Molecular cloning of human GDP-mannose 4,6-dehydratase and reconstitution of GDP-fucose biosynthesis in vitro.

We have cloned the cDNA encoding human GDP-mannose 4,6-dehydratase, the first enzyme in the pathway converting GDP-mannose to GDP-fucose. The message is expressed in all tissues and cell lines examined, and the cDNA complements Lec13, a Chinese Hamster Ovary cell line deficient in GDP-mannose 4,6-dehydratase activity. The human GDP-mannose 4,6-dehydratase polypeptide shares 61% identity with the enzyme from Escherichia coli, suggesting broad evolutionary conservation. Purified recombinant enzyme utilizes NADP+ as a cofactor and, like its E. coli counterpart, is inhibited by GDP-fucose, suggesting that this aspect of regulation is also conserved. We have isolated the product of the dehydratase reaction, GDP-4-keto-6-deoxymannose, and confirmed its structure by electrospray ionization-mass spectrometry and high field NMR. Using purified recombinant human GDP-mannose 4,6-dehydratase and FX protein (GDP-keto-6-deoxymannose 3,5-epimerase, 4-reductase), we show that the two proteins alone are sufficient to convert GDP-mannose to GDP-fucose in vitro. This unequivocally demonstrates that the epimerase and reductase activities are on a single polypeptide. Finally, we show that the two homologous enzymes from E. coli are sufficient to carry out the same enzymatic pathway in bacteria.

Fucose is found as a component of glycoconjugates such as glycoproteins and glycolipids in a wide range of species from humans to bacteria. For example, fucose is a component of the capsular polysaccharides and antigenic determinants of bacteria, while in mammals fucose is present in many glycoconjugates, the most widely known being the human blood group antigens. Fucose-containing glycoconjugates have been implicated as playing key roles in embryonic development in the mouse (1) and more recently in the regulation of the immune response, specifically as a crucial component of the selectin ligand sialyl Lewis X (reviewed in Refs. 1 and 2). In all cases, fucose is transferred from GDP-fucose to glycoconjugate acceptors by specific transferases. Thus, defects in GDP-fucose biosynthesis will affect all fucosylation within the cell. Recently, individuals deficient in the biosynthesis of GDP-fucose have been identified (3,4) and suffer from the immune disorder leukocyte adhesion deficiency type II (LADII). 1 These patients fail to synthesize fucosylated blood groups, and their leukocytes do not express the fucose containing carbohydrate sialyl Lewis X. The patient's leukocytes do not extravasate normally, which leads to recurrent infections.
In his pioneering work in the early 1960s, Ginsberg (5, 6) elucidated the enzymatic pathway converting GDP-mannose to GDP-fucose. Later, Yurchenco and Atkinson (7) showed that this was the primary biosynthetic route to GDP-fucose. As shown in Fig. 1, GDP-mannose is converted to GDP-fucose by GDP-mannose 4,6-dehydratase via the oxidation of mannose at C-4 followed by the reduction of C-6 to a methyl group, yielding GDP-4-keto-6-deoxymannose. The reaction has been reported to proceed with transfer of a hydride from C-4 to C-6 (8) by a tightly bound cofactor, thought to be NADP ϩ or NAD ϩ , which is regenerated during the reaction. This intermediate is then epimerized at C-3 and C-5 to yield GDP-4-keto-6-deoxy-glucose and finally reduced by NADH or NADPH at C-4 to produce GDP-fucose. In the initial studies, it was not certain if the last two steps, the epimerizations and reduction, were performed by one enzyme or two. One potential regulatory mechanism in the pathway was first revealed in the studies of Kornfeld and Ginsberg (9), who demonstrated that GDP-mannose 4,6-dehydratase was inhibited by the final product in the biosynthetic pathway, GDP-fucose.
Recent studies have addressed several open questions about the enzymes in this pathway. Two studies have shown that GDP-mannose 4,6-dehydratase utilizes NADP ϩ and not NAD ϩ as a cofactor. Yamamoto et al. demonstrated this with the enzyme from Klebsiella pneumoniae (10) and more recently Sturla et al. detected NADP ϩ bound to the Escherichia coli enzyme (11). This requirement for NADP ϩ differentiates GDPmannose 4,6-dehydratase from the two other well characterized sugar nucleotide 4,6-dehydratases, dTDP-glucose 4,6-dehydratase and CDP-glucose 4,6-dehydratase, both which require NAD ϩ (12,13). Additionally, recent studies have addressed the question of whether the epimerase and reductase activities are present in one protein or are two separate proteins as is the case in dTDP-rhamnose biosynthesis (14,15). Serif and co-workers (16) suggested that the 3,5-epimerase and * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  4-reductase activities were present on a single polypeptide when they purified small amounts of the enzyme from pig thyroids (16). This was confirmed by Tonetti et al. (17) when they cloned the human protein FX. Sequencing the gene revealed homology to the bacterial sugar nucleotide reductases. Using antibody depletion experiments, purified protein, and cell extracts as a source for the GDP-4-keto-6-deoxymannose, they demonstrated that FX combined both the epimerase and reductase activities in one polypeptide.
To understand better the human enzymes involved in this pathway, their role in selectin-mediated cell adhesion, and the LADII defect, we have undertaken the molecular cloning of the human gene encoding GDP-mannose 4,6-dehydratase. Furthermore, utilizing purified recombinant enzymes expressed in E. coli, we have reconstituted the GDP-fucose biosynthetic pathway in vitro. We demonstrate that two enzymes, GDPmannose 4,6-dehydratase and GDP-4-keto-6-deoxymannose 3,5-epimerase, 4-reductase, are sufficient to synthesize GDPfucose from GDP-mannose, confirming earlier studies suggesting that both epimerase and reductase activities are encoded in a single polypeptide. Additionally, we show that human GDPmannose 4,6-dehydratase has a strict specificity for NADP ϩ over NAD ϩ . Using the homologous E. coli enzymes, GDP-mannose 4,6-dehydratase (GMD) and GDP-4-keto-6-deoxymannose 3,5-epimerase, 4-reductase (WCAG), we demonstrate that the same is true in bacteria. We also show that human GDPmannose 4,6-dehydratase is subject to feedback inhibition by GDP-fucose and that this, along with its differential levels of gene expression, provides potential mechanisms for regulating its activity.

Data Base Searching and Sequence Alignments-The National
Center for Biotechnology Information (NCBI) EST data base was searched with BLAST on the NCBI server. The human and E. coli dehydratase peptides were aligned with the program GAP of the Genetics Computer Group analysis package. Amino terminal peptides of the GDP-mannose 4,6-dehydratase from different species were aligned using the GeneWorks program from IntelliGenetics, Inc.
Cloning Human GDP-mannose 4,6-Dehydratase-A cDNA library was constructed from HL-60 cells as described by Sako et al. (18). This plasmid-based cDNA expression library was assembled into 19 pools, each representing 60,000 -100,000 individual clones/pool. The pools were screened by PCR using DNA primers based on both the mouse EST (accession number W29220) and the published E. coli sequence: 5Ј-TGATGAGCCAGAGGACTTTGTCATAGCTAC-3Ј and 5Ј-CAGAAA-GTCCACTTCAGTCGGTCGGTAGTA-3Ј.
Two pools gave the expected 200-base pair fragment. The 200-base pair PCR product was reamplified by PCR, random primer-labeled with 32 P, and used to identify positive clones from the two library pools by colony hybridization. This approach yielded plasmid pMT-hGMD containing the cDNA of human GDP-mannose dehydratase in a eukaryotic expression vector.
The positive pools were probed with the 32 P-labeled PCR primers. This yielded plasmid pMT-hFX, containing the cDNA of human GDPmannose epimerase-reductase in a eukaryotic expression vector.
Transfection of Lec13 and Cell Staining-The Chinese hamster ovary (CHO) cell line Lec13, was obtained from Professor P. Stanley at Albert Einstein Collage of Medicine. This line was first transfected with pMT-NeoFTIV, a vector expressing human fucosyltransferase IV and the neomycin resistance genes, by the calcium phosphate method as described previously (19). To make this cell line capable of replicating vectors containing the polyoma virus origin of replication, one of the Fuc-T IV-positive cell lines, clone 9E9A, was again transfected with pCDNA3.1 ZeoPyLT, a plasmid expressing the early region of polyoma virus including large T, by the lipofectamine method according to the manufacturer's instructions (Invitrogen). One zeocin-resistant clone had Fuc-T IV activity and was replication-competent. This cell line, 9E9A LT2.9, was transfected with pMT-hGMD or pMT-hFX by the lipofectamine method. After 48 h, cells were analyzed for Lewis X expression by immunofluorescence after staining with CD15 antibody (Immunotech) and goat anti-mouse fluorescein isothiocyanate secondary antibody (Boehringer Manheim).
In Vitro Assays of CHO Cell Extracts-Mutant CHO cells transfected with human dehydratase cDNA, human epimerase-reductase cDNA, or wild type CHO cells were lysed under nitrogen pressure in 0.75 ml of 25 mM Hepes, pH 7.4, 100 mM NaCl, 10 mM EDTA, 10 mM DTT for 5 min in a Parr Bomb at 900 p.s.i. on ice. The cell debris was pelleted at 50,000 ϫ g for 1 h, and soluble extracts were assayed in 25 mM Hepes, pH 7.4, 100 mM NaCl, 15 mM MgCl 2 , 10 mM DTT, 10 M GDP-mannose, with 100,000 cpm of 14 C-labeled GDP-mannose for 2 h at 37°C. The reactions were stopped by boiling for 5 min followed by centrifugation for 5 min at 15,000 rpm in a microcentrifuge. Unlabeled GDP-mannose and GDP-fucose were added as standards. GDP-mannose, GDP-fucose and the 4-keto 6-deoxy intermediate were separated as described by Yamamoto et al. (10) except the amide-80 column (Tosohaas) was run in 66% acetonitrile and 7.5 mM citric acid/Na 2 HPO 4 buffer, pH 4.0. The 14 C-labeled sugar nucleotides were detected with a flow through scintillation counter Beta-1 (Packard) run with a solid scintillant cell. The unlabeled sugar nucleotides were detected at 254 nm.
Expression and Purification of Human Enzymes from E. coli-The human dehydratase and epimerase-reductase genes were cloned by PCR into the EcoRI and HindIII sites of vector pRSETB (Invitrogen) for expression in E. coli. This yielded vectors pRSEThGMD and pRS-EThFX. The inserts in the resulting vectors were sequenced in their entirety. The resulting dehydratase fusion protein had the sequence MRGSHHHHHHGMASMTGGQQMGRDLYDDDDKDPSSPSAGTM-EF added to amino acid 20 of the predicted sequence, the position homologous to the start of the E. coli enzyme. For the epimerasereductase, the same 43-amino acid fusion peptide was added to amino FIG. 1. GDP-fucose biosynthetic pathway. The pathway for the conversion of GDP-mannose to GDP-fucose is shown. GDP-mannose 4,6-dehydratase catalyzes the oxidation of C-4 of mannose to the ketone and the reduction of C-6 to the methyl group to yield GDP-4-keto-6deoxymannose. The NADP ϩ cofactor is reduced and then oxidized during these two steps. GDP-4-keto-6-deoxy-mannose 3,5epimerase, 4-reductase catalyzes the epimerization at C-3 and C-5 to yield GDP-4-keto-6-deoxyglucose, followed by the reduction of C-4 by NADPH or NADH, yielding GDP-fucose. Chemical reduction of the GDP-4-keto-6-deoxymannose intermediate produces GDP-6-deoxytalose and GDP-rhamnose. acid 2 of the published sequence. The two expression vectors were transformed into E. coli strain BL21/DE3 and the bacteria were grown in LB media containing ampicillin and chloramphenicol. Both the dehydratase-expressing cells and epimerase-reductase-expressing cells were grown at room temperature to an A 600 of 0.5 and induced with 0.3 mM isopropyl-1-thio-␤-D-galactopyranoside for 3-4 h. The cells were resuspended in Tris-buffered saline containing DNase I and lysed in a French press at 12,000 p.s.i. After clarification, the dehydratase was purified by successive chromatography over nitrolotriacetic acid-Ni 2ϩagarose (Qiagen) eluting with 200 mM imidazole, chromatography over ceramic hydroxyapatite (Bio-Rad) eluting with a phosphate gradient (0.0 -0.5 M), and chromatography over MonoQ (Pharmacia) eluting with a NaCl gradient (0.0 -0.5 M). The epimerase-reductase was purified by successive chromatography over nitrolotriacetic acid-Ni 2ϩ -agarose eluting with 200 mM imidazole and chromatography over 2Ј-5Ј ADP-Sepharose eluting with 10 mM NADP ϩ . The resulting proteins were greater than 90% pure as judged by SDS-PAGE and staining with Coomassie Blue. Amino terminal sequencing of the first 15 amino acids of the proteins confirmed their identity as fusion protein expressed from the pRSET vector.
A second expression construct was made for each gene, which had a 20-amino acid amino-terminal leader, MRGSHHHHHHGSDYKD-DDDK, added to the Met 19 of hGMD or Met 1 of hFX. To construct these expression plasmids, synthetic DNA was used to rebuild the hGMD gene from the AgeI site near the 5Ј-end of the gene. The synthetic DNA was hybridized and ligated to the AgeI-HindIII fragment of pRSETh-GMD into the NdeI and HindIII sites of pRSETB to yield pT7hGMD. A similar approach was taken to construct pT7hFX using the BamHI-HindIII fragment of pRSEThFX. The resulting vectors pT7hMGD and pT7hFX were transformed into E. coli strain BL21/DE3, and protein was expressed and purified as above. No significant differences were observed in the activity, substrate, or cofactor preference for both the human dehydratase or epimerase-reductase having either the longer or shorter N-terminal fusion.
In Vitro Assays of Purified Dehydratase and Epimerase-Reductase-Purified proteins were assayed in 25 mM MOPS, pH 7.0, 100 mM NaCl, 5 mM EDTA, 10 mM DTT, 25 M GDP-mannose with 100,000 cpm of 14 C-labeled GDP-mannose for 1 h at 37°C. When sequential reactions were performed, the mix was incubated for an additional 1 h at 37°C after adding second enzyme or cofactor. The reactions were stopped by boiling for 2 min followed by centrifugation for 1 min at 15,000 rpm in a microcentrifuge. Unlabeled GDP-mannose and GDP-fucose were added at this point as internal standards. Kinetic constants were determined graphically using S/V versus S plots with GDP-mannose concentrations ranging from 1 ⁄2 to 5 K m . The reaction conditions were as above with 1 mM NADP ϩ added to reactions.
Paper Chromatography-14 C-Labeled GDP-mannose, GDP-fucose, and reaction products were incubated in 0.2 M NaBH 4 for 15 min at room temperature to reduce keto intermediates and then were cleaved from the sugar by incubating in 1 M trichloroacetic acid for 10 min in boiling water. This mixture was chromatographed on Whatman 3MM paper in descending mode using three different solvent systems (solvent I, water-saturated methyl ethyl ketone for 24 h; solvent II, ethyl acetate/pyridine/water (3.6:1.0:1.15) for 7 h; solvent III, ethyl acetate/ pyridine/water (10:2.5:1.5) for 17 h. Free sugar standards were localized by staining with AgNO 3 in acetone followed by NaOH in methanol (20). Radioactivity was detected by cutting the paper into strips, and 1-cm sections of each strip were counted in 1 ml of water plus 10 ml of formula 989 (Packard).
Synthesis and Characterization of GDP-4-keto-6-deoxymannose-GDP-4-keto-6-deoxymannose was synthesized from GDP-mannose using the purified human and purified E. coli dehydratases. Typical reaction conditions for human dehydratase were 2.5 mM GDP-mannose, 0.1 mg/ml human dehydratase, 25 mM MOPS, pH 7.0, 100 mM NaCl, 10 mM DTT, 5 mM EDTA, 10 M NADPH, and 10 M NADP ϩ at 37°C for 3-6 h. Typical reaction conditions for E. coli dehydratase were 10 mM GDP-mannose, 0.1 mg/ml E. coli dehydratase, 10 mM MOPS, pH 6.5, 100 mM NaCl, 2 mM DTT, 1 mM EDTA, 10 M NADPH, and 100 M NADP ϩ at 37°C for 3-6 h. The reactions were allowed to proceed to completion as judged by HPLC, and the protein was removed by ultrafiltration on a Centricon-10. The resulting mix was desalted on Sephadex G-10 run in water and lyophilized. GDP-4-keto-6-deoxymannose prepared this way was stable frozen at Ϫ20°C either as a dried powder or in water and was judged essentially pure by HPLC and ESI-MS analysis. For ESI-MS analysis, 500 mM citric acid-sodium phosphate buffer was added to a 20 pmol/l solution of GDP-4-keto-6-deoxymannose (in 50% acetonitrile) to give a final concentration of 0.5 mM at pH 4. Flow injection at 10 l/min was used to introduce this sample into the Micromass Platform II electrospray ionization-equipped mass spectrometer, which was operated in the negative ion mode with a cone voltage of 40 V. The intermediate was further characterized by high field NMR. All NMR experiments were performed on a Varian Unity plus 600-MHz spectrometer. Samples were dissolved in D 2 O, and HDO was used as an internal reference, set to 4.76 ppm at 25°C. Twodimensional total correlation spectroscopy, nuclear Overhauser effect spectroscopy, and heteronuclear multiple-quantum coherence were performed with standard Varian pulse sequences.
Cloning of E. coli GMD and E. coli WCAG-The gmd and wcaG genes were cloned from the E. coli K-12 genome using PCR oligonucleotides that created an NcoI site overlapping the initiating methionine of each gene and a HindIII site following the termination codon of each gene. The PCR products were cloned into the NcoI and HindIII sites of pSE380 (Invitrogen), yielding pSEGMD and pSEWCAG. Sequences were confirmed by DNA sequencing. During the sequencing of the wcaG gene, it was discovered that multiple clones contained two nucleotide differences from the published sequence (21). The last nucleotide in codon 255 and the first nucleotide in codon 256 changed from a CG in the published sequence to a GC in the cloned gene. This changed the predicted amino acids in those positions from an aspartate at amino acid 255 and a valine at amino acid 256 in the published sequence to glutamate and leucine in the cloned gene.
Expression and Purification of E. coli Enzymes from E. coli-E. coli strain GI934 harboring either pSEGMD or pSEWCAG were grown in LB media containing ampicillin at 370°C to an A 600 of 0.5. The cells were induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside and grown for an additional 4 h. Cell pellets were broken in a French press at 12,000 p.s.i. After clarification, the E. coli dehydratase was purified by successive chromatography over Toyopearl DEAE (TosoHaas) eluting with a NaCl gradient (0.0 -0.5 M), chromatography over ceramic hydroxyapatite eluting with a phosphate gradient (0.0 -0.5 M), and chromatography over MonoQ eluting with a NaCl gradient (0.0 -0.5 M). The E. coli epimerase-reductase was purified by successive chromatography over Toyopearl DEAE eluting with a NaCl gradient (0.0 -0.5 M), followed by chromatography over heparin-Toyopearl eluting with 200 mM NaCl. The resulting proteins were greater than 90% pure as judged by SDS-PAGE and staining with Coomassie Blue. Amino-terminal sequencing of the first 15 amino acids of the proteins confirmed their identities.

Molecular
Cloning of Human GDP-mannose 4,6-Dehydratase-To clone human GDP-mannose 4,6-dehydratase, we first performed a TBLASTN search of the NCBI EST data base using the sequence of the E. coli enzyme. This identified a mouse EST, accession number W29220, with a high degree of homology to the E. coli enzyme. We designed two oligonucleotide primers based on conserved amino acids present in both the E. coli sequence and the partial sequence of the mouse gene. Using the oligonucleotides as primers, we obtained a 200-base pair PCR fragment from a human promyelocytic cell line HL-60 cDNA library. This fragment was then used as a probe to isolate two apparently full-length cDNA clones for the putative human GDP-mannose 4,6 dehydratase, the sequence of which is shown in Fig. 2. There are two potential initiator methionines located at nucleotides 76 and 130 of this sequence. As shown in Fig. 3A, the downstream methionine of the human protein more closely aligns with the initiating methionine of the E. coli protein. However, alignment of the human sequence to the recently cloned arabidopsis GDP-mannose 4,6-dehydratase gene and a putative C. elegans translation product (Fig. 3B) suggests that translation of the human protein initiates at the first methionine, at nucleotide 76, and that dehydratases from nonbacterial sources contain an amino-terminal extension. The human and E. coli proteins are 61% identical over their entire lengths, and both contain an extended consensus sequence, GXXGXXG, identifying the ␤␣␤ fold found in many NAD ϩ -and NADP ϩ -binding proteins (22). This sequence is found between amino acids 9 and 15 of the E. coli enzyme. Fig. 4 shows that the human dehydratase gene encodes a single mRNA transcript of about 1.7 kilobase pairs that is expressed in all tissue and cell types examined, albeit at varying levels.

Cloning of Human GDP-mannose 4,6-Dehydratase
The mRNA levels are highest in pancreas followed by small intestine, liver, colon, and prostate and lowest in ovary, brain, lung, spleen, and peripheral blood lymphocytes. Likewise, a varied level of expression was seen in the human cell lines examined (Fig. 4C).
GDP-mannose 4,6-Dehydratase cDNA Complements the Dehydratase Defect in Lec13-To demonstrate the isolated cDNA encoded GDP-mannose 4,6-dehydratase activity, we transiently transfected it into Lec13, a CHO cell line that previously has been identified as lacking GDP-mannose 4,6-dehydratase activity (23). To monitor the GDP-mannose 4,6dehydratase activity, we first stably transfected Lec13 with human fucosyltransferase IV (19), an enzyme that utilizes GDP-fucose to synthesize the Lewis X (CD15) epitope. This gave a readily identifiable cell surface marker dependent upon the biosynthesis of GDP-fucose. Restoration of dehydratase activity in the mutant cells should restore GDP-fucose biosynthesis and produce Lewis X antigen on the cell surface. As demonstrated in Fig. 5A, culture of these Fuc-T IV-expressing Lec13 cells in media containing fucose allowed the synthesis of GDP-fucose through the salvage pathway and generated CD15positive cells. Transient transfection of the same cell line with the vector expressing the human GDP-mannose 4,6-dehydratase, in media lacking fucose, also causes the cells to stain positive for Lewis X, demonstrating that the dehydratase gene complements the CHO cell defect (Fig. 5B). By contrast, transfection with the same vector expressing the human epimerasereductase gene, again in media lacking fucose, shows no staining by CD15 (Fig. 5C). As further evidence, lysates of Lec13 cells transiently transfected with human GDP-mannose 4,6dehydratase cDNA readily converted 14 C-labeled GDP-mannose to GDP-fucose (Fig. 6B), whereas lysates of Lec13 cells transfected with the human epimerase-reductase cDNA in the same expression vector did not (Fig. 6A). The level of activity in dehydratase-transfected cells (Fig. 6B, 25 g) is higher than that of wild type CHO cells (Fig. 6C, 125 g). Based on this data, we conclude that the cDNA encodes GDP-mannose 4,6-dehydratase.  Chromosomal Localization of Human GDP-mannose 4,6-Dehydratase and FX Gene-To determine if the two genes for GDP-fucose biosynthesis are linked on the human genome we mapped their location using fluorescence in situ hybridization. Full-length cDNA inserts encoding the human dehydratase (pMT-hGMD) and epimerase reductase (pMT-hFX) genes were used to probe a human genomic PAC (hGMD) or P1 (hFX) libraries (24,25) (Genome Systems, Inc., St. Louis, MO). Two genomic clones were obtained for each probe. We confirmed that the genomic clones contained the hGMD and hFX genes by subcloning and sequencing (data not shown). The hGMD and hFX genes were mapped using two-color fluorescence in situ hybridization utilizing the genomic clone for each gene and a centromere-specific probe (26) (Genome Systems, Inc., St. Louis, MO). The two genes are not linked in the human genome. The human dehydratase gene was localized to the p terminus of chromosome 6, an area corresponding to band 6p25. A total of 80 metaphase cells were analyzed, with 64 exhibiting specific labeling. In a similar fashion, human epimerase-reductase was mapped to the q terminus of chromosome 8, an area corresponding to band 8q24.3. A total of 80 metaphase cells were analyzed, with 70 exhibiting specific labeling (data not shown).
Human GDP-mannose Dehydratase and Epimerase-Reduc- tase Are Sufficient to Convert GDP-mannose to GDP-fucose-To further characterize GDP-mannose 4,6-dehydratase, and to reconstitute GDP-fucose biosynthesis in vitro, we needed to obtain purified proteins for both the dehydratase and epimerasereductase enzymes. To this end, we expressed both the human GDP-mannose 4,6-dehydratase and GDP-4-keto-6-deoxy-mannose 3,5-epimerase 4-reductase in E. coli as fusion proteins. The fusion proteins were purified (Fig. 7A), and their identities were confirmed by sequencing the first 15 amino acids of each peptide (data not shown). The human dehydratase protein migrated on SDS-PAGE near the position expected based upon its calculated molecular mass (42.7 kDa), but the epimerasereductase migrated more slowly than expected (38.3 kDa), as previously reported by Tonetti et al. (17). We confirmed that recombinant hFX had the mass predicted from its cDNA sequence by ESI-LC-MS (data not shown).
To characterize the reactions of the purified dehydratase and epimerase-reductase, we incubated the enzymes with 14 C-labeled GDP-mannose and identified the reaction products by HPLC and paper chromatography. As shown in Fig. 8A, purified GDP-mannose 4,6-dehydratase converts 14 C-labeled GDPmannose to a new species that runs at the position reported for GDP-4-keto-6-deoxymannose (10). We confirmed the identity of GDP-4-keto-6-deoxymannose by descending paper chromatography. As shown in Fig. 1, the expected monosaccharides resulting from reduction of GDP-4-keto-6-deoxymannose by borohydride and cleavage from the nucleotide with acid would be rhamnose and 6-deoxytalose. Fig. 9A (filled triangles) shows that when the reaction products obtained by incubating GDPmannose, human dehydratase, and NADP ϩ were reduced, cleaved, and run on paper, four spots resulted. The major species, running at 16 cm, co-migrates with an unlabeled standard for rhamnose. This component also co-migrates with rhamnose in solvent systems II and III (data not shown). The species at 5 cm co-migrates with mannose and was presumably derived from the unreacted starting material. The species at 24 cm runs faster than rhamnose in this solvent, as expected for 6-deoxytalose, but has an R F that differs from the published value for 6-deoxytalose (6,27). This is also true in solvents II and III. An additional species was observed in this sample, running near 37 cm. Without authentic 6-deoxytalose, we have not been able to confirm the identity of the two faster running spots. Using the purified enzyme, we were able to determine the cofactor utilized by human GDP-mannose 4,6-dehydratase. Comparing panels A and B of Fig. 8, we see that the human dehydratase utilizes NADP ϩ as a cofactor and cannot utilize NAD ϩ even at a 10-fold higher concentration.
Using the purified human dehydratase and epimerase-reductase, we could demonstrate that these two enzymes alone were sufficient to convert GDP-fucose to GDP-mannose. Sequential incubation of GDP-mannose with human dehydratase and NADP ϩ followed by human epimerase-reductase and NADPH converts the GDP-mannose to GDP-fucose, based upon co-chromatography with authentic GDP-fucose (Fig. 8C). We confirmed the identify of the product as GDP-fucose by descending paper chromatography of the free monosaccharide after reduction with NaBH 4 and cleavage from GDP with acid. Fig. 9A (open squares) shows the free monosaccharide co-migrates with an identically treated 14 C-labeled GDP-fucose standard. Identical results were obtained in solvents systems II and III (data not shown). Unlike the human dehydratase, which shows a strict cofactor preference for NADP ϩ over NAD ϩ , the human epimerase-reductase can utilize either NADPH or NADH, although NADPH is used more efficiently (data not shown).
Characterization of GDP-4-keto-6-deoxymannose by ESI-MS and High Field NMR-To confirm that the reaction product of human dehydratase and GDP-mannose was GDP-4-keto-6-deoxymannose, we isolated the product and analyzed it by mass spectrometry and NMR. Ϫ peak for residual GDP-mannose at 604.0 is also present in the spectra. A combination of one-dimensional homonuclear, two-dimensional homonuclear, and two-dimensional heteronuclear NMR experiments revealed that the isolated reaction product was a mixture of at least two related sugar nucleotides. The observed chemical shifts and coupling constants are listed in Table I. From these data, we have identified the two major species as GDP-4-keto-6-deoxymannose and GDP-3-keto-6-deoxymannose. NMR analysis of the isolated product of E. coli dehydratase also showed a similar mixture of GDP-4-keto-6-FIG. 9. Analysis of the reaction products of human and E. coli dehydratase and epimerase-reductase by descending paper chromatography. 14 C-Labeled reaction products were reduced with NaBH 4 , and the resulting sugar was cleaved with acid and spotted on Whatman 3MM paper developed in water-saturated methyl ethyl ketone for 24 h. The paper was cut into strips, and 1-cm sections of each strip were counted. A, the reactions contained GDP-mannose, human dehydratase, and NADP ϩ , followed by human epimerase-reductase plus NADPH (open squares) or GDP-mannose, human dehydratase, and NADP ϩ (filled triangles). B, the reactions contained GDP-mannose, E. coli dehydratase, NADP ϩ , E. coli epimerase-reductase, and NADPH (open squares) or GDP-mannose, E. coli dehydratase, and NADP ϩ (filled triangles). The position of 14 C-labeled GDP-mannose and GDPfucose that had been treated identically to the reaction mixtures is shown above labeled as mannose and fucose, respectively. The position of unlabeled free rhamnose is also shown at the top of the trace.
E. coli GDP-mannose Dehydratase and Epimerase-Reductase Are Sufficient to Convert GDP-mannose to GDP-fucose-To address whether the bacterial enzymes catalyze the same reactions as the human dehydratase and epimerase-reductase, we cloned, expressed, and purified the E. coli enzymes (Fig. 7B). We examined the reactions of the E. coli enzymes in the same way we characterized the human enzymes. Fig. 8E shows that incubation of E. coli dehydratase with GDP-mannose and NADP ϩ resulted in production of the reaction intermediate GDP-4-keto-6-deoxymannose. This reaction product produced only two spots in paper chromatography after reduction and cleavage. One migrated with the rhamnose standard at 17 cm, and one migrated faster at 37 cm (Fig. 9B, filled triangles). The spot seen at 25 cm in the reaction with human dehydratase is missing (Fig. 9, compare A and B (filled triangles)). The E. coli dehydratase used NADP ϩ as a cofactor and could not substitute NAD ϩ at concentrations up to 1 mM (Fig. 8, D and E).
As with the human enzymes, E. coli dehydratase and epimerase-reductase were sufficient to convert GDP-mannose to GDP-fucose. Incubation of GDP-mannose with E. coli dehydratase and E. coli epimerase-reductase in the presence of NADP ϩ and NADPH converted the GDP-mannose to GDPfucose (Fig. 8F). We confirmed the identity of GDP-fucose by paper chromatography (Fig. 9B, open squares). The E. coli epimerase-reductase can use NADH but not as efficiently as NADPH (data not shown).
We performed a preliminary characterization of the two dehydratases, and the results are shown in Table II. The E. coli enzyme has a significantly higher K m for GDP-mannose than the human enzyme but also has a significantly higher V max . Additionally, we monitored inhibition of the dehydratase by GDP-fucose, which has been proposed as a mechanism to regulate the enzyme's activity. Both human and E. coli dehydratases were inhibited by GDP-fucose with IC 50 values lower than the IC 50 values for inhibition by GDP, suggesting a specific effect and a potential role in regulation of the enzyme's activity. Quite unexpectedly, both enzymes are stimulated by NADPH at micromolar concentrations, although this co-factor does not  a The identification of the 3-keto compound was complicated by the small coupling constant of the proton on C-4 of mannose, which reduced magnetization transfer in the total correlation spectroscopy experiment. During repeated lyophilization of the samples in D 2 O to exchange protons, new peaks were evident in the NMR, suggesting the GDP-4keto-6-deoxymannose and its related compounds are not entirely stable under these conditions. play a role in catalysis. It is not clear if this stimulation by NADPH is relevant to the in vivo regulation of the enzyme. DISCUSSION We have cloned the gene encoding human GDP-mannose 4,6-dehydratase using homology between the E. coli enzyme and a mouse EST. The cloned gene complemented the previously identified GDP-mannose 4,6-dehydratase defect in the CHO cell line Lec13, demonstrating that the cDNA encodes a functional protein. The dehydratase gene shows high levels of identity between bacteria and human and indeed across the spectrum of species examined (for a comparison of dehydratases from a variety of bacterial and nonmammalian sources see Bonin et al. (28). The message for human GDP-mannose 4,6dehydratase is expressed in all tissues examined, albeit at varying levels. The varying levels of expression of the dehydratase message suggest the enzyme may be regulated at the level of transcription, and in fact there is evidence for developmental regulation in rat and nereids (29,30). It also appears that, in both humans and E. coli, GDP-fucose biosynthesis is regulated by feedback inhibition of the dehydratase by GDPfucose, the final product in the pathway. This mechanism of inhibition was noted for Aerobacter aerogenes by Kornfeld and Ginsberg (9) and was suggested for the porcine enzyme by Serif and co-workers (33).
The cloning of the human dehydratase gene, along with the recent cloning of the human epimerase-reductase by Tonetti et al. (17) has allowed us to reconstitute GDP-fucose biosynthesis in vitro using purified, recombinant enzymes. Thus, we have definitively shown that two enzymes, a dehydratase and an epimerase-reductase, are sufficient to convert GDP-mannose to GDP-fucose. In doing so, we demonstrated, using purified recombinant proteins, that in humans and in E. coli, both 3,5epimerase and 4,6-reductase activities are present in a single protein. This confirms the earlier studies of Serif and co-workers (16) on the enzyme purified from porcine thyroids and the recent work of Tonetti et al. (17) with the human FX protein.
Additionally, we find that human dehydratase has a strict cofactor requirement for NADP ϩ for which NAD ϩ cannot substitute. This is consistent with earlier reports demonstrating that GDP-mannose 4,6-dehydratase from K. pneumoniae requires NADP ϩ (10) as well as the recent work of Sturla et al., who reported that purified E. coli dehydratase contains NADP ϩ .
We have characterized the product of the dehydratase reaction, GDP-4-keto-6-deoxy-mannose. This product was first reported by Ginsberg (6) for the bacterial enzyme and later by Overton and Serif (27) for the porcine enzyme. Using purified dehydratase, we have analyzed the reaction products by both HPLC and paper chromatography and find results consistent with the expected product. For unambiguous structural confirmation, we isolated and analyzed the intermediate by ESI-MS and high field NMR. Mass spectrometric analysis yielded a molecular mass consistent with GDP-4-keto-6-deoxymannose but no evidence of any other GDP-mannose derivatives of different masses. High field NMR revealed the presence of two compounds, one being the expected product GDP-4-keto-6-deoxymannose and the other the related GDP-3-keto-6-deoxymannose. This was the case for the product of both the human and E. coli dehydratases. The presence of both the 4-keto-and 3-keto-6-deoxy-sugars was also seen in the dTDP-rhamnose pathway where a 4,6-dehydratase converts dTDP-glucose to dTDP-4-keto-6-deoxy-glucose (31,32). The biological significance of both GDP-4-keto-6-deoxymannose and GDP-3-keto-6deoxymannose intermediates is unclear, although apparently both are epimerized and reduced to GDP-fucose by the epimerase reductase (Fig. 8, C and F). It is possible that the 3-ketosugar arose during work-up of the isolated reaction product. However, the isolated, unlabeled intermediate also was converted to GDP-fucose by purified human epimerase-reductase (data not shown).
Cells from two patients having LADII do not fucosylate their cell surfaces (4) and as such lack both blood group antigens and the sialyl Lewis X and related epitopes that function as selectin ligands. The molecular basis of this disorder is still unknown. As with Lec13, this phenotype can be rescued in cell lines derived from these patients by culturing them in the presence of fucose, suggesting that GDP-fucose transport and the complement of fucosyltransferases are intact in these cells. 2 This would imply that the defect in the LADII patients lies in the pathway converting GDP-mannose to GDP-fucose, i.e. either the dehydratase or epimerase-reductase. With the human genes for both enzymes now cloned, we can determine if either is responsible for the LADII phenotype. There are suggestions from E. coli that there may be an additional gene that plays a role in GDP-fucose biosynthesis in vivo. Sequencing of the capsular polysaccharide operon in E. coli led to the identification of an open reading frame (wcaH) immediately downstream of the dehydratase, gmd, and epimerase-reductase genes, wcaG (21). This putative protein has been assigned to the GDP-fucose biosynthetic pathway, yet it is clearly not necessary for conversion of GDP-mannose to GDP-fucose in vitro. WcaH may play a role in GDP-fucose biosynthesis in vivo or play another, yet unidentified, role in capsular polysaccharide biosynthesis. The cloning of the GDP-mannose 4,6-dehydratase gene provides a valuable tool to address outstanding questions in the regulation, biosynthesis, and role of GDP-fucose in vivo.