A Bifunctional Enzyme with L-Fucokinase and GDP-L-fucose Pyrophosphorylase Activities Salvages Free L-Fucose in Arabidopsis*

Monomeric sugars generated during the metabolism of polysaccharides, glycoproteins, and glycolipids are imported to the cytoplasm and converted to respective nucleotide sugars via monosaccharide 1-phosphates, to be reutilized as activated sugars. Because l-fucose (l-Fuc) is activated mainly in the form of GDP derivatives in seed plants, the salvage reactions for l-Fuc are expected to be independent from those for Glc, Gal, l-arabinose, and glucuronic acid, which are activated as UDP-sugars. For this study we have identified, in the genomic data base of Arabidopsis, the gene (designated AtFKGP) of a bifunctional enzyme with similarity to both l-fucokinase and GDP-l-Fuc pyrophosphorylase. Recombinant AtFKGP (rAt-FKGP) expressed in Escherichia coli showed both l-fucokinase and GDP-l-Fuc pyrophosphorylase activities, generating GDP-l-Fuc from l-Fuc, ATP, and GTP as the starting substrates. Point mutations in rAtFKGPs at either Gly133 or Gly830 caused loss of GDP-l-Fuc pyrophosphorylase and l-fucokinase activity, respectively. The apparent Km values of l-fucokinase activity of rAtFKGP for l-Fuc and ATP were 1.0 and 0.45 mm, respectively, and those of GDP-l-Fuc pyrophosphorylase activity for l-Fuc 1-phosphate and GTP were 0.052 and 0.17 mm, respectively. The expression of AtFKGP was detected in most cell types of Arabidopsis, indicating that salvage reactions for free l-Fuc catalyzed by AtFKGP occur ubiquitously in Arabidopsis. Loss-of-function mutants with tDNA insertion in AtFKGP exhibited higher accumulation of free l-Fuc in the soluble fraction than the wild-type plant. These results indicate that AtFKGP is a bifunctional enzyme with l-fucokinase and GDP-l-Fuc pyrophosphorylase activities, which salvages free l-Fuc in Arabidopsis.

highly probable that such salvage reactions utilizing free L-Fuc released during degradation of cell wall polysaccharides and glycoproteins also occur in normal plants; however, little is known about the molecular mechanisms of salvage reactions for free L-Fuc in seed plants.
The salvage pathways convert free monosaccharides to nucleotide sugars as their activated forms and involve two reaction processes, phosphorylation of the monosaccharides and formation of nucleotide sugars from monosaccharide 1-phosphates (monosaccharide 1-Ps) and corresponding nucleoside triphosphates. Whereas Glc, Gal, glucuronic acid (GlcA), galacturonic acid (GalA), and L-arabinose (L-Ara) are activated mainly as UDP derivatives, L-Fuc and Man appear as GDP derivatives in the salvage pathways (9). Recently, we have identified a UDP-sugar pyrophosphorylase (EC 2.7.7.64), designated PsUSP (Pisum sativum UDP-sugar pyrophosphorylase), with broad substrate specificity toward various monosaccharide 1-Ps in pea sprouts. Consistent with the activated forms of monosaccharides, PsUSP catalyzed the formation of UDP-Glc, UDP-Gal, UDP-GlcA, UDP-GalA, and UDP-L-Ara from respective monosaccharide 1-Ps and UTP but failed to convert L-Fuc-1-P and Man-1-P to respective UDP-sugars (13,14). This suggests that free L-Fuc and Man are converted to the respective GDP-sugars by specific enzymes distinct from those acting on Glc, Gal, GlcA, GalA, and L-Ara in the salvage pathways. The enzymes participating in the salvage reactions for free L-Fuc have been well characterized in mammals, particularly in pigs. A pig kidney L-fucokinase (EC 2.7.1.52) specifically converts free L-Fuc to L-Fuc-1-P using ATP as the phosphate donor (15). Based on the amino acid sequence, the L-fucokinase is categorized into the GHMP (Galactokinase, Homoserine kinase, Mevalonate kinase, Phosphomevalonate kinase) family, which includes galactokinase (EC 2.7.1.6) and L-arabinokinase (EC 2.7.1.46) found in seed plants (16,17). GDP-L-Fuc pyrophosphorylase (EC 2.7.7.30) catalyzing the formation of GDP-L-Fuc from L-Fuc-1-P and GTP has also been characterized in the pig kidney (18). By the action of the GDP-L-Fuc pyrophosphorylase, L-Fuc-1-P, possibly formed by the L-fucokinase in intact cells, is converted to GDP-L-Fuc in the presence of GTP. In short, free L-Fuc is probably collaboratively salvaged by L-fucokinase and GDP-L-Fuc pyrophosphorylase in mammalian cells. On the other hand, a bifunctional enzyme with both L-fucokinase and GDP-L-Fuc pyrophosphorylase activities, designated Fkp, has been identified in Bacteroides fragilis, which mammals harbor as a symbiont in their intestines (19). Although several genes have been annotated as GHMP family proteins or nucleotide sugar pyrophosphorylases in the genomes of Arabidopsis and rice, neither L-fucokinase nor GDP-L-Fuc pyrophosphorylase has been identified so far in seed plants.
To clarify the molecular mechanism of free L-Fuc salvage in seed plants, we searched the genomic data base and identified a gene that encodes a protein with similarity to both L-fucokinase and GDP-L-Fuc pyrophosphorylase in Arabidopsis. From the properties of the recombinant protein expressed in Escherichia coli, it can be inferred that the gene product has both L-fucokinase and GDP-L-Fuc pyrophosphorylase activities. Loss-of-function mutants were used to confirm the physiological functions of the gene product in Arabidopsis.
Protein Determination-The concentration of protein was determined by the method of Bradford (22) with bovine serum albumin as the standard. In the analysis of the enzymatic properties, substrate specificities, and kinetic values, the concentration of the recombinant enzyme was determined based on the color intensity of the stain with Coomassie Brilliant Blue R-250 on SDS-PAGE (23) compared with the bovine serum albumin standard.
cDNA Cloning-Total RNA was extracted from 2-week-old Arabidopsis seedlings containing cotyledons, true leaves, and roots. The seedlings were frozen in liquid nitrogen and then homogenized with a mortar and pestle. The RNA was extracted with an Isogen kit (Nippon Gene, Tokyo, Japan) according to the manufacturer's instructions. Single strand cDNA was synthesized from 2 g of total RNA from Arabidopsis seedlings using a reverse transcriptase, ReverTra Ace-␣-(Toyobo, Osaka, Japan), and an oligo(dT) adaptor primer (5Ј-GCGACATCAT-CGAATTCCGATGTTTTTTTTTTTTTTT-3Ј). For cloning of At1g01220 cDNA, a set of primers, At1g01220-F-1 (5Ј-CGAA-TTCATGTCTAAGCAGAGGAAG-3Ј) and At1g01220-R-1 (5Ј-GGAATTCAAATACAGATGCTCC-3Ј), was designed based on the genomic data base of Arabidopsis. The PCR was performed with proofreading polymerase (KOD-Plus, Toyobo) and the set of primers, using the single strand cDNA as a tem-plate under the following conditions: 0.5 min denaturing at 94°C, 0.5 min annealing at 55°C, and 3.0 min amplification at 68°C, 35 cycles. The amplified cDNA fragment was treated with ExTaq at 72°C for 5 min to allow the subsequent cloning into a pGEM T-Easy vector (Promega, Madison WI). Point mutations were introduced at Gly 133 and Gly 830 of At1g01220 into cDNA by PCR using sets of primers, FKGP-PM1-F (5Ј-AGGTGACTCCAAAAGGGTTC-3Ј) and FKGP-PM1-R (5Ј-GCAGCATGAAGCATCAATAC-3Ј), and FKGP-PM2-F (5Ј-TCTAGGAACCTCGAGCATTC-3Ј) and FKGP-PM2-R (5Ј-GCACTGCCACGAGGAACATTG-3Ј), respectively. The nucleotide sequence of the cloned fragment was determined with an ABI PRISM 310 genetic analyzer (Applied Biosystems, Foster City, CA).
Expression and Purification of Recombinant Enzyme-The coding region of the cloned At1g01220 cDNA was excised from the pGEM T-Easy vector and then subcloned into an EcoRI site of a pET32a expression vector (Novagen, Madison, WI) to give rise to a plasmid construct, At1g01220/pET32a, which was designed to express the recombinant enzyme fused to thioredoxin and His 6 tags at the N terminus. The plasmid construct was transfected into the BL21 (DE3) gold strain of E. coli (Stratagene, La Jolla, CA). The cells were grown at 10°C, and the recombinant protein was induced by treatment with 1.0 mM isopropyl ␤-D-thiogalactopyranoside for 24 h. To stabilize the recombinant protein in E. coli, chaperone proteins, GroES and GroEL, were co-expressed using the pGro7 plasmid (Takara Bio Inc.). The cells were harvested and lysed in a buffer containing 50 mM Tris-HCl buffer, pH 8.0, 0.2 M NaCl, 20% (w/v) glycerol, 1.0 mM phenylmethylsulfonyl fluoride (PMSF), and 0.2% lysozyme from chicken egg (Wako, Osaka, Japan). The lysate was put on a 1.5 ϫ 6.0-cm chelating Sepharose FF column (GE Healthcare, Tokyo, Japan) that had been equilibrated with the buffer without PMSF and lysozyme. The column was washed with 50 mM Tris-HCl buffer, pH 7.5, 0.2 M NaCl, 20% (w/v) glycerol, and 50 mM imidazole, and the bound protein was then eluted with the same buffer containing 250 mM imidazole. The purified recombinant enzyme (0.2 mg from a 500-ml culture of E. coli) was digested with 1.0 unit of thrombin (Novagen) at 22°C for 24 h to split off the fused thioredoxin and His 6 tags. The purified recombinant enzyme was examined for its purity, properties, and substrate specificity.
Enzyme Assay-The L-fucokinase activity of the recombinant enzyme was determined by monitoring the formation of L-Fuc-1-P in the presence of L-Fuc and ATP. The reaction mixture contained 50 mM Tris-HCl buffer, pH 7.5, 2 mM MgCl 2 , 10 mM L-Fuc, 2 mM ATP, 5% (w/v) glycerol, and enzyme in a final volume of 0.2 ml. After incubation at 35°C, the reaction was terminated by dipping the mixture in a boiling water bath for 2 min; this assay method was defined as the standard assay condition for L-fucokinase activity. The reaction products were detected using high performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) system equipped with a CarboPac PA-1 column (4 ϫ 250 mm, Dionex Japan, Osaka, Japan). The column was eluted with a linear gradient of 0.1-0.2 M sodium acetate in 0.1 M NaOH (0 -20 min), followed by isocratic elution with 0.5 M sodium acetate in 0.1 M NaOH (20 -30 min) at a flow rate of 1 ml per min. The amount of L-Fuc-1-P produced was estimated from the peak area based on the curve of standard L-Fuc-1-P. One unit of L-fucokinase activity was defined as the amount capable to produce 1 mol of L-Fuc-1-P from L-Fuc and ATP per min. Elution times of standard sugars were 2.3 min for L-Fuc and 10.3 min for L-Fuc-1-P (supplemental Fig. S1A).
The GDP-L-Fuc pyrophosphorylase activity of the recombinant enzyme was determined using a reaction mixture consisting of 50 mM MOPS-KOH buffer, pH 7.0, 2 mM MgCl 2 , 0.25 mM L-Fuc-1-P, 1 mM GTP, and enzyme in a final volume of 40 l. After incubation at 35°C, the reaction was terminated as above (this was defined as the standard assay condition for GDP-L-Fuc pyrophosphorylase activity), and the reaction products were analyzed according to the method described previously (13,24). The reaction products were applied to a high performance liquid chromatography (HPLC) system with a Shimadzu LC-10AD equipped with a CarboPac PA-1 column. The column was eluted with 0.05 M sodium acetate for the initial 2 min, then by a linear gradient of 0.05-1.0 M sodium acetate (2-40 min), followed by isocratic elution with 1 M sodium acetate (40 -46 min) at a flow rate of 1 ml per min. The reaction products were monitored by measuring absorbance at 262 nm, and the amount of GDP-sugar produced was estimated from the peak area based on the curve of standard GDP-L-Fuc. Elution times of the standards were as follows: GDP-L-Fuc, 30.2 min; GTP, 38.9 min (supplemental Fig. S1B). One unit of GDP-L-Fuc pyrophosphorylase activity was defined as the amount capable of producing 1 mol of GDP-L-Fuc from L-Fuc-1-P and GTP per min. The assay for pyrophosphorolysis activity of GDP-L-Fuc (reverse direction of the enzyme action) was performed in a 0.1-ml reaction mixture containing 50 mM MOPS-KOH buffer, pH 7.0, 2 mM MgCl 2 , 1 mM GDP-L-Fuc, 1 mM PP i , and enzyme. Here, the enzyme activity was measured in a manner analogous to the above with GTP as the calibration standard and expressed as mol of GTP produced per min.
Successive reactions of L-fucokinase and GDP-L-Fuc pyrophosphorylase activities were carried out using a reaction mixture consisting of 50 mM Tris-HCl buffer, pH 7.5, 1 mM MgCl 2 , 20 mM L-Fuc, 2 mM ATP, 2 mM GTP, 5% (w/v) glycerol, and enzyme in a final volume of 100 l. After incubation at 35°C, the reaction was terminated, and the reaction products were analyzed as described above. To identify GDP-L-Fuc (the reaction product) formed from L-Fuc, ATP, and GTP by the recombinant enzyme, the product was purified by chromatography on a charcoal column (Wako), paper chromatography, then HPLC (the details of product preparation are given in the legend of supplemental Fig. S2), and analyzed by NMR spectroscopy. The sample was dissolved in D 2 O, and 1 H NMR spectra were recorded at 400 MHz and at room temperature with a Bruker DPX-400 spectrometer. HDO was used as the internal standard (4.78 ppm). Ring-proton assignments in NMR were made by first-order analysis of the spectra and confirmed by H-H COSY experiments.
Determination of Free Monosaccharide Contents in Plants-A soluble fraction containing free monosaccharides was prepared from Arabidopsis plants grown for 35 days. Aerial parts of the plants were homogenized with a mortar and pestle in water. The homogenate was centrifuged at 10,000 ϫ g for 5 min, and the supernatant was collected as the soluble fraction. To recover free monosaccharides completely, the precipitate was homogenized again and centrifuged, and the resulting supernatant was combined with the soluble fraction. The soluble fraction was applied to a 1.5 ϫ 2.0-cm DEAE-Sephadex A-25 (GE Healthcare) column. Neutral monosaccharides were eluted with 40 ml of water, concentrated by an evaporator, and then analyzed by HPAEC-PAD as described previously (25). Separation of standard monosaccharides is shown in supplemental Fig. S3.

Identification of Bifunctional L-Fucokinase/GDP-L-fucose
Pyrophosphorylase in the Arabidopsis Genome-To elucidate the salvage pathway for free L-Fuc in plants, we searched for open reading frames (ORFs) with similarity to the mouse L-fucokinase gene (Mus musculus, AJ297482) (26) in the Arabidopsis genome. The BLAST search revealed that the protein encoded by ORF At1g01220 has the highest similarity (23% identical) with murine L-fucokinase. Although Arabidopsis contains several ORFs other than At1g01220 encoding GHMP kinase family proteins, including galactokinase (GAL1, At3g06580) (27) and L-arabinokinase (ARA1, At4g16130) (28), these ORFs did not show significant similarity to the L-fucokinase sequence. We also used the amino acid sequence of human (Homo sapiens) GDP-L-Fuc pyrophosphorylase (AF017445) (18) for a BLAST search of the Arabidopsis genome. The protein encoded by At1g01220 appears to possess significant similarity (22% identical) to the GDP-L-Fuc pyrophosphorylase if one considers only the N-terminal half-region (from Trp 34 to Leu 565 ). Other ORFs possibly encoding nucleotide sugar pyrophosphorylases were less similar to the human GDP-L-Fuc pyrophosphorylase sequence. Because At1g01220 encodes a protein that shares similarities with both L-fucokinase and GDP-L-Fuc pyrophosphorylase, we designated it as AtFKGP (Arabidopsis thaliana L-fucokinase/GDP-L-Fuc pyrophosphorylase). Although the sequence similarity to the murine L-fucokinase was observed over nearly all of AtFKGP, that to the human GDP-L-Fuc pyrophosphorylase was limited to the N-terminal half-region. At1g01220 also exhibited low but significant sequence similarity (17% identical) to a bacterial bifunctional L-fucokinase/GDP-L-Fuc pyrophosphorylase, Fkp, from B. fragilis (19).
The cDNA for AtFKGP was isolated by reverse transcription-PCR using cDNA from Arabidopsis seedlings as the template, and the nucleotide sequence of AtFKGP was determined (supplemental Fig. S4). The comparison of the cDNA sequence with the genomic sequence of Arabidopsis (TAIR) revealed that the AtFKGP gene consists of seven exons and six introns, which is in accordance with the computer-based annotation by the Arabidopsis Genome Initiative. AtFKGP encodes a polypeptide of 1,055 amino acids with a calculated molecular mass of 116,351.5 Da and a theoretical isoelectric point (pI) of 5.87. The peptide sequence of AtFKGP did not contain a secretory signal or transmembrane domain, suggesting that AtFKGP exists inside the cells. A BLAST search of the rice (Oryza sativa) genome identified an ORF closely related (62% identical) to AtFKGP, which we designated OsFKGP (Oryza sativa L-fucokinase/GDP-L-Fuc pyrophosphorylase) (supplemental Fig. S4). Based on the amino acid sequence deduced from the cDNA, AtFKGP appeared to contain a region (from Leu 129 to Lys 146 ) similar to the pyrophosphorylase consensus motif (L(X) 2 GXGTXM(X) 4 PK) conserved for nucleotide sugar pyrophosphorylases (13,29). On the other hand, a region from Pro 826 to Ser 835 of AtFKGP was highly similar to the known ATP-binding motif (PX 3 GL(G/S)SSA) that forms a phosphatebinding loop wrapping around the ␤-phosphate group of ATP in GHMP family proteins (Fig. 1A) (30). Both the pyrophosphorylase consensus and the ATP-binding motifs were also conserved in OsFKGP (supplemental Fig. S4 and Fig. 1A), suggesting that these motifs are important for the functions of FKGP proteins in seed plants. Phylogenetic analysis revealed that AtFKGP forms a plant FKGP family together with OsFKGP, which is apparently independent from the vertebrate L-fucokinase family and bacterial Fkp (19) (Fig. 1B).
Properties of the Recombinant Protein-To characterize the enzymatic properties of the gene product of AtFKGP, the cDNA fragment corresponding to the full-length AtFKGP was subcloned into an expression vector, and the recombinant AtFKGP (rAtFKGP) was expressed in E. coli. The recombinant protein was not stable in E. coli cells, accumulating as an inclusion body, but co-expression of the chaperone proteins GroEL and GroES remarkably improved the stability of rAtFKGP. The recombinant protein fused to thioredoxin and His tags was partially purified by chelating chromatography (Table 1) and appeared as a protein with a molecular mass of more than 120 kDa on SDS-PAGE (Fig. 2). Thrombin digestion of the purified rAtFKGP decreased the relative molecular mass by elimination of the fused thioredoxin and His 6 tags. The purified rAtFKGP still contained a considerable amount of other proteins, including GroEL protein expressed by the pGro7 plasmid (asterisk in Fig. 2). The protein band with a relative molecular mass of 70 kDa may correspond to an endogenous DnaK protein of E. coli. Second purification of rAtFKGP by chelating (chelating Sepharose FF) and gel permission (Sephacryl S-200) chromatography was performed, but these proteins could not be removed. The chaperone proteins seem to bind to rAtFKGP, maintaining its conformation.
In the presence of L-Fuc and ATP, rAtFKGP formed L-Fuc-1-P and generated GDP-L-Fuc from L-Fuc-1-P and GTP, whereas in the control experiment, a thioredoxin-His tag pro-tein expressed by the empty pET32a vector showed neither enzyme activity. These results indicate that AtFKGP encodes a protein with L-fucokinase and GDP-L-Fuc pyrophosphorylase activities as expected from the amino acid sequence similarity discussed above. To confirm the bifunctionality of AtFKGP, point mutations were introduced at the Gly 133 and Gly 830 residues important for GDP-L-Fuc pyrophosphorylase and L-fu-cokinase activities of rAtFKGP, respectively, and the effect of the mutations on the activities was examined. The point mutation substituting Ala for Gly 133 (designated PM1) abolished GDP-L-Fuc pyrophosphorylase activity of rAtFKGP but hardly affected the L-fucokinase activity (Fig. 3). On the other hand, the point mutation substituting Ala for Gly 830 (PM2) reduced L-fucokinase activity by more than 90%, while preserving the GDP-L-Fuc pyrophosphorylase activity. The recombinant protein having both PM1 and PM2 lost both activities. These results indicate that AtFKGP has two distinct activities, L-fucokinase and GDP-L-Fuc pyrophosphorylase activity.   The recombinant AtFKGP was stable in 50% (w/v) glycerol, 1 mM PMSF, 0.2 M NaCl, 50 mM Tris-HCl buffer, pH 7.5, at Ϫ20°C for 1 month, but more than half of the GDP-L-Fuc pyrophosphorylase activity was lost within 1 day when it was stored without glycerol at 4°C. Thrombin digestion did not increase but rather decreased specific activity of rAtFKGP ( Table 1), suggesting that the fused protein does not impair the enzyme activity of rAtFKGP. The purified rAtFKGP fused to thioredoxin and His tags was therefore used in the following experiments.
Enzymatic Properties-The enzymatic properties, substrate specificities, and kinetics of rAtFKGP were determined. The specific activity of rAtFKGP for L-fucokinase was estimated at 0.17 unit/mg protein and that for GDP-L-Fuc pyrophosphorylase at 0.71 unit/mg protein (here the protein content was assessed based on the color intensity of rAtFKGP stained with Coomassie Brilliant Blue R-250 on SDS-PAGE). The specific activities of rAtFKGP apparently differ from those (1.27 and 0.36 units/mg protein, respectively) of pig L-fucokinase (15) and GDP-L-Fuc pyrophosphorylase (18). As has been observed for many nucleotide sugar pyrophosphorylases from other sources, rAtFKGP catalyzed both synthesis of GDP-L-Fuc from L-Fuc-1-P and GTP (GTP:L-fucose-1-phosphate guanylyltransferase action, the forward direction), and pyrophosphorolysis producing L-Fuc-1-P and GTP from GDP-L-Fuc and PP i (GDP-L-Fuc pyrophosphorylase action, the reverse).
The L-fucokinase activity of rAtFKGP absolutely required divalent cations such as Mg 2ϩ and Mn 2ϩ , as had the pig L-fucokinase in Ref. 15. The divalent cations were also necessary for the GDP-L-Fuc pyrophosphorylase activity of rAtFKGP (Table  2). Although Mn 2ϩ was a better cation than Mg 2ϩ for the L-fucokinase activity, it had only a poor effect on the GDP-L-Fuc pyrophosphorylase activity of rAtFKGP compared with Mg 2ϩ .
The maximal L-fucokinase activity of the enzyme occurred at pH 10.5, whereas the maximal GDP-L-Fuc pyrophosphorylase activity of the enzyme was observed in the pH range from 6.5 to   8.0 (Fig. 4). Because the optimum pH values for L-fucokinases from mammals are around 8.0 (15,32), and because it is not likely that AtFKGP functions as L-fucokinase at pH 10.5 in Arabidopsis tissues, a reaction mixture buffered to pH 7.5 was used in the assays for L-fucokinase activity of rAtFKGP throughout the study. It is conceivable that the glycine-NaOH buffer used for the determination of the optimum in the high pH range affected the L-fucokinase activity. However, neither glycine nor sodium ion (50 mM) elevated the activity at low pH levels, indicating that the high pH optimum of the L-fucokinase activity of AtFKGP is attributable to the nature of the enzyme protein itself. The optimum temperature for L-fucokinase activity of AtFKGP was 40°C, and the enzyme lost more than 99% of its activity at 50°C (Fig. 5). GDP-L-Fuc pyrophosphorylase activity of rAtFKGP was maximal in the range from 30 to 45°C (Fig. 5). The GDP-L-Fuc pyrophosphorylase activity was thus somewhat more stable than the L-fucokinase activity.
Substrate Specificity of L-Fucokinase Activity-Generally, monosaccharide kinases catalyze phosphorylation of monosac-charides using ATP as the phosphate donor. Our results above show that rAtFKGP possesses L-fucokinase activity, forming L-Fuc-1-P in the presence of L-Fuc and ATP. We examined whether rAtFKGP catalyzes other monosaccharide kinase reactions using 10 mM L-Ara, Glc, Gal, L-Gal, Man, Xyl, L-Rha, GalA, GlcA, and GlcNAc in the presence of 2 mM ATP under standard assay conditions. None of these monosaccharides served as a substrate for rAtFKGP at all. Although L-fucokinase purified from pig kidney shows weak activity on Glc, forming Glc-1-P (15), rAtFKGP did not act on Glc. The substrate specificity of the L-fucokinase activity toward nucleotide phosphate donors was also determined. The enzyme failed to form L-Fuc-1-P, when CTP, GTP, ITP, or UTP was added to the reaction mixture instead of ATP. Although GTP served as a nucleotide donor for the GDP-L-Fuc pyrophosphorylase activity of rAt-FKGP (see below), it did not serve as phosphate donor for the L-fucokinase activity of rAtFKGP.
Substrate Specificity of GDP-L-Fuc Pyrophosphorylase Activity-To determine the substrate specificity of the GDP-L-Fuc pyrophosphorylase activity of rAtFKGP toward nucleotides, the effect of adding ATP, CTP, ITP, or UTP to the reaction mixtures on the formation of GDP-L-Fuc from L-Fuc-1-P and GTP was examined ( Table 3). The formation of GDP-L-Fuc was barely inhibited by these nucleotides, and no product corresponding to nucleotide sugars other than GDP-L-Fuc was detected on HPLC analysis. Addition of PP i , however, strongly inhibited the formation of GDP-L-Fuc by rAtFKGP. The inhibition of nucleotide sugar synthesis by PP i has also been observed for a UDP-Glc pyrophosphorylase from Acanthamoeba castellanii (31) and PsUSP (13), where it is due to product inhibition. These results indicate that rAtFKGP specifically uses GTP as the nucleotide donor.
The specificity of the enzyme toward monosaccharide 1-Ps was determined using the following substrates: L-Ara-1-P, Gal-1-P, Glc-1-P, Man-1-P, Xyl-1-P, GlcA-1-P, and Glc-NAc-1-P, together with L-Fuc-1-P. We found that rAtFKGP fails to utilize monosaccharide 1-Ps other than L-Fuc-1-P as the glycosyl donor. This strict substrate specificity of GDP-L-Fuc pyrophosphorylase toward L-Fuc-1-P has also been observed for enzymes from other origins, such as pig kidney (18). Together with the substrate specificity of L-fucokinase activity of rAtFKGP toward monosaccharides, the specific action of GDP-L-Fuc pyrophosphorylase activity of rAt-FKGP on L-Fuc-1-P indicates that AtFKGP specifically cata-  a GDP-L-Fuc pyrophosphorylase activity of rAtFKGP was determined in the presence of 1 mM of other nucleotides or PP i under standard assay conditions. b Relative activities are expressed as a percentage of that without addition of these compounds.
lyzes conversion of L-Fuc to GDP-L-Fuc via L-Fuc-1-P, and it is not involved in the salvage reactions for other monosaccharides in Arabidopsis. Successive L-Fucokinase and GDP-L-Fuc Pyrophosphorylase Reactions-Generation of GDP-L-Fuc from L-Fuc via L-Fuc-1-P by successive L-fucokinase and GDP-L-Fuc pyrophosphorylase activities of rAtFKGP was performed in a reaction mixture containing L-Fuc, ATP, and GTP as the starting substrates (Fig. 6). The amount of GDP-L-Fuc in the reaction mixture increased proportionally to the reaction time. L-Fuc-1-P also accumulated in the reaction mixture, but the relative amount was very low compared with GDP-L-Fuc. The results indicate that AtFKGP catalyzes two successive reactions, phosphorylation of L-Fuc and generation of GDP-L-Fuc from L-Fuc-1-P and GTP in such a way that L-Fuc-1-P formed by the L-fucokinase activity is immediately consumed as the substrate for the GDP-L-Fuc pyrophosphorylase activity of AtFKGP. The GDP-L-Fuc generated from L-Fuc, ATP, and GTP by the successive reactions was purified and analyzed for its structure by 1 H NMR spectroscopy (supplemental Fig. S1). The signals of a doublet of doublets at ␦ 3.54 ppm (J 2Љ3Љ 9.9 Hz, J 1Љ2Љ , 7.  Table 4. The K m value (1.0 mM) of the L-fucokinase activity of rAtFKGP for L-Fuc was manifestly higher than those (0.027 and 0.12) for known L-fucokinases from pig kidney (15) and pig liver (32), respectively. The kinetic values of GDP-L-Fuc pyrophosphorylase were also determined. The protein exhibited high affinity to L-Fuc-1-P (K m value, 0.052 mM). A similar K m value (0.060) for L-Fuc-1-P has been reported for GDP-L-Fuc pyrophosphorylase from pig kidney (18). The low accumulation of L-Fuc-1-P in the successive reactions ( Fig. 6) may be attributed to the low K m value toward L-Fuc-1-P, which enables the enzyme to rapidly convert L-Fuc-1-P to GDP-L-Fuc. The k cat values of GDP-L-Fuc pyrophosphorylase activity of rAtFKGP for GDP-L-Fuc (7.86 milliunits/mg protein) and PP i (8.02) in the reverse reaction (pyrophosphorolysis of GDP-L-Fuc) were apparently higher than those for L-Fuc-1-P (1.59) and GTP (1.82) in the forward reaction. It would seem that the equilibrium of the nucleotide sugar synthesis and pyrophosphorolysis catalyzed by AtFKGP is controlled by the concentrations of L-Fuc-1-P, GTP, PP i , and GDP-L-Fuc in intact cells.
Expression of AtFKGP in Tissues of Arabidopsis-To analyze the expression pattern of AtFKGP, quantitative analyses of the mRNA were performed in various tissues of Arabidopsis (Fig. 7). The expression of AtFKGP was detected in all tissues examined, indicating that the salvage reactions for free L-Fuc catalyzed by AtFKGP occur ubiquitously in Arabidopsis. Free L-Fuc is likely released from L-Fuc-containing glycoconjugates by the action of ␣-L-fucosidases (EC 3.2.1.51). AtFXG1 is a unique gene known to encode ␣-Lfucosidase in Arabidopsis (35). Whereas the expression of AtFXG1 is reported to be high in young leaves and inflorescence of Arabidopsis (35), relatively high expression of AtFKGP was observed in flower buds.

L-Fuc Accumulation in the Mutant Plants-
The dual activity of rAtFKGP as both L-fucokinase and GDP-L-Fuc pyrophosphorylase suggests that AtFKGP is involved in the salvage of free L-Fuc in intact plants. To address the physiological functions of AtFKGP in Arabidopsis, L-Fuc content in the plant tissues was measured in loss-of-function mutants of AtFKGP. There are two independent mutants with tDNA insertion in the AtFKGP gene, namely SALK-012400 (designated fkgp-1) and SALK-053913 (fkgp-2), in which the insertions occurred in the third and sixth exons of the AtFKGP gene, respectively (Fig. 8A). The genotypes of the mutants were confirmed by genomic PCR using specific primers (see "Experimental Procedures"), and homozygous lines of fkgp mutants (fkgp-1#1, fkgp-1#16, fkgp-2#1, and fkgp-2#14) were isolated. The content of free L-Fuc in the soluble fraction extracted from the Arabidopsis plants was determined by HPAEC-PAD. Both fkgp-1 and -2 mutants appeared to accumulate more than 40 times as much free L-Fuc in the soluble fraction as the wild-type plants (Fig. 8B), whereas the content of the other monosaccharides (Glc, Gal, and L-Ara) detected in this experiment was hardly affected by the mutations (supplemental Fig. S3). The results indicate that AtFKGP plays central roles in the salvage of free L-Fuc in Arabidopsis. Although the mutations abolish the function of AtFKGP completely, thus causing strong accumulation of free L-Fuc, the fkgp mutants did not show any altered visible phenotype. Moreover, the sugar composition of cell wall polysaccharides was not affected by the fkgp mutations (data not shown), whereas it has been reported that the mur1 mutation results in a considerable reduction of L-Fuc content in cell wall polysaccharides (36). These results suggest that GDP-L-Fuc is mainly supplied through the de novo pathway, as reported for mammalian cells (26), and that the services of AtFKGP are not necessarily required for the synthesis of L-Fuc-containing cell wall polysaccharides in Arabidopsis, at least under normal growth conditions.

DISCUSSION
Free L-Fuc released during metabolism of glycoconjugates is imported and converted to GDP-L-Fuc via L-Fuc-1-P as an intermediate metabolite in the salvage pathway in mammals (15,18). In the symbiotic bacterium B. fragilis, GDP-L-Fuc is generated from exogenous L-Fuc via L-Fuc-1-P by a bifunctional L-fucokinase/GDP-L-Fuc pyrophosphorylase, Fkp (19). L-Fucokinase and GDP-L-Fuc pyrophosphorylase activities of rAtFKGP expressed in E. coli and high accumulation of free L-Fuc in fkgp mutants demonstrate that seed plants also possess a salvage pathway for free L-Fuc similar to that of mammals and the bacterium. AtFKGP has weak but significant similarities to both mammalian L-fucokinase and GDP-L-Fuc pyrophosphorylase and contains ATP-binding and pyrophosphorylase consensus motifs. AtFKGP also has significant similarity to Fkp. These facts suggest that the AtFKGP gene has evolved from an ancestral gene common to mammalian L-fucokinase and bacterial Fkp genes. It is possible that the mammalian L-fucokinases have then lost their GDP-L-Fuc pyrophosphorylase activity in the evolutional process, whereas plant FKGP proteins and bacterial Fkp preserved both activities. The low sequence sim-