Cloning, Sequencing, and High Level Expression of the Genes Encoding Adenosylcobalamin-dependent Glycerol Dehydrase of Klebsiella pneumoniae *

The gld genes encoding adenosylcobalamin-depend- ent glycerol dehydrase of Klebsiella pneumoniae were cloned by cross-hybridization with a DNA fragment of Klebsiella oxytoca diol dehydrase genes. Since the Escherichia coli clones isolated did not show appreciable enzyme activity, plasmids for high level expression of cloned genes were constructed. The enzyme expressed in E. coli was indistinguishable from the wild-type glycerol dehydrase of K. pneumoniae by the criteria of poly- acrylamide gel electrophoretic, immunochemical, and catalytic properties. It was also shown that the recom- binant functional enzyme consists of M r 61,000, 22,000, and 16,000 subunits. Sequence analysis of the genes re- vealed four open reading frames separated by 2–12 bases. The sequential three open reading frames from the first to the third ( gldA , gldB , and gldC genes) encoded polypeptides of 555, 194, and 141 amino acid res- idues with predicted molecular weights of 60,659( (cid:97) ), 21,355( (cid:98) ), and 16,104( (cid:103) ), respectively. High level expression of these three genes in E. coli produced more than 14-fold higher level of fully active apoenzyme than that in K. pneumoniae . It was thus concluded that these are the genes encoding the endonuclease mapping. Nucleotide Sequencing— The plasmid recovered from one of the weakly hybridizing E. coli clones was designated pUCGD25 and used for analysis of the nucleotide sequence. Restriction fragments of pUCGD25 were subcloned into pUC118 or pUC119 (11), and the tem- plate single-stranded DNAs were prepared from them. DNA sequencing was performed by the dideoxyribonucleotide chain termination method of Sanger et al. (12) using a Sequencing Pro kit (Toyobo Co., Osaka, Japan). To resolve compression artifacts, urea-acrylamide gel contain*

Glycerol dehydrase (glycerol hydro-lyase, EC 4.2.1.30) and diol dehydrase (DL-1,2-propanediol hydro-lyase, EC 4.2.1.28) are enzymes that catalyze the AdoCbl 1 -dependent conversion of 1,2-diols to the corresponding deoxy aldehydes (1,2). They are formed by some genera of Enterobacteriaceae, such as Klebsiella and Citrobacter, when the bacteria grow anaerobically in a medium containing glycerol and 1,2-propanediol, respectively, and participate in the fermentation of these substrates (3,4). The enzymologic characteristics of these enzymes have been extensively studied. It has been shown that these enzymes are similar in molecular weights and substrate spectra and dissociation into two dissimilar protein components but are different in immunochemical reactivity toward anti-diol dehydrase antiserum, monovalent cation selectivity patterns, affinity for AdoCbl, and substrate specificity (for review, see Ref. 4).
In the previous paper (5), we have reported cloning and sequence analysis of the pdd genes encoding the three subunits of diol dehydrase of Klebsiella oxytoca (formerly Klebsiella pneumoniae and Aerobacter aerogenes) ATCC 8724. Since glycerol dehydrase and diol dehydrase are mutually related enzymes, it seems of much help in deducing the functional sites of these AdoCbl-requiring enzymes to elucidate the similarity and difference between these enzymes at a molecular level. We attempted to isolate the genes encoding glycerol dehydrase whose amino acid sequences have not yet been reported. Tong et al. (6) reported that Escherichia coli harboring a plasmid containing the dha regulon of K. pneumoniae showed very low glycerol dehydrase activity. Daniel and Gottschalk (7) found the growth temperature-dependent activity of glycerol dehydrase in E. coli expressing the Citrobacter freundii dha regulon.
This article describes cloning and sequence analysis of the gld genes encoding glycerol dehydrase of K. pneumoniae (formerly A. aerogenes) ATCC 25955. High level expression systems for the glycerol dehydrase genes in E. coli are also reported here.
DNA Manipulations-Standard recombinant DNA techniques were performed as described by Sambrook et al. (8). Genomic DNA from K. pneumoniae was isolated according to the method of Marmur (9).
Construction and Screening of the K. oxytoca Genomic DNA Library-Genomic DNA library of K. pneumoniae ATCC 25955 was constructed as described previously (5). Transformation of E. coli JM109 was performed by the electroporation method using BRL Gene porator as described by Dower et al. (10). Screening procedures were carried out with a DIG labeling kit (Boehringer Mannheim) using Millipore HATF nitrocellulose filters. The 1.6-kb PstI fragment of plasmid pUCDD11, which covers most of pddA gene encoding the ␣ subunit of K. oxytoca diol dehydrase (5), was used as a probe. Hybridization was carried out at 55°C. Positive clones were isolated and characterized by restriction endonuclease mapping.
Nucleotide Sequencing-The plasmid recovered from one of the weakly hybridizing E. coli clones was designated pUCGD25 and used for analysis of the nucleotide sequence. Restriction fragments of pUCGD25 were subcloned into pUC118 or pUC119 (11), and the template single-stranded DNAs were prepared from them. DNA sequencing was performed by the dideoxyribonucleotide chain termination method of Sanger et al. (12) using a Sequencing Pro kit (Toyobo Co., Osaka, Japan). To resolve compression artifacts, urea-acrylamide gel contain-ing 35% formamide was used for electrophoresis according to the manufacturer's instructions.
Construction of Expression Plasmid for Glycerol Dehydrase Genes-The DNA fragment encoding the N-terminal region of the 61,000 polypeptide was prepared by hybridization of two synthetic oligodeoxyribonucleotides, TATGAAAAGATCAAAACGATTTGCAGT (the initiation codon is underlined) and ACTGCAAATCGTTTTGATCTTTTCA. This 26-base pair fragment and 5-kb ScaI-EcoRI fragment of pUCGD25 were ligated to the NdeI-EcoRI-digested pRK172 (13) to produce expression plasmid pRK172(GD EcoRI ). Plasmid pUSI2E(DD) (5) was digested partially with NdeI and completely with EcoRI. The resulting 5.0-kb NdeI-EcoRI fragment and the 5-kb NdeI-EcoRI fragment obtained from pRK172(GD EcoRI ) were ligated to construct expression plasmid pUSI2E(GD EcoRI ). pUSI2E(GD EcoRI ) was digested with EcoRI, treated with a large fragment of E. coli DNA polymerase I in the presence of a mixture of the four deoxyribonucleoside triphosphates, and digested with XhoI. Resulting 7.4-kb DNA fragment was ligated to the 0.4-kb XhoI-EcoRV fragment of pUSI2E(GD EcoRI ) to produce pUSI2E(GD), another expression plasmid for glycerol dehydrase.
Transformation of E. coli with expression plasmids pUSI2E-(GD EcoRI ), pUSI2E(GD), and pRK172(GD EcoRI ) and cultivation of the transformants were carried out as described before (5).
Enzyme and Protein Assays-Homogenates and cell-free extracts prepared by sonication were assayed for glycerol dehydrase activity by the 3-methyl-2-benzothiazolinone hydrazone method (14). Since glycerol serves as both substrate and suicide inactivator for glycerol dehydrase, 1,2-propanediol was used as a substrate for routine assay of the enzyme. One unit of glycerol dehydrase is defined as the amount of enzyme activity that catalyzes the formation of 1 mol of propionaldehyde/min at 37°C. Protein was assayed by the method of Lowry et al. (15) with crystalline bovine serum albumin as a standard. Specific activity is expressed as units/mg protein.
Polyacrylamide Gel Electrophoresis and Activity Staining of Glycerol Dehydrase-Polyacrylamide gel electrophoresis of cell-free extracts was performed under nondenaturing conditions as described by Davis (16) in the presence of 0.1 M 1,2-propanediol or under denaturing conditions as described by Laemmli (17). Protein staining was carried out with Coomassie Brilliant Blue G-250. Activity staining for glycerol dehydrase was performed as described previously for diol dehydrase (5).

RESULTS
Cloning of the Glycerol Dehydrase Genes-Genomic DNA from K. pneumoniae ATCC 25955 was digested partially with restriction enzyme Sau3AI, and resulting 6 -20-kb fragments were inserted into the BamHI site of plasmid pUC119 (11). E. coli JM109 was transformed with the plasmids. Ampicillinresistant transformants were screened by using the 1.6-kb PstI fragment of pUCDD11 that carries most of the gene encoding the ␣ subunit of K. oxytoca diol dehydrase (5). Thirteen strongly hybridizing and 14 weakly hybridizing E. coli clones were isolated from 5 ϫ 10 3 transformants. Southern blot analysis of the plasmids from hybridization-positive clones with the probe revealed that plasmids from the former 13 clones possessed 1.2-kb and/or 0.4-kb PstI fragments and plasmids from 12 of the latter 14 clones possessed a 0.8-kb PstI fragment.
When grown in a glycerol medium or a glycerol/1,2-propanediol medium (18), five of the nine strongly hybridizing clones examined showed definite 1,2-propanediol-dehydrating activity, irrespective of aeration during cultivation. In contrast, none of the weakly hybridizing clones showed significant 1,2propanediol-dehydrating activity (Ͻ0.01 unit/mg protein) irrespective of growth conditions. Since K. pneumoniae produces both diol dehydrase and glycerol dehydrase, it may be possible that the plasmids from the strongly hybridizing clones contain genes encoding diol dehydrase, whereas plasmids from the weakly hybridizing clones contain genes encoding glycerol dehydrase.
Characterization of the Glycerol Dehydrase Genes by High Level Expression-Further characterization of the glycerol de-   hydrase genes in the insert DNA fragment of a weakly hybridizing clone was attempted by high level expression. pUCGD25, a plasmid recovered from one of such clones, was analyzed for the nucleotide sequence. Sequence analysis of the hybridization-positive 0.8-kb PstI fragment and its flanking regions revealed an ORF (Fig. 1) encoding a polypeptide whose amino acid sequence was highly homologous with that of the ␣ subunit of K. oxytoca diol dehydrase (5). Since the diol dehydrase genes encoding the ␣, ␤, and ␥ subunits exist in tandem in this order, we constructed expression plasmids pUSI2E(GD EcoRI ) and pRK172(GD EcoRI ), as described previously (5), which contain the DNA region from this ORF to the EcoRI site downstream of tac and T7 promoters, respectively. Both E. coli JM109 carrying pUSI2E(GD EcoRI ) and E. coli BL21(DE3)/pLysS carrying pRK172(GD EcoRI ) exhibited 1,2propanediol-dehydrating activity when cultured in LB ϩ 1,2propanediol medium in the presence of isopropyl-1-thio-␤-Dgalactopyranoside. Specific activity of the homogenates was 15 and 8.3 units/mg protein, respectively, which was 12 and 6 times higher than that of cell-free extract of K. pneumoniae ATCC 25955 grown on glycerol (Table I). It is therefore evident that the insert DNA fragment includes genes for functional dehydrase.
Identification as Glycerol Dehydrase and Characterization of the Expressed Gene Products-Upon polyacrylamide gel electrophoresis under nondenaturing conditions in the presence of 1,2-propanediol (19), cell-free extracts of the recombinant E. coli strains carrying plasmid pUSI2E(GD EcoRI ) or pRK172(GD EcoRI ) contained a new protein band (marked with arrowhead in Fig. 2A). When the gel was subjected to activity staining by visualizing propionaldehyde through reaction with 2,4-dinitrophenylhydrazine, this band was the only one that converted 1,2-propanediol to propionaldehyde in the presence of AdoCbl (Fig. 2B). Propionaldehyde was not formed without AdoCbl (data not shown). The cell-free extract of K. pneumoniae grown on glycerol showed the band of glycerol dehydrase in the same position (Fig. 2B), whereas the extract of E. coli JM109 transformed with pUSI2E did not. The recombinant diol dehydrase of K. oxytoca formed in E. coli (5) migrated more slowly than this band.
The substrate specificity expressed by a ratio of glyceroldehydrating activity to 1,2-propanediol-dehydrating activity (G/P ratio) was also shown in Table I. Cell-free extract from E. coli JM109 carrying pUSI2E(GD EcoRI ) showed a G/P ratio of 3.0 in the 1-min assay and 0.6 in the 10-min assay. These values coincided well with those for glycerol dehydrase of K. pneumoniae but not with those for diol dehydrase (3,18). Furthermore, the 1,2-propanediol-dehydrating activity in the extract was not immunoprecipitated by rabbit antiserum against diol dehydrase of K. oxytoca (18) (data not shown). From these results, it was concluded that the dehydrase formed in E. coli carrying pUSI2E(GD EcoRI ) or pRK172(GD EcoRI ) is identical to the wild-type glycerol dehydrase of K. pneumoniae.
When analyzed by SDS-polyacrylamide gel electrophoresis, the cell-free extract of E. coli carrying pUSI2E(GD EcoRI ) or pRK172(GD EcoRI ) contained three thick protein bands with M r of 61,000, 22,000, and 16,000, but the extract of E. coli carrying expression vector pUSI2E (control) did not (Fig. 3). Recombinant glycerol dehydrase expressed in E. coli carrying pUSI2E(GD EcoRI ) was further characterized by two-dimensional gel electrophoresis, i.e. polyacrylamide gel electrophoresis in the presence of 1,2-propanediol (nondenaturing condi- tions) followed by SDS-polyacrylamide gel electrophoresis (denaturing conditions). As shown in Fig. 4, the functional glycerol dehydrase that migrated as a single band under nondenaturing conditions in the presence of the substrate (marked with an arrowhead on the top) then dissociated into the three polypeptides with M r of 61,000, 22,000, and 16,000 upon SDSpolyacrylamide gel electrophoresis. Thus, it is evident that glycerol dehydrase apoenzyme is composed of the 61,000, 22,000, and 16,000 subunits, which were designated ␣, ␤, and ␥ subunits, respectively.
Sequence Analysis and Identification of the Glycerol Dehydrase Genes-As glycerol dehydrase genes were located in the DNA region from ORF1 to the EcoRI site, the region was subjected to nucleotide sequence analysis according to the strategy shown in Fig. 1. As summarized in Figs. 1 and 5, there existed four successive ORFs (ORF1 to ORF4). ORF1 to ORF4 were separated by 2 to 12 bases, respectively. Shine-Dalgarno sequences were found 6 to 11 bases upstream of the putative initiation codon (ATG) (GTG for ORF2) for each ORF. ORF1 to ORF4 encode polypeptides of 555, 194, 141, and 607 amino acid residues with predicted molecular weights of 60,659, 21,355, 16,104, and 63,594, respectively. The first three predicted molecular weights coincided well with M r of the ␣, ␤, and ␥ subunits of glycerol dehydrase, respectively, suggesting that ORF1 to ORF3 are the genes encoding subunits of glycerol dehydrase. To confirm this, we constructed expression plasmid pUSI2E(GD) that contained ORF1 to ORF3 but lacked ORF4 except for the first 24 bases. The homogenate of E. coli JM109 carrying this plasmid exhibited glycerol dehydrase activity of 18 units/mg protein, which was slightly higher than that of E. coli JM109 carrying pUSI2E(GD EcoRI ) ( Table I). The enzyme expressed with pUSI2E(GD) was indistinguishable from that expressed with pUSI2E(GD EcoRI ) in the following properties, electrophoretic mobilities (Figs. 2 and 3), the reactivity with anti-diol dehydrase antiserum, and a G/P ratio in 1-and 10min assays (Table I). These results indicate that inclusion of the first three ORFs (ORF1-ORF3) encoding 61,659, 21,355, and 16,104 polypeptides in an expression plasmid was sufficient to form high levels of functional glycerol dehydrase. It is evident that the polypeptide encoded by ORF4 is not a subunit of glycerol dehydrase. No additional subunits were required for activity, because specific activity of the recombinant glycerol dehydrase purified from E. coli carrying pUSI2E(GD) was essentially the same as that reported (20) with the enzyme purified from K. pneumoniae (data not shown). From all the results presented in this paper, it was concluded that ORF1, ORF2, and ORF3 are the genes encoding the ␣, ␤, and ␥ subunits of glycerol dehydrase, respectively. These were designated gldA, gldB, and gldC genes, respectively. The amino acid sequences of the ␣, ␤, and ␥ subunits of glycerol dehydrase deduced from the nucleotide sequences of gldA, gldB, and gldC genes, respectively, are shown in Fig. 5.
Sequence Homologies-The deduced amino acid sequences of the subunits of glycerol dehydrase were compared with those of the corresponding subunits of K. oxytoca diol dehydrase (Fig.  6). When properly aligned, identities of amino acid sequences of ␣, ␤, and ␥ subunits between glycerol dehydrase and diol dehydrase were 71, 58, and 54%, and similarities including the substitutions among chemically similar amino acids (21) reached 87, 78, and 73%, respectively. Thus, it became clear that remarkable homologies exist in the sequences of each subunit between these enzymes. The hydropathy profiles (22) and the predicted secondary structures (23) of the subunit polypeptides were similar to those of diol dehydrase (data not shown), suggesting that these polypeptides possess similar three-dimensional structures and functions in both enzymes.
The nucleotide sequence homologies of the coding regions for the ␣, ␤, and ␥ subunits between glycerol dehydrase genes and diol dehydrase genes were 72, 62, and 57%, respectively. No sequence homology was found between these enzymes in the regions upstream of the ␣ subunit gene. DISCUSSION Glycerol dehydrase has been reported to dissociate into components A and B upon chromatography on DEAE-cellulose or Sephadex G-100 in the absence of substrate (20,24). Neither component alone is active, but enzymatic activity is restored when both are combined. The molecular weight of components A and B determined by gel filtration are 22,000 and 189,000, respectively. We confirmed that recombinant glycerol dehydrase was also separated into components A and B, which were composed of the 22,000(␤) subunit and the 61,000(␣) ϩ 16,000(␥) subunits, respectively (data not shown).
The nucleotide sequences of the K. pneumoniae dha regulon have been submitted to GenBank (Accession U30903). According to the data, the genes corresponding to gldA, gldC, and ORF4 in this paper have been assigned as genes encoding a medium subunit, a small subunit, and a large subunit of glycerol dehydrase, respectively. In clear contrast to this, however, we reached the conclusion that the polypeptide encoded by ORF4 is not a subunit of glycerol dehydrase. Functions of the product of ORF4 are under current investigation.
The homology search revealed other highly homologous genes, i.e. the C. freundii dhaB, dhaC, and dhaE genes that lie in an opposite direction in the region upstream of the 1,3propanediol dehydrogenase gene (GenBank; Accession U09771). The deduced amino acid sequences of C. freundii dhaB, dhaC, and dhaE genes were 94, 89, and 86% identical to those of K. pneumoniae gldA, gldB and gldC genes, respectively. Since C. freundii is known to produce glycerol dehydrase (3), the dhaB, dhaC, and dhaE genes of C. freundii can be regarded as the genes encoding the ␣, ␤, and ␥ subunits of glycerol dehydrase of this bacterium.
Comparison of the deduced amino acid sequences of the ␣, ␤, and ␥ subunits of glycerol dehydrase with those of the ␣, ␤, and ␥ subunits of diol dehydrase showed a conspicuous homology between them (Fig. 6). This indicates that these dehydrases are evolutionarily related. Alignment of the amino acid sequences of the subunits of glycerol dehydrase with those of diol dehydrase indicated that the ␣ subunit may be divided into the three regions. The middle region (amino acid residues 121-406 in the glycerol dehydrase ␣ subunit) showed the highest regional homology (83% between the enzymes), whereas the amino-terminal (residues 1-120) and carboxyl-terminal (residues 407-555) regions are relatively divergent (regional homology 57 and 59%, respectively) (Fig. 6). In ␤ subunit, a highly homologous region existed in the middle of the ␤ subunit (residues 90 -143 in the glycerol dehydrase ␤ subunit, regional homology 91%), although homologies in the rest of the ␤ and ␥ subunits were lower than 60%. Amino acid deletions were observed in the amino-terminal region of the ␤ and ␥ subunits of glycerol dehydrase. From these regional homologies, it seems likely that the ␣ subunit and the middle region of the ␤ subunit is structurally and functionally important for both enzymes in common.
The deduced amino acid sequences of the glycerol dehydrase subunits did not show significant homology to the other proteins listed in the PIR and SWISSPROT data bases when analyzed using FASTA program (25). Recent x-ray crystallographic analyses of the Cbl-binding domain of E. coli methionine synthase (26) and Propionibacterium shermanii methylmalonyl-CoA mutase (27) revealed that Cbl is bound to these enzymes with the 5,6-dimethylbenzimidazole moiety displaced by an imidazole of the histidine residue. The sequence DX-HXXG, which contains this histidine residue, is reported to be conserved in these and some of the other cobalamin-dependent enzymes (26). However, this motif was not found in either diol dehydrase (5) or glycerol dehydrase, suggesting that Cbl is bound to these dehydrases in a different manner.