalamin from distances shorter by 0.1 Å and longer by 0.1 Å than those in diol dehydratase are colored in yellow and green, respectively. The label ‘B’ after the residue number refers to residues of the β subunit. The red hydrogen bond is present only glycerol dehydratase. (C) Orientation of the a-acetamide side chain of pyrrole ring A of the corrin ring. A possible hydrogen bond between the a-acetamide side chain and an amino acid residue is also shown. (a) Glycerol dehydratase·cobalamin complex. (b) Diol dehydratase·cobalamin complex whose structure was determined at 4 °C (PDB: 1DIO) [9]. (c) Diol dehydratase·cyanocobalamin complex (PDB: 1EGM) [10]. (d) Diol dehydratase·adeninylpentylcobalamin complex (PDB: 1EEX) [10].

In the case of glycerol dehydratase, no electron density of the cyanide ligand was seen even though diffraction data were collected at 100 K. The Co–N bond distance between the cobalt atom and N(3) of 5,6-dimethylbenzimidazole in the glycerol dehydratase-bound cobalamin is 2.48 Å. This value is close to that in the diol dehydratase·cobalamin complex (2.50 Å) whose structure was determined at 4 °C [9] and significantly longer than those in the complexes of diol dehydratase with cyanocobalamin (2.18 Å) and with adeninylpentylcobalamin (2.22 Å) [10]. We assigned the former as the diol dehydratase·cob(II)alamin complex, because no electron density corresponding to the cyano group was observed [10]. It has been reported that the Co–CN bond is cleaved by X-ray irradiation during data collection with diol dehydratase [58]. Kratky and coworkers have reported that free and glutamate mutase-bound cyanocobalamin is reduced to cob(II)alamin by X-ray irradiation [59]. Therefore, we believe that the structure reported in this paper is also that of the glycerol dehydratase·cob(II)alamin complex. The dihedral angle of the northern and southern least-squares planes is 5.5°, indicating that the corrin ring of the glycerol dehydratase-bound cobalamin is also almost planar, as compared with that of free cyanocobalamin (14.1°). This value is close to those in the diol dehydratase-bound cobalamins (2.9–5.1°) [9,10].

Figure 4C indicates the comparison of the position of the a-acetamide side chain of pyrrole ring A of the corrin ring in the glycerol dehydratase-bound cobalamin (Fig. 4Ca) with those in the diol dehydratase-bound cobalamins. It is clear that the direction of the a-acetamide side chain is very close to that in the structure of the diol dehydratase·cobalamin complex determined at 4 °C [9] (Fig. 4Cb). It seems that this side chain turns to the opposite direction to the cobalt atom, depending upon the steric bulk of the upper axial ligand (CN or adeninylpentyl group) (Fig. 4Cc,d). Thus, this offers evidence that the glycerol dehydratase-bound cobalamin exists in a five-coordinated, square-pyramidal complex, suggesting again that the bound cobalamin is actually cob(II)alamin.

Substrate- and K+-binding sites

Substrate propane-1,2-diol and the essential cofactor K+ are bound inside the TIM barrel of the α subunit (Fig. 4A). This suggests that K+ bound in the active site of glycerol dehydratase in the presence of substrate is also not exchangeable with NH4+ in the crystallization solution, as in diol dehydratase [9]. The two hydroxyl groups of substrate directly coordinate to K+(Fig. 5A). The O(2) and O(1) atoms of the substrate are fixed in the active site by hydrogen bonding with Gluα171 and Glnα297 and Hisα144 and Aspα336, respectively (Fig. 4A). K+ is hepta-coordinated by the two hydroxyls of the substrate and five oxygen atoms of the active-site residues. Such characteristics of the substrate- and K+-binding sites are quite similar to those seen in diol dehydratase [9]. Although racemic propane-1,2-diol was used for purification and crystallization, the (R)-enantiomer is better fitted to the electron-density map (Fig. 5A). When R-values were compared with (R)- and (S)-isomers in the active site, the (R)-isomer gave slightly lowervalues. Furthermore, when Fo − Fc maps were compared, there was no significant electron density left for the (R)-isomer, while slight electron density remained for the (S)-isomer. Thus, we assigned the (R)-isomer to the electron-density map. The kinetic results, however, indicate that glycerol dehydratase shows almost equal affinity toward the (S)- and (R)-isomers (Table 3). The reason for this discrepancy is at present not clear. In contrast, diol dehydratase prefers the (S)-isomer (Km(R)/Km(S) = 3.1–3.2) [9]. The subtle differences between glycerol and diol dehydratases in the positions of Valα301, Serα302, and Aspα336 (Fig. 5B) might explain the less marked preference of glycerol dehydratase to the (S)-enantiomer in the substrate binding. Glycerol serves as a very good substrate as well as a potent suicide inactivator for both glycerol dehydratase [39] and diol dehydratase [3]. It is well known that diol dehydratase undergoes the inactivation by glycerolat a faster rate than glycerol dehydratase [39,60]. It was reported by Bachovchin et al. that dioldehydratase distinguishes between ‘R’ and ‘S’ binding conformations, the enzyme·(R)-glycerol complex being predominantly responsible for the product-forming reaction, while the enzyme·(S)-glycerol complex results primarily in the inactivation reaction [60]. Therefore, the less marked preference of the glycerol dehydratase toward the (S)-isomer explains why it is inactivated by glycerol during catalysis at a slower rate than the diol dehydratase.

Figure 5.

Interaction of glycerol dehydratase with the substrate. (A) Stereo drawing of a part of the electron-density map (omit map, 2Fo − Fc) contoured for the substrate and nearby amino acid residues. K+ and propane-1,2-diol are shown as a cyan CPK model and a green stickmodel, respectively. (B) Stereo drawing of the substrate and nearby residues. Colored stick model, glycerol dehydratase; gray stick model, diol dehydratase.


We would like to thank Dr Keiko Miura for her kind help in data collection at the BL40B2 beamline, SPring-8, Japan. We thank Ms. Yukiko Kurimoto for her assistance in manuscript preparation.