An FMN Hydrolase Is Fused to a Riboflavin Kinase Homolog in Plants*

Riboflavin kinases catalyze synthesis of FMN from riboflavin and ATP. These enzymes have to date been cloned from bacteria, yeast, and mammals, but not from plants. Bioinformatic approaches suggested that diverse plant species, including many angiosperms, two gymnosperms, a moss (Physcomitrella patens), and a unicellular green alga (Chlamydomonas reinhardtii), encode proteins that are homologous to riboflavin kinases of yeast and mammals, but contain an N-terminal domain that belongs to the haloacid dehalogenase superfamily of enzymes. The Arabidopsis homolog of these proteins was cloned by RT-PCR, and was shown to have riboflavin kinase and FMN hydrolase activities by characterizing the recombinant enzyme produced in Escherichia coli. Both activities of the purified recombinant Arabidopsis enzyme (AtFMN/FHy) increased when the enzyme assays contained 0.02% Tween 20. The FMN hydrolase activity of AtFMN/FHy greatly decreased when EDTA replaced Mg2+ in the assays, as expected for a member of the Mg2+-dependent haloacid dehalogenase family. The functional overexpression of the individual domains in E. coli establishes that the riboflavin kinase and FMN hydrolase activities reside, respectively, in the C-terminal (AtFMN) and N-terminal (AtFHy) domains of AtFMN/FHy. Biochemical characterization of AtFMN/FHy, AtFMN, and AtFHy shows that the riboflavin kinase and FMN hydrolase domains of AtFMN/FHy can be physically separated, with little change in their kinetic properties.

The cofactors FMN and FAD participate in numerous vital processes in all organisms. Some of these processes are mitochondrial electron transport, photosynthesis, fatty acid oxidation, and metabolism of vitamins B6, B12, and folates. FMN and FAD are respectively synthesized by the enzymes riboflavin kinase (EC 2.7.1.26) and FAD synthetase (EC 2.7.7.2) in the presence of ATP and Mg 2ϩ . Bifunctional enzymes with riboflavin kinase and FAD synthetase activities have been characterized and cloned from bacteria (1)(2)(3)(4). Biochemical characterization of the Corynebacterium ammoniagenes enzyme has established that phosphorylation of riboflavin to FMN is essentially irreversible, and that adenylylation of FMN to FAD is readily reversible (1). Homologs of bifunctional enzymes with riboflavin kinase and FAD synthetase activities from many other bacterial species exist in GenBank TM , which suggests that this type of enzyme may be ubiquitous among bacteria. Yet, this is not the only type of bacterial riboflavin kinase; monofunctional enzymes from Bacillus subtilis and Streptococcus agalactiae have been cloned and characterized (5,6).
Only monofunctional enzymes with riboflavin kinase or FAD synthetase activity have so far been studied and cloned in eukaryotes. Both enzymes have been purified from rat tissues and biochemically characterized (7)(8)(9)(10)(11). Three monofunctional riboflavin kinases (one from mammals and two from yeast), with sequence homology to the C-terminal domains of the bacterial bifunctional enzymes, have been cloned (12)(13)(14). The Saccharomyces cerevisiae enzyme resides in microsomes and the inner mitochondrial membrane (12). Crystal structures have been determined for riboflavin kinases from humans and Schizosaccharomyces pombe (13)(14)(15). Fad1p from S. cerevisiae is the only eukaryotic FAD synthetase that has been cloned (16). This enzyme shares little or no sequence similarity to the bacterial bifunctional enzymes with riboflavin kinase and FAD synthetase activities and probably resides in the cytosol.
Although FMN and FAD play vital roles in metabolism, little is known about the enzymes that synthesize these cofactors in plants. Monofunctional riboflavin kinases or FAD synthetases have been assayed in various plant species (17)(18)(19)(20), and a monofunctional riboflavin kinase has been purified from mung bean (17). However, no plant riboflavin kinases or FAD synthetases have been cloned and fully characterized. Subcellular localization of these enzymes has not been investigated, except for a single study showing riboflavin kinase activity in the cytosol and in an organellar fraction containing chloroplasts and mitochondria in spinach (21).
Little is known about membrane transport of riboflavin and flavin nucleotides in plants. It is known that plants can transport flavins across membranes only from the finding that some plant species excrete flavins into the growth medium because of iron deficiency (38,39). Membrane transport of riboflavin and flavin nucleotides has been observed in other eukaryotes and in prokaryotes. Uptake of riboflavin from medium has been studied in B. subtilis, in riboflavin-deficient mutants of S. cerevisiae, and in some mammalian cell types (40 -42). Excretion of riboflavin into growth medium is a well known phenomenon in yeast, fungi, and flavinogenic bacteria. This phenomenon enabled the development of fermentation methods for commercial production of riboflavin as alternatives to chemical synthesis (43). Both import and export of riboflavin across the vacuolar membrane exist in the filamentous fungus Ashyba gossypii (44). Import and export of FAD have been investigated in rat liver microsomes (45). S. cerevisiae mitochondria can import riboflavin, FMN, and FAD; they can also export FAD, presumably through the Flx1p transporter (46,47).
In plants, we know little about the enzymes responsible for turnover of FMN and FAD, and their subcellular localization, despite the crucial roles of flavin nucleotides in metabolism. Hence, we searched protein and DNA sequence databases for homologs of monofunctional and bifunctional enzymes with riboflavin kinase and FAD synthetase activities. Based on these bioinformatic data, we propose a model for the subcellular distribution of these enzymes and for the transport of flavins across membranes inside plant cells. We cloned and characterized a homolog of monofunctional riboflavin kinases from Arabidopsis. Our results show that in this model plant, the homolog of monofunctional riboflavin kinase is fused to an FMN hydrolase that belongs to the haloacid dehalogenase (HAD) 2 superfamily of enzymes.
Plants and Growth Conditions-Arabidopsis thaliana plants (ecotype Columbia) were grown for 3 weeks in 12-h days (photosynthetic photon flux density 80 E m Ϫ2 s Ϫ1 ) in potting soil at 23°C.
cDNA Cloning, Constructs, Sequence Analysis, and Expression in E. coli-To clone the AtFMN/FHy cDNA, total Arabidopsis RNA was isolated using the RNeasy Plant Mini Kit (Qiagen, Valencia, CA), and reverse-transcribed using Superscript II reverse transcriptase (Invitrogen) and an oligo(dT) primer. The AtFMN/FHy open reading frame was then amplified using Taq2000 polymerase (Stratagene, La Jolla, CA) and the primers 5Ј-GACGACGACAAGATGTCGATGAGCAATTC-3Ј (HKForward) and 5Ј-GAGGAGAAGCCCGGTTCATTTAGTCA-GATAAGGAT-3Ј (HKReverse). The generated PCR fragment was purified using a Wizard PCR column (Promega) and cloned into the pGEM-T Easy vector (Promega). The primers comprised gene-specific sequences flanked by vector-specific sequences needed for cloning into pET Ek/LIC expression vectors (Novagen). (The vector-specific sequences are underlined.) The AtFMN/FHy ORF was reamplified from the generated construct using Pfu polymerase (Stratagene) and the primers HKForward and HKReverse. The generated PCR fragment was purified using a Wizard PCR column, treated with T4 polymerase, and ligated into pET-30 Ek/LIC following the manufacturer's protocol. To test if riboflavin kinase and FMN hydrolase activities reside in separate domains, the N-terminal (AtFHy, amino acids 1-234) and C-terminal (AtFMN, amino acids 227-379) domains of AtFMN/FHy were subcloned into the expression vector pET-30 Ek/LIC. The sequence encoding the N-terminal domain of AtFMN/FHy was amplified from the full-length AtFMN/FHy cDNA using Pfu polymerase and the primers HKForward and HReverse (5Ј-GAGGAGAAGCCCGGTTCATA-AAGTGTTCTCTATCCAGT-3Ј). The generated PCR fragment was purified using a Wizard PCR column, treated with T4 polymerase, and ligated into pET-30 Ek/LIC following the manufacturer's protocol. The construct for the overexpression of the C-terminal domain was prepared as described for the N-terminal domain, but with the PCR primers KForward (5Ј-GACGACGACAAGATGCAAGACTGGATA-GAGAACAC-3Ј) and HKReverse. All constructs were verified by sequencing. The pET-30 Ek/LIC constructs above were introduced into the Rosetta strain of E. coli (Novagen) to express the recombinant proteins. Bacteria were cultured at 37°C in LB medium containing 100 g/ml kanamycin and 34 g/ml chloramphenicol. When A 600 reached 0.6 -1, isopropyl-D-thiogalactopyranoside was added to a final concentration of 200 M, and incubation was continued for 4 -6 h at 25°C.
Protein Isolation and Molecular Mass Determination-The recombinant enzymes were first purified by affinity chromatography on Ni 2ϩnitriloacetic acid resin (Novagen) under native conditions. To purify AtFMN/FHy and AtFMN, E. coli cells from 200-ml cultures were harvested by centrifugation (5,000 ϫ g, 15 min) and resuspended in 3 ml of 50 mM potassium phosphate buffer, pH 8.0, containing 300 mM NaCl and 10 mM imidazole (buffer A). To purify AtFHy, buffer A containing 10% glycerol, 0.25% Tween 20, and 1 mM MgCl 2 (buffer AϩGTM) was used. This buffer was also used to purify AtFMN/FHy for the enzyme assays with added 0.02% Tween 20. Subsequent operations were carried out at 0 -4°C. The resuspended cells were broken using the BugBuster reagent (Novagen) as described by the manufacturer. The extracts were cleared by centrifugation (20,000 ϫ g, 20 min), and desalted on a PD-10 column (Amersham Biosciences, Piscataway, NJ) equilibrated either in buffer A or AϩGTM. The affinity purifications were done following the manufacturer's protocols. The obtained 150-mM imidazole eluates were desalted on PD-10 columns equilibrated with 20 mM Tris-HCl, pH 7.5.
After digestion with recombinant enterokinase to release the tags, the affinity-purified enzymes were treated with EKapture agarose to remove enterokinase. AtFMN/FHy was then treated with Ni 2ϩ -nitriloacetic acid resin to remove the undigested enzyme and released tag. The purified enzyme was desalted into 20 mM Tris-HCl, pH 7.5, containing 10% glycerol. When 0.25% Tween 20 and 1 mM MgCl 2 were used during the purification, they were also added to the desalting buffer. AtFMN and AtFHy were purified by ion exchange chromatography using an Ä KTA FPLC system equipped with a Mono Q 5/50 GL column (Amersham Biosciences). To purify AtFMN, the column was equilibrated with 20 mM Tris-HCl, and the bound proteins were eluted with a 0 -0.35 M NaCl gradient. AtFMN eluted at about 180 mM NaCl. Glycerol (10% final concentration) was added to the eluate fractions before storage. To purify AtFHy, the column was equilibrated with 20 mM Tris-HCl, containing 10% glycerol, 0.25% Tween 20, and 10 mM MgCl 2 . AtFHy eluted in the flow-through. The eluate fractions were frozen in liquid N 2 and stored at Ϫ80°C until use. Freezing did not affect the enzyme activity.
The purification of AtFHy in buffer A resulted in very low yields because of nonspecific proteolytic cleavage by enterokinase and because of the separation of the enzyme into numerous peaks during ion exchange chromatography. The purification of AtFHy (and AtFMN/ FHy) in buffer AϩGTM improved yield and purity.
Enzyme Assays-Riboflavin kinase and FMN hydrolase activities were measured using an Alliance HPLC system with a fluorescence detector (Waters, Milford, MA). Concentrations of riboflavin and FMN substrates were determined spectrophotometrically (12). Unless otherwise indicated, the procedures described below were used. Initial velocity data at steady state were measured. Substrates were saturating, and product formation was proportional to enzyme concentration and time. Less than 5% of the substrates were typically consumed. Exceptions were those riboflavin kinase enzyme assays (to determine the kinetic parameters) that were carried out in the presence of Ͻ5 nM riboflavin. These assays consumed 10 -15% of riboflavin, as the fluorescence measurements were close to the detection limit. Final assay volumes were 100 or 250 l. As reported elsewhere (2), riboflavin kinase activity was assayed in 100 mM potassium phosphate buffer, pH 7.5, containing 1 mM dithiothreitol, 15 mM MgCl 2 , 10 mM Na 2 SO 3 , 3 mM ATP, and 1 M riboflavin. The catalytic efficiency (k cat /K m ) of this activity was virtually identical in phosphate and Tris-HCl buffers at pH 7.5 (see TABLE  TWO). Phosphate buffer was selected for future experiments because the K m value of AtFMN/FHy for riboflavin was slightly lower in this buffer than in Tris-HCl. FMN hydrolase activity was assayed in 50 mM Tris-HCl, pH 7.5, containing 1 mM dithiothreitol, 10 mM MgCl 2 , and 100 M FMN. Phosphate buffer was not used because preliminary results (not shown) suggested that the FMN hydrolase activity of AtFMN/FHy is inhibited by phosphate. Tween 20 (0.02%) was added to these assays when indicated under "Results." Tween 20 in the 0.01-0.20% range gave similar results.
After incubation at 30°C for 15-20 min, the assays were stopped by adding saturated formic acid (1:20 of final assay volume), and centrifuged to remove the precipitated protein. The flavins were separated by HPLC using a Waters NovaPak C18 column (3.9 ϫ 150 mm), and measured with a Waters 2475 fluorescence detector. Excitation and emission wavelengths were 470 and 530 nm, respectively. The mobile phase contained 25% methanol, 100 mM formic acid, and 100 mM ammonium formate (2). Product formation was determined from fluorescence relative to a blank wherein the enzyme was added after incubation. The kinetic parameters (K m , k cat ) were calculated from Hanes plots. The standard error for k cat /K m was calculated by error propagation (49).
Phylogenetic Analyses-We selected the protein sequences for the phylogeny as follows. For the HAD domains, similarity searches of the GenBank TM protein data base were conducted using the BLASTP program and using AtFHy (residues 18 -166) as a query sequence. This search resulted in 501 protein sequences. After setting a cutoff value of 1ϫ10 Ϫ7 , 204 protein sequences remained. Next, we selected representative prokaryotic and eukaryotic sequences for the phylogeny. Protein sequences from one fully sequenced species per phylum were selected for nine phyla of bacteria (E. coli K12, Proteobacteria; Synechococcus elongates PCC 6301, Cyanobacteria; B. subtilis, Firmicutes; Propionibacterium acnes KPA 171202, Actinobacteria; Dehalococcoides ethanogenes, Chloroflexi; Deinococcus radiodurans, Deinococcus-Thermus; Thermotoga maritima, Thermotogae; Chlorobium tepidum, Chlorobi; Rhodopirellula baltica, Planctomycetes). The sequence of ␤-phosphoglucomutase (␤-PGM) from Lactococcus lactis was the most similar to AtFHy among the biochemically characterized enzymes. This ␤-PGM was therefore used as the outgroup for the phylogenetic analyses despite the score value of 2ϫ10 Ϫ4 . The eukaryotic sequences were selected from representative organisms that belong to the kingdoms Metazoa (Homo sapiens, Drosophila melanogaster, and Caenorhabditis elegans), Fungi (S. cerevisiae and S. pombe), and Viridiplantae (A. thaliana); and from a single cellular organism Entamoeba histolytica, classified as Entamoebidae (no rank). To investigate the phylogenetic relationship between the plant FHy domains and the enzymes listed above, we included the FHy protein sequences from Oryza sativa (monocot), Picea glauca (gymnosperm), Physcomitrella patens (moss), and C. reinhardtii (unicellular green alga) in the analysis. The sequences from P. glauca and P. patens were obtained by translating the corresponding ESTs. This selection process resulted in 36 protein sequences, which were further processed to eliminate the N-terminal and C-terminal protrusions and were aligned using ClustalW. To investigate the phylogenetic relationships among riboflavin kinases from different organisms, the protein sequences from the organisms listed above (except L. lactis) were obtained using BLASTP and AtFMN as a query sequence. This selection process resulted in 22 protein sequences, which were then further processed as described before.
For the phylogeny of the haloacid dehalogenases, we included the sequences: Arabidopsis The sequences BJ187273 and BJ593554 from P. patens were obtained from an EST that was sequenced from the 5Ј-and 3Ј-end, respectively. The sequence C_90019 was from the DOE Joint Genome Institute Eukaryotic Genomics data base (genome.jgi-psf.org).
The phylogenetic trees were assembled using the Phylip (3.6.3) package programs for MacOS X (50). First, 1,000 bootstrap samples were generated using the Seqboot program. Second, the protein distance matrices were calculated from the Seqboot outputs using the Protdist program. Third, the phylogenetic trees were generated from the Protdist outputs using the Neighbor program. Fourth, the consensus phylogenetic tree was generated from the Neighbor outputs using the Consense program. Last, the consensus tree was drawn using the Drawtree program.

RESULTS
Bioinformatic Sequence Analyses-Two types of riboflavin kinases and FAD synthetases have been cloned to date: monofunctional enzymes from mammals and yeast, and bifunctional enzymes from prokaryotes (7)(8)(9)(10)(11)(12)(13)(14)(15). To explore the presence of these enzymes in plants, we conducted BLAST searches of the Arabidopsis data base using the protein sequences of both monofunctional and bifunctional enzymes with riboflavin kinase and FMN hydrolase activities. NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 BLAST searches using the protein sequence of monofunctional riboflavin kinase (Fmn1p) from S. cerevisiae revealed a single homolog (GenBank™ AAP21181) encoded by the gene At4g21470 (TABLE ONE). Unexpectedly, the deduced Arabidopsis protein has an N-terminal extension of 225 residues relative to Fmn1p. This extension has FMN hydrolase activity (see below), so the enzyme encoded by the gene At4g21470 was named riboflavin kinase/FMN hydrolase (AtFMN/FHy).

Plant Riboflavin Kinase-FMN Hydrolase
BLAST searches using the protein sequence of monofunctional FAD synthetase (Fad1p) from S. cerevisiae revealed a single sequence homolog encoded by the gene At5g03430 (TABLE ONE). The searches using the bifunctional enzyme with riboflavin kinase and FAD synthetase activities from B. subtilis revealed two sequence homologs encoded by the genes At5g23330 and At5g08340 (TABLE ONE). Proteins encoded by these genes have N-terminal extensions with characteristics of organellar targeting peptides.
Searches of EST databases using TBLASTN and the Arabidopsis sequences listed in TABLE ONE, showed that diverse plant species have sequence homologs of monofunctional enzymes with riboflavin kinase or FAD synthetase activity, as well as sequence homologs of bifunctional enzymes with both activities (not shown). The apparent presence of both classes of riboflavin kinases and FAD synthetases separates plants from other eukaryotes.
BLAST searches of ESTs from GenBank TM showed that diverse plant species, including many angiosperms, two gymnosperms, a moss (P. patens), and a unicellular green alga (C. reinhardtii), have putative FMN/FHy enzymes. (For representative sequences, see Fig. 1.) The presence of these bipartite enzymes in angiosperms, gymnosperms, a moss, and a unicellular green alga suggests that they evolved before the speciation of vascular plants.
The N-terminal domains of the FMN/FHy enzymes share sequence similarity to members of the HAD superfamily of enzymes. This superfamily is large and diverse, yet few members have known biochemical functions (51)(52)(53)(54). Of those that do, ␤-phosphoglucomutase from bacterium L. lactis (55) (Fig. 1, A and B) is the most closely related to the N-terminal HAD domains of the FMN/FHy enzymes (see Phylogenetic Analyses under "Experimental Procedures").
Members of the HAD superfamily have low (Ͻ15%) sequence identity (53), but have three conserved motifs (52). The first two motifs are hhhhDXXG(T/V) and hhhh(T/S), where h is a hydrophobic residue. The third motif consists of an N-terminal lysine and a C-terminal pair of aspartates, separated by 3-7 residues. All members of the HAD superfamily have the lysine and at least one of the aspartates. The N-terminal domains of the FMN/FHy enzymes contain all three motifs (Fig. 1A), indicating that these domains belong to the HAD superfamily.
Furthermore, the FMN/FHy enzymes from two gymnosperms, P. glauca (Fig. 1A) and Pinus taeda (EST CF672665, not shown), have extra ϳ25-amino acid N-terminal extensions with characteristics of mitochondrial targeting peptides. These extensions are not found in the FMN/FHy sequences from angiosperms and C. reinhardtii. Thus, gymnosperms may contain FMN/FHy enzymes in organelles. We cannot predict the presence of non-organellar FMN/FHy enzymes in these plants because no gymnosperm genome has been fully sequenced yet. Neither can we predict subcellular localization of FMN/FHy sequences from mosses and ferns, because few sequences from these plants are now available.
Eukaryotes and bacteria contain many uncharacterized enzymes that are related to the HAD domains of the FMN/FHy enzymes, and to the ␤-phosphoglucomutase from L. lactis (Fig. 1B). Some of these enzymes may catalyze hydrolysis of FMN. However, all the non-plant eukaryotic HAD enzymes that are represented on the tree cluster apart from the FHy domains, as does also a different set of plant HAD enzymes (Fig. 1B,  At1-4). Likewise, all the prokaryotic HAD enzymes cluster apart from the FHy domains, as do also At5 and the ␤-phosphoglucomutase from L. lactis. This suggests that the FHy enzymes might be unique to the plant lineage. In contrast, the riboflavin kinase domains of the FMN/ FHy enzymes appear to be closely related to the monofunctional riboflavin kinases from other eukaryotes (Fig. 1D). Taken together, these results suggest that the FMN/FHy enzymes originated early in plant evolution by fusion of a haloacid dehalogenase to a eukaryotic-type riboflavin kinase.
Cloning and Expression of AtFMN/FHy, AtFMN, and AtFHy in E. coli-Full-length cDNA for AtFMN/FHy was cloned by RT-PCR using mRNA isolated from stems as a template. This cDNA was then amplified by PCR, subcloned into the pET-30 Ek/LIC vector, and functionally expressed in E. coli. Despite the N-terminal tag attached to the recombinant protein, the desalted extracts of cells harboring the AtFMN/FHy cDNA had much higher riboflavin kinase and FMN hydrolase activities than did the extracts of cells harboring the empty vector (Fig. 2). The riboflavin kinase activity of AtFMN/FHy increased roughly three times when the enzyme assays contained Tween 20; the FMN hydrolase activity did not change significantly (Fig. 2). The FMN hydrolase activity of the extracts containing the recombinant AtFMN/FHy greatly decreased when EDTA replaced MgCl 2 in the assays (Fig. 2B), as expected for a member of the Mg 2ϩ -dependent haloacid dehalogenase family (51)(52)(53)(54).
We hypothesized from the bioinformatic data that the riboflavin kinase and FMN hydrolase activities of AtFMN/FHy reside in separate domains. To test this hypothesis, the N-terminal (AtFHy, amino acids 1-234) and C-terminal (AtFMN, amino acids 227-379) domains of AtFMN/FHy were subcloned into the pET-30 Ek/LIC vector, functionally expressed in E. coli, and characterized as described in the next section. A small overlapping region (amino acids 227-234) was included in AtFMN and AtFHy because we could not accurately determine the boundary between the two domains from the protein sequence alignment (Fig. 1C).
Purification and Biochemical Characterization of AtFMN/FHy, AtFMN, and AtFHy-The recombinant enzymes carrying N-terminal S-protein and hexahistidine tags were purified by Ni 2ϩ chelate affinity chromatography, and digested with recombinant enterokinase to cleave the tags. Uncleaved AtFMN/FHy and the tag were removed by the second Ni 2ϩ chelate affinity chromatography (Fig. 3). Uncleaved AtFMN or AtFHy, and the tags, were removed by ion exchange chromatography

Homologs of riboflavin kinases and FAD synthetases from Arabidopsis
The putative riboflavin kinase and FAD synthetase genes from Arabidopsis were identified using the TBLASTN search program and the indicated query sequence.

Query sequence
Arabidopsis homolog E-value  (Fig. 3). These enzyme preparations with cleaved His tags were used in all subsequent work. Mobility of the purified AtFMN/FHy, AtFMN, and AtFHy on the SDS-PAGE gel agreed, respectively, with the theoretical molecular masses of 42.1, 17.5, and 25.7 kDa, which were calculated from the amino acid sequence. The molecular mass of native AtFMN/FHy was estimated by gel filtration chromatography. The enzyme activity migrated as a symmetrical peak with an apparent mass of 43.6 kDa, in close agreement with the theoretical molecular mass of 42.1 kDa. Thus, this measurement indicates that AtFMN/FHy is active as a monomer. Riboflavin kinases from rat and the plant Phaseolus aureus are also active as monomers (8,17).

Plant Riboflavin Kinase-FMN Hydrolase
To test how Tween 20 affects the riboflavin kinase and FMN hydrolase activities of the purified AtFMN/FHy, both enzyme activities were measured in the temperature range of 25-65°C, with or without detergent added to the assays (Fig. 4, A and B). Both enzyme activities increased (ϳ11 times for riboflavin kinase and ϳ4 times for FMN hydrolase at 30°C; even higher increases were observed at higher assay temperatures) and the associated temperature optima shifted to 5-10°C higher temperatures in the presence of Tween 20. The deter-gent apparently has a stabilizing effect on AtFMN/FHy; it was thus included in all the subsequent enzyme assays.
The activation energy of riboflavin kinase activity, with versus without detergent added to the assays, did not change significantly (Fig. 4C). However, the higher activation energy of the FMN hydrolase activity with Tween 20 suggested that the detergent affects the mechanism of the FMN hydrolase reaction (Fig. 4D).
The low molecular weight acid phosphatase from bovine kidney is the only FMN-hydrolyzing enzyme for which activation energy (28.6 kJ mol Ϫ1 ) is known (37). The activation energy of AtFMN/FHy for FMN hydrolysis is higher, with (72.8 kJ mol Ϫ1 ) or without (56.7 kJ mol Ϫ1 ) the   detergent, than that of the bovine enzyme. Activation energy has not been reported for any riboflavin kinase.
The riboflavin kinase activity of AtFMN/FHy exhibited alkaline pH optimum (Fig. 5A), as did the activities from rat and the plant Solanum nigrum (8,18). The FMN hydrolase activity of AtFMN/FHy exhibited acid pH optimum (Fig. 5B), as did the previously characterized acid phosphatases that hydrolyze FMN (21,35,37). The FMN hydrolase activity of AtFMN/FHy declined sharply when assayed above pH 8.0. By contrast, the activities of all other known FMN-hydrolyzing enzymes declined sharply when assayed at above pH 5.5-6.0.
The riboflavin kinase activity of AtFMN/FHy was slightly higher in Tris-HCl buffer than in phosphate (pH 7.5) or CHES (pH 9.0) buffer (Fig. 5A). To examine if the buffer effect at pH 7.5 is due to a change in one or both K m and k cat values, we determined the catalytic constants of AtFMN/FHy for riboflavin in Tris-HCl and phosphate buffers (TABLE  TWO). When the riboflavin kinase activity of AtFMN/FHy was assayed in Tris-HCl buffer, both K m and k cat values for riboflavin were slightly higher, and the catalytic efficiency (k cat /K m ) was virtually unchanged. Thus, the investigated buffer effect is small. Detailed biochemical characterization of AtFMN/FHy at various pH values was beyond the scope of this study. The buffer effects for the riboflavin kinase activity at pH 9.0 and for the FMN hydrolase activity at pH 6.0 (Fig. 5) were therefore not investigated.
To test the hypothesis that the riboflavin kinase and FMN hydrolase activities of AtFMN/FHy reside in separate domains, purified AtFMN and AtFHy were assayed for both enzyme activities (TABLE TWO). AtFMN has only riboflavin kinase activity; AtFHy has only FMN hydrolase activity. Hence, the two enzyme activities do reside in separate domains of AtFMN/FHy.
Catalytic constants of AtFMN/FHy, AtFMN, and AtFHy are given in TABLE TWO. These data establish that the riboflavin kinase and FMN hydrolase domains of AtFMN/FHy can be physically separated, with little change in their catalytic properties. No evidence is observed for substantial mutual modulation of the riboflavin kinase and FMN hydrolase activities in the fused enzyme. However, we cannot exclude that fusion of the riboflavin kinase and FMN hydrolase domains affects their temperature stability, pH response, or sensitivity to effectors. Investigation of these effects was beyond the scope of this study.
The K m value of AtFMN/FHy for riboflavin is ϳ10 times lower than the value of the riboflavin kinase that was partially purified from the plant S. nigrum (18), and is ϳ10 3 times lower than the value of the enzyme that was purified to homogeneity from rat (7,8). The enzyme assays were done at 37°C with the rat enzyme, so the published k cat value of the rat enzyme for riboflavin is not directly comparable to that of AtFMN/FHy. We estimated from the temperature curve (Fig. 4A) that AtFMN/FHy has ϳ15 times higher turnover number and ϳ10 4 times higher catalytic efficiency (k cat /K m ) compared with the rat enzyme at 37°C. The K m value of AtFMN/FHy for ATP is two times lower than that of the rat enzyme (7). Though Fmn1p from S. cerevisiae has been cloned, the kinetic properties of this enzyme have not been determined (12). Thus, our data are the first reported kinetic properties for a purified recombinant eukaryotic-type riboflavin kinase.
The kinetic constants for FMN hydrolysis have been reported for three acid phosphatases from spinach, and for the low molecular weight  acid phosphatases from bovine kidney (37) and human liver (31,34). The K m value of AtFMN/FHy for FMN is ϳ10 2 -10 3 times lower than those of the three enzymes from spinach, which were partially purified (21). The K m value of AtFMN/FHy for FMN is ϳ10 3 -10 4 times lower, and the catalytic efficiency is ϳ10 2 -10 3 times higher, compared with the values of the bovine and human enzymes, which were purified to apparent homogeneity (31,34). AtFMN/FHy is thus the most efficient catalyst of FMN hydrolysis reported to date.

DISCUSSION
Despite the vital roles of FMN and FAD in metabolism, much remains to be learned about the enzymes that catalyze conversion of riboflavin into these cofactors in plants. Searches of genome and EST databases suggested that plants have sequence homologs of both mono-functional and bifunctional enzymes with riboflavin kinase and FAD synthetase activities (TABLE ONE). Here we report on cloning and biochemical characterization of the Arabidopsis homolog of monofunctional enzymes with riboflavin kinase activity.
The enzyme AtFMN/FHy catalyzes hydrolysis of FMN to riboflavin, and phosphorylation of riboflavin to FMN. The FMN hydrolase activity of AtFMN/FHy resides in its N-terminal domain, apparently exclusive to plant riboflavin kinases. Sequence homology indicated that this domain is a member of the HAD superfamily of enzymes (Fig. 1A). The N-terminal domain of AtFMN/FHy is to our knowledge the first member of the HAD superfamily that catalyzes hydrolysis of FMN.
This novel FMN hydrolase activity raises the question whether other enzymes with similar biochemical properties exist in plants, other eukaryotes, and/or bacteria. The phylogenetic analysis (Fig. 1B) revealed that cellular organisms have many phylogenetically related enzymes of unknown biochemical functions. It is conceivable that some of these enzymes catalyze FMN hydrolysis. However, it is also conceivable that the FHy domains here described are the only members of the superfamily that catalyze FMN hydrolysis, as their protein sequences from algae to angiosperms group apart from all other HAD enzymes from representative cellular organisms (Fig. 1B).
The ability of AtFMN/FHy to hydrolyze FMN as well as the genomic evidence indicating that the bifunctional enzymes with riboflavin kinase and FMN hydrolase activities originated early in plant evolution are surprising findings. This is because FMN hydrolase activity was not detected in the previously purified riboflavin kinase from mung bean (17), and we found no evidence of a monofunctional riboflavin kinase without the extra N-terminal domain after an extensive search of plant EST and genome databases.
Three possibilities help reconcile our findings with earlier studies. First, bioinformatic evidence shows that plants contain sequence homologs of the bacterial bifunctional enzymes with riboflavin kinase and FAD synthetase activities (Arabidopsis has two). The plant riboflavin kinase described before (17) could be a sequence homolog of the bacterial bifunctional enzyme, but lacking FAD synthetase activity. Such an enzyme exists in S. agalactiae (6). Second, our results show that the FMN hydrolase activity of AtFMN/FHy requires Mg 2ϩ (Fig. 2B). This is also true for other phosphohydrolases of the HAD superfamily (56). Thus, the riboflavin kinase from mung bean could have FMN hydrolase activity that is undetectable when assayed without Mg 2ϩ . Third, plants could contain a monofunctional riboflavin kinase with no sequence homology to the riboflavin kinases investigated to date.
Plants use flavin nucleotides in mitochondria, plastids, and the cytosol. Three of the plant enzymes involved in the biosynthesis of the precursor riboflavin (lumazine synthase, bifunctional GTP cyclohydrolase II/3,4-dihydroxy-2-butanone 4-phosphate synthase, and 2,5-diamino-6-ribosylamino-4(3H)-pyrimidinone 5Ј-phosphate deaminase) have recently been cloned and characterized (57)(58)(59). Notably, all three contain N-terminal extensions with characteristics of chloroplast targeting peptides. Data base searches for plant homologs of riboflavin biosynthetic enzymes (not shown) indicated that all candidate sequences encode such extensions. In addition, pea chloroplasts import the lumazine synthase from spinach in vitro (59). A, assuming that plastids and mitochondria contain enzymes with riboflavin kinase and FAD synthetase activities; B, with only riboflavin kinase activity; and C, with only FAD synthetase activity. This model is based on biochemical and bioinformatic evidence suggesting that plastids alone contain the enzymes catalyzing synthesis of riboflavin (white arrows), that plastids and mitochondria contain enzymes with one or both riboflavin kinase and FAD synthetase activities (1), and that the cytosol contains a bifunctional enzyme with both activities (2) and an FAD synthetase (3). The predicted membrane transport is in black arrows.
If riboflavin is synthesized only in plastids, as the biochemical (57)(58)(59) and bioinformatic data suggest, then mitochondria and the cytosol must either import flavin nucleotides or import riboflavin to synthesize flavin nucleotides. Plants apparently contain homologs of monofunctional and bifunctional enzymes with riboflavin kinase and FAD synthetase activities (TABLE ONE). The sequence homologs of the monofunctional enzymes appear to be cytosolic; the homologs of the bifunctional enzymes have N-terminal extensions with characteristics of organellar targeting peptides. However, sequence data cannot be used to predict presence of one or both riboflavin kinase and FAD synthetase activities. An enzyme from S. agalactiae is a sequence homolog of the bifunctional enzyme with riboflavin kinase and FAD synthetase activities (RibC) from B. subtillis, but has only riboflavin kinase activity (6). We propose from these findings that the cytosol synthesizes both flavin nucleotides, and that organelles synthesize one or both FMN and FAD (Fig. 6).