Biosynthesis of Mannosylglycerate in the Thermophilic Bacterium Rhodothermus marinus

The biosynthetic reaction scheme for the compatible solute mannosylglycerate in Rhodothermus marinusis proposed based on measurements of the relevant enzymatic activities in cell-free extracts and in vivo 13C labeling experiments. The synthesis of mannosylglycerate proceeded via two alternative pathways; in one of them, GDP mannose was condensed withd-glycerate to produce mannosylglycerate in a single reaction catalyzed by mannosylglycerate synthase, in the other pathway, a mannosyl-3-phosphoglycerate synthase catalyzed the conversion of GDP mannose and d-3-phosphoglycerate into a phosphorylated intermediate, which was subsequently converted to mannosylglycerate by the action of a phosphatase. The enzyme activities committed to the synthesis of mannosylglycerate were not influenced by the NaCl concentration in the growth medium. However, the combined mannosyl-3-phosphoglycerate synthase/phosphatase system required the addition of NaCl or KCl to the assay mixture for optimal activity. The mannosylglycerate synthase enzyme was purified and characterized. Based on partial sequence information, the corresponding mgs gene was identified from a genomic library of R. marinus. In addition, the mgs gene was overexpressed inEscherichia coli with a high yield. The enzyme had a molecular mass of 46,125 Da, and was specific for GDP mannose andd-glycerate. This is the first report of the characterization of a mannosylglycerate synthase.

In Rhodothermus marinus, mannosylglycerate accumulates in response to growth at supraoptimal temperature and salinity while the amide form, mannosylglyceramide, accumulates exclusively in response to salt stress (5,6). Furthermore, mannosylglycerate was found to protect several enzymes against inactivation by temperature and freeze-drying (12). The elucidation of metabolic pathways for the synthesis of compatible solutes is essential to understand the regulatory mechanisms involved in the adaptation of many thermophilic and hyperthermophilic organisms to fluctuations in salt or temperature. Moreover, the biochemical and genetic characterization of key enzymes is also a prerequisite for achieving overproduction of these compounds in suitable hosts.
In this study we elucidated the biosynthetic pathways of mannosylglycerate in R. marinus, and characterized the enzyme mannosylglycerate synthase, catalyzing a final step in the synthesis of mannosylglycerate. In addition, we report the cloning and overexpression of the respective gene in Escherichia coli.
* This work was supported in part by the European Community Biotech Program Extremophiles as Cell Factories, BIO4-CT96-0488, and by PRAXIS XXI and FEDER, Portugal PRAXIS/2/2.1/BIO/1109/95. 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 /EBI Data Bank with accession number(s) AF173987.
ʈ Supported by invited Scientist Fellowship BCC/11978 from PRAXIS XXI.

Strains, Plasmids, and Culture Conditions
The type strain of R. marinus DSM 4252 (Deutsche Sammlung von Mikroorganismen und Zellkulturen, Braunschweig, Germany) was grown on Degryse medium containing 2.5 g of tryptone and 2.5 g of yeast extract/liter (24), and supplemented with 1, 4, and 6% NaCl (w/v). Cultures were grown in a 5-liter fermentor at 65°C, with continuous gassing with air and stirred at 150 rpm. Cell growth was monitored by measuring the turbidity at 600 nm.

Preparation of Cell-free Extracts
Cells were harvested by centrifugation (8,000 ϫ g, 15 min, 20°C) during the late exponential phase of growth and frozen at Ϫ80°C. The cell sediment was suspended in Tris-HCl (20 mM, pH 7.6) containing DNase I (10 g/ml extract) and MgCl 2 (5 mM), and a mixture of protease inhibitors: phenylmethylsulfonyl fluoride (0.08 mg/ml), leupeptine (0.02 mg/ml), and antipain (0.02 mg/ml). Cells were disrupted in a French pressure cell, followed by centrifugation (20,000 ϫ g, 1 h, 10°C) to remove cell debris. Cell extracts were applied to a Sephadex G-25 column (Amersham Pharmacia Biotech), equilibrated with 20 mM Tris-HCl (pH 7.6) to remove mannosylglycerate (MG) and other low molecular weight compounds prior to measuring enzyme activities. The protein content was determined by the Bradford assay (25) using bovine serum albumin as standard.
Discontinuous or continuous methods were used to measure the activities of these enzymes at 65°C. In the discontinuous method, cell extracts (generally, 100 l) were incubated with the substrates, in a total reaction volume of 0.6 ml, for various periods of time, and the reactions were stopped by cooling and acidification with 25 mM HCl. After centrifugation, the pH of the supernatants was adjusted to 7.0 before freeze-drying. The residues were dissolved in 2 H 2 O and analyzed by NMR. In the continuous method, the substrates were added to a 5-mm NMR tube and the reactions started by addition of aliquots of the enzyme preparation. The time course of product formation was followed by sequential acquisition of spectra. The reactions were stopped by rapid cooling followed by acidification. After centrifugation, the supernatant was neutralized and MG quantified by 1 H NMR. Control assays lacking the cell extracts or the substrates were also carried out. All specific activities are the mean values of six assays in crude extracts prepared from cells derived from three independent growths.

NMR Spectroscopy
NMR spectra were acquired on Bruker DRX500 or AMX300 spectrometers. 1 H NMR spectra were acquired with a 5-mm broad band inverse probe head with presaturation of the water signal. For quantification spectra were acquired with a repetition delay of 35 s and acetate was used as an internal concentration standard. Proton chemical shifts are relative to 3-(trimethylsilyl)propanesulfonic acid (sodium salt).

Purification of the Native Mannosylglycerate Synthase
The enzyme, catalyzing the synthesis of MG from GDP mannose and D-glycerate, was purified by fast protein liquid chromatography (Amersham Pharmacia Biotech) at room temperature from R. marinus cells grown in medium containing 4% NaCl. Fractions were examined for the presence of the enzyme by the mannosylglycerate synthase assay method described above and visualization of MG formation by TLC.
Ion-exchange Chromatography-Cell-free extract was applied to a column (XK50/30; bed volume, 250 ml) packed with DEAE-Sepharose fast flow (Amersham Pharmacia Biotech) equilibrated with Tris-HCl (20 mM, pH 7.6). Elution was carried out with a two-step linear NaCl gradient (0 -0.2 M and 0.2-0.4 M) in the same buffer. Mannosylglycerate synthase activity was found in the fractions eluting at approximately 0.3 M NaCl.
Hydrophobic Interaction Chromatography-Ammonium sulfate was added to the pooled fractions from the previous step to a final concentration of 0.5 M. This sample was applied to a column (XK16/20; bed volume, 90 ml) packed with phenyl-Sepharose (Amersham Pharmacia Biotech) equilibrated with 50 mM potassium phosphate (pH 7.0), containing 0.5 M ammonium sulfate. Elution was carried out with a decreasing linear gradient of ammonium sulfate (0.5-0.0 M) in potassium phosphate buffer. Mannosylglycerate synthase eluted toward the end of the gradient.
Anion-exchange Chromatography (Resource Q)-Active fractions were pooled, concentrated, and equilibrated to 20 mM Tris-HCl (pH 6.7). The resulting sample was applied to a 6-ml Resource Q column. Elution was carried out with a two-step linear NaCl gradient (0 -0.2 M and 0.2-1 M). The fractions eluting between 0.22 and 0.25 M NaCl contained mannosylglycerate synthase.
Electroelution of Mannosylglycerate Synthase from a Native PAGE-Electrophoresis of native proteins was performed on 7.5% polyacrylamide (30). Four protein bands were located by cutting two thin longitudinal strips from the sides of the gel slab and staining with Coomassie Blue. The four bands were recovered by electrophoretic elution from the gel using a BIOTRAP BT 1000 (Schleider & Schuell) at 200 V for 4 h, and each protein was assayed for mannosylglycerate synthase activity. Of these bands, only one had this activity, originating a single band by SDS-PAGE.
The N-terminal amino acid sequence of mannosylglycerate synthase was determined by the method of Edman and Begg (31) using an Applied Biosystem Model 477A protein sequencer. The internal sequences were determined by Edman degradation after digestion with trypsin and separation of the peptides by micro-HPLC at the Microchemical Facility, Emory University School of Medicine, GA.

DNA Methodology, Cloning, and Analysis
Most DNA manipulations followed standard molecular techniques and procedures (32). Total DNA from R. marinus was purified by the method of Marmur (33). For sequencing purposes, recombinant plasmid DNAs were prepared using plasmid kits (QIAGEN). Southern blots of restriction endonuclease-cleaved genomic or plasmid DNAs, and colony blots were hybridized with DNA probes labeled with a digoxigenin DNA labeling and detection kit (Roche Molecular Biochemicals).
To obtain a genomic library from R. marinus, total DNA was partially digested with restriction enzyme Sau3A. Fragments ranging from 1 to 5 kbp were purified from agarose gels, and ligated into the BamHI site of dephosphorylated pUC18. The ligation mixture was used to transform E. coli TG1 cells. Transformed cells were plated on YT-broth supplemented with ampicillin, 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside, and IPTG. Approximately, 2 ϫ 10 4 transformants were obtained after 18 h of growth at 37°C. Similarly, a partial genomic library from R. marinus, containing DNA fragments selected by Southern analysis of genomic DNA preparations with the mgs probe, was obtained by complete digestion of total DNA with restriction enzyme PstI, followed by ligation of purified fragments of about 2.1 kbp in size into the PstI site of pUC18, and subsequent transformation of E. coli XL1-Blue cells with the ligation mixture. Positive clones were detected by colony hybridization with digoxigenin-labeled probes as described above.
Nucleotide sequences were determined by MWG-BIOTECH (Ebersberg, Germany) using the LI-COR 4200 automated sequencing system. Inserts of positive pUC18 clones (pMG721, pMG7, and pMG161), pGEM-T Easy-clones, and plasmid clone pMG37 were sequenced in both orientations using vector-and insert-specific oligodeoxynucleotide primers. Nucleic acid and protein sequence analyses were conducted with programs in the Wisconsin Genetics Computer Group (GCG) software package (34). The European Bioinformatics Institute data bases, and functionally annotated genomes in the Kyoto Encyclopedia of Genes and Genomes, as well as the archaeal genome sequence data base were screened for homologies using the (T)FASTA (35,36) and (T)BLAST (37) algorithms.
Overexpression of the mgs Gene and Purification of the Enzyme-The mgs gene was amplified by PCR using the forward primer 5Ј-GCG-GAATTCATGAGCCTGGTCGTTTTTCCC-3Ј with an additional EcoRI recognition sequence (underlined) immediately upstream of the ATG start codon, and the reversed primer 5Ј-GCGCTGCAGGGATCCTCAG-GCGGTCGACAGTGCC-3Ј with additional PstI and BamHI recognition sequences (underlined) directly behind the TGA stop codon. The PCR product was purified after digestion with EcoRI and PstI, and ligated into the corresponding sites in the pKK223-3 expression vector to obtain plasmid pMG37. E. coli XL1-Blue cells containing pMG37 were grown to the midexponential growth phase (OD 600 ϭ 0.4), induced with IPTG, and growth continued for a further 3 to 4 h. Cells were harvested and treated as described above for the preparation of cell-free extracts. The resulting extract was incubated for 30 min at 75°C and centrifuged. After dialysis against 20 mM Tris-HCl (pH 7.6), the sample was loaded onto a Mono Q (Amersham Pharmacia Biotech) column that was eluted by a linear gradient of NaCl (0 -500 mM). Fractions with activity were pooled, concentrated, and applied to a Superdex-200 (Amersham Pharmacia Biotech) column.

Characterization of the Native and the Recombinant Mannosylglycerate Synthase
The temperature profile for the activity of mannosylglycerate synthase was determined between 35 and 95°C, by using the discontinuous method described above. The effect of pH on mannosylglycerate synthase activity was determined at 90°C in 50 mM MES (pH 5.5-6.5) and 50 mM BisTris-propane buffer (pH 6.5-9.5). All pH values were measured at room temperature; pH values at 90°C were calculated by using ⌬pK a /⌬T°C ϭ Ϫ0.011 and Ϫ0.015 for MES and BisTris-propane, respectively. Enzyme thermostability was determined at 65 and 90°C by incubating an enzyme solution in 20 mM Tris-HCl (pH 7.6). At appropriate times, samples were withdrawn and immediately examined for residual mannosylglycerate synthase activity using the discontinuous assay at 90°C. Kinetic parameters were determined at 90°C. Reaction mixtures contained GDP mannose (0.165-10.0 mM) plus 10 mM D-glycerate, or D-glycerate (0.188 -10.0 mM) plus 10 mM GDP mannose. Samples were preheated for 3 min, the reaction was initiated by the addition of the enzyme preparation, and MG was quantified by 1 H NMR. Experiments were performed in duplicate. Values for V max and K m were determined from Hanes plots.
The molecular mass of native mannosylglycerate synthase was determined on a Superose 12 column (Amersham Pharmacia Biotech) equilibrated with 50 mM sodium phosphate buffer (pH 7.6) containing 0.15 M NaCl. Cytochrome c (12.4 kDa), carbonic anhydrase (29 kDa), albumin (66 kDa), aldolase (158 kDa), and ferritin (440 kDa) were used as standards. The isoelectric point of native mannosylglycerate synthase was determined by isoelectric focusing (model 111 mini-IEF cell, Bio-Rad), according to the manufacturer. A pH 3-10 isoelectric focusing gel and standards in a range of pI 4.5-9.6 were used.

Nucleotide Sequence Accession Number
The nucleotide sequence of mannosylglycerate synthase has been deposited in GenBank under accession number AF173987.

RESULTS
Enzymatic Activities Involved in the Synthesis of Mannosylglycerate-By analogy with the known biosynthetic pathways of other sugar derivatives, such as glucosylglycerol (16) and trehalose (38), the terminal reaction in the synthesis of MG should involve a sugar nucleotide as donor of the glycosyl moiety, and a phosphorylated acceptor (3-P-glycerate). From an array of experiments with GDP mannose, UDP mannose, and ADP mannose as donors, and D-3-P-glycerate as acceptor, we were unable to detect the formation of MG in cell extracts prepared from R. marinus grown in medium containing 4% NaCl. Activity for the synthesis of MG, however low, was detected upon the addition of NaCl or KCl (150, 300, and 450 mM), to the enzyme assay, when GDP mannose was used as a glycosyl donor. Optimal activity was observed at 300 mM NaCl or KCl. Unexpectedly, the activity for the formation of MG in cell extracts was significantly higher when D-glycerate, instead of D-3-P-glycerate, was used in addition to GDP mannose (Table I, compare the values for mannosylglycerate synthase and mannosyl 3-P-glycerate synthase/mannosyl 3-P-glycerate phosphatase). Furthermore, this activity did not depend on the salt concentration in the assay mixture (0 -450 mM NaCl or KCl). 1 H NMR was used to detect and monitor the time course for the formation of MG in crude cell extracts at 65°C. In the anomeric region of the spectra clear resonances due to the anomeric protons of MG and the ribose and mannose moieties in GDP mannose were detected, allowing to monitor on line the consumption of the substrate and build up of the product (Fig. 1).
The use of two different substrates, D-glycerate and D-3-Pglycerate, combined with the differences in the salt dependence, led us to suspect that R. marinus contained two enzymatic systems for the synthesis of MG. This hypothesis was further supported by the observation that D-3-P-glycerate was not a substrate for purified mannosylglycerate synthase, the enzyme catalyzing the synthesis of mannosylglycerate from GDP mannose and D-glycerate, and the activity of the enzyme was independent of the presence of NaCl or KCl. Firm evidence for the salt dependent activity, which used GDP mannose and D-3-P-glycerate as substrates, was obtained from experiments performed with the flow-through of the DEAE-Sepharose column in the chromatographic procedure for purification of mannosylglycerate synthase. This fraction did not contain mannosylglycerate synthase activity, as assayed with GDP mannose and D-glycerate. On the other hand, incubation of an aliquot with GDP mannose and D-3-P-glycerate, at 65°C, did not lead to the immediate formation of MG, but to a compound that remained at the origin of TLC plates, suggesting that a phosphorylated product had been formed. To confirm this hypothesis, alkaline phosphatase was added to the reaction mixture after the initial incubation period with GDP mannose and D-3-P-glycerate. The analysis of the reaction mixture by TLC revealed the presence of MG (Fig. 2). The identity of the compound was confirmed by 1 H NMR. These results showed that mannosyl-3-phosphoglycerate synthase activity was present in the fractions that did not adsorb onto the DEAE-Sepharose column, and demonstrated the presence of an enzymatic system for MG synthesis, that did not involve mannosylglycerate synthase. Evidence for the presence of a phosphatase, which acted upon mannosyl-3-phosphoglycerate in cell extracts, was further obtained from the decrease in the formation of MG when cell extracts were treated with the phosphatase inhibitor, sodium fluoride (results not shown).
To study the effect of salinity of the growth medium on the expression of the enzymes involved in the biosynthetic pathway of MG, cell extracts from R. marinus grown in medium containing 1, 4, and 6% NaCl were examined, and the enzymatic activities responsible for catalysis of the sequence of reactions from glucose 6-phosphate to the formation of GDP mannose, as well as those catalyzing the final step in the biosynthesis of MG, were determined. The enzymes committed to the synthesis of MG were not affected by the salinity of the growth medium, but the three activities involved in the formation of mannose-1-P from glucose-6-P decreased significantly. However, all the activities were constitutive under the conditions tested ( Table I).
Synthesis of Mannosylglycerate in Whole Cells from 13 C-Labeled Glucose-The constitutive character of the enzymes involved in the synthesis of MG was also apparent from the immediate synthesis of this solute observed in non-growing cell suspensions of R. marinus in response to a salt upshock. The production of MG by cells of R. marinus was investigated by in vivo 13 C NMR, in experiments where [1-13 C]glucose was supplied to cells grown at 65°C in a defined medium containing 1% NaCl and glucose as the only carbon source. Cells were suspended in phosphate buffer (20 mM, pH 7.5) containing 1% NaCl. Under these conditions labeling of MG could not be detected; however, increasing the NaCl concentration to 3% resulted in the appearance of resonances due to C 1 (at 98.7 ppm) and C 3 (at 63.3 ppm) of the mannose and glycerate moieties of MG, respectively (spectra not shown). The MG content increased in the first 50 min after the salt upshock and remained constant thereafter. These data also show that the mannose moiety is derived directly from glucose, and glycolysis proceeds via the Embden-Meyerhoff pathway in R. marinus since label from [1-13 C]glucose ends on carbon 3 of glycerate (39).
Purification and Sequence Analysis of Mannosylglycerate Synthase-The enzyme, catalyzing the synthesis of mannosylglycerate from GDP mannose and D-glycerate was purified using four chromatographic steps described under "Materials and Methods." At this stage, approximately 30% of the mannosylglycerate synthase activity present in the initial crude extract was recovered. This preparation contained four bands in native PAGE (Fig. 3). The enzyme was obtained by electroelution and migrated as a single band in SDS-PAGE with an apparent molecular mass of 45 kDa. A value of 122 Ϯ 9 kDa was found by gel filtration at room temperature. The isoelectric point was 4.8. The N-terminal amino acid sequence of the   purified enzyme and those of five peptides generated by a tryptic digestion were determined to be SLVVFPFKHEHPEV-LLHNVRVAAXHPRXH, IHFYDADITSFGPD, ASTDAMITW-MITR, LYGGLDD, QVELLELFTTPVR, and GYDYAQQYLYR, respectively.
Cloning, Sequencing and Overexpression of mgs Gene in E. coli-PCR amplification from the genomic DNA of R. marinus with primers based on the N-terminal and three internal amino acid sequences yielded unique products with lengths of approximately 300, 900, and 1100 nucleotides. The nucleotide sequence of the 1100-bp fragment predicted an open reading frame encoding a peptide of 360 amino acids, which contained the N-terminal sequence as well as all internal, partial sequences of mannosylglycerate synthase. This fragment was used as a probe to isolate clones from the constructed genomic libraries of R. marinus. In total, three positive clones were isolated. One plasmid derived from the genomic library, designated pMG721, contained a 3.4-kbp partial Sau3A fragment, whereas the two plasmids of the size selected library, designated pMG7 and pMG161, contained each orientation of the expected 2.1-kbp PstI fragment. Nucleotide sequence analysis showed, that the 3.4-kbp fragment overlapped the 2.1-kbp fragment almost completely (Fig. 4).
Several open reading frames were found on both strands (Fig. 4). An open reading frame, designated mgs, encoding a protein of 397 amino acids, was found showing a perfect correlation with the sequence of the 1100-bp PCR fragment used as probe. The ATG codon (at position 689) functions as initiation codon for the mgs gene (Fig. 5), since it was directly followed by a stretch of 23 amino acids identical to the N-terminal sequence of the purified enzyme. A putative ribosome binding sequence is positioned 7-12 bp upstream of the ATG codon. The molecular mass based on the deduced amino sequence was calculated to be 46,125 Da, which is in excellent agreement with the value of 45 kDa obtained by SDS-PAGE with the native enzyme. Comparison of the amino acid sequence of mannosylglycerate synthase with sequences in several data bases showed only low similarities. The highest similarity (29% identity) was found with a 382-amino acid long hypothetical protein of Pyrococcus horikoshii.
Two open reading frames located directly upstream and downstream of the mgs gene with lengths of 219 and 308 amino acids, respectively, did not show any significant similarities with other known sequences in the data bases. However, in the opposite orientation upstream of the mgs gene, an open reading frame (with an ATG start codon at position 382) encodes a 120-amino acid partial sequence with high similarity (47% identity) to a glucosamine-6-phosphate deaminase of E. coli (40). The 3Ј end of the cloned region (position 2687 to end) was identical to a previously cloned sequence of the lig gene coding for the DNA ligase of R. marinus (41).
For overexpression in E. coli, the PCR-amplified mgs coding sequence was cloned under the control of the strong inducible FIG. 3. Native PAGE and SDS-PAGE of mannosylglycerate synthase. A, four protein bands were found by native PAGE (7.5%, w/v, acrylamide) after the chromatographic purification steps (total protein applied was 20 g). The band indicated by the arrow had mannosylglycerate synthase activity and was recovered by electroelution. B, SDS-PAGE of the electroeluted band (protein applied was 10 g) gave rise to a single band (lane 2). Lane 1, molecular mass markers.
FIG. 4. Physical and genetic map of the mannosylglycerate synthase gene (mgs) of R. marinus. Open reading frames (orf) are indicated by arrows. The horizontal lines represent the inserts of the isolated plasmids. NagB' indicates a partial open reading frame with homology to the glucosamine-6-phosphate deaminase (NagB) from E. coli (40). Lig' represents the (partial) lig gene from R. marinus (41).
FIG. 5. Nucleotide sequence and deduced amino acid sequence of the mannosylglycerate synthase gene. The underlined amino acid sequences indicate the N terminus of the purified mannosylglycerate synthase and the peptides whose sequences were determined after tryptic digestion of the purified mannosylglycerate synthase. The putative ribosome binding sites are highlighted with a black background. tac promoter in pKK223-3. The sequence of the plasmid insert was identical to that of the native gene. Moreover, the Nterminal sequence of the recombinant protein was also identical to the native N-terminal sequence (data not shown). SDS-PAGE analysis of crude extracts from E. coli XL1-Blue (pMG37) grown in the presence of IPTG revealed the presence of an extra band of 45 kDa, that was not observed in crude extracts from E. coli XL1-Blue (pKK223-3) (Fig. 6). A heat treatment at 75°C for 30 min resulted in an extensive purification of the crude extract with the 45-kDa band being almost exclusively present (Fig. 6). Incubation of the crude extract from E. coli XL1-Blue (pMG37) at 80°C with GDP mannose and D-glycerate resulted in the formation of MG, but this compound was not detected after incubation of those substrates with a crude extract from E. coli XL1-Blue (pKK223-3). The specific activity of mannosylglycerate synthase in crude extracts of E. coli XL1-Blue (pMG37) was 10 mol/min/mg of protein at 90°C.
Catalytic Properties of Mannosylglycerate Synthase-The native enzyme preparation obtained by electroelution had a specific activity of 70 mol/min/mg of protein. The enzyme did not have detectable activity for the backward reaction. The substrate specificity of mannosylglycerate synthase was investigated using GDP mannose, ADP mannose, UDP mannose, mannose 1-phosphate, and mannose 6-phosphate as sugar donors, and D-glycerate, L-glycerate, and D-3-P-glycerate as sugar acceptors. To investigate the ability of mannosylglycerate synthase to catalyze the synthesis of other glycosylconjugates, GDP glucose, UDP glucose, and ADP glucose were tested as sugar donors, and glycerol 3-phosphate, glucose 6-phosphate, glucose 1-phosphate, and glucose as sugar acceptors. The enzyme was specific for GDP mannose and D-glycerate. An absolute stereospecificity for D-glycerate was found. UDP mannose was the only sugar nucleotide, other than GDP mannose, that was used by the enzyme, but to a very low extent (less than 3% of that of GDP mannose). K m values were determined for GDP mannose and D-glycerate (Table II). NaCl and KCl, in the concentration range of 50 -500 mM, had no effect on the enzyme activity. On the other hand, addition of Mg 2ϩ was required for maximal activity; the activity of mannosylglycerate synthase was 2.5-fold lower in the absence of this divalent cation. The activity was undetectable at room temperature, and maximal activity was reached between 85 and 90°C (Fig. 7), which is well above the temperature range for growth of R. marinus (6). From the linear part of an Arrhenius plot, the activation energy of 63 kJ/mol was calculated. At the temperature required for maximal activity, 90°C, the stability of the enzyme was only moderate, 50% of the activity being lost after 30 min incubation (Fig. 8). At 65°C the specific enzymatic activity was 5-fold lower than the maximum value, and the half-life for inactivation increased to 170 min. The enzyme thermostability was not improved by Ca 2ϩ (5 mM), nor was it enhanced under anaerobic conditions. Within the pH range examined (5.5-8.5), the activity of the enzyme at 90°C remained nearly constant between pH 5.5 and 7.0, but decreased as the pH was increased above 7.0 (Fig. 9). These properties were determined using a partially purified enzyme preparation with a specific activity of 13 mol/ min/mg of protein.
The temperature profile for activity (Fig. 7), the thermostability at 65 and 90°C (Fig. 8) as well as the kinetic parameters K m and V max were also determined for the recombinant enzyme (Table II). Overall, the properties of the native and recombinant enzyme were similar, but the recombinant form was less thermostable.

DISCUSSION
In this study, two alternative routes for the synthesis of the compatible solute mannosylglycerate in R. marinus are proposed based on measurements of the relevant enzymatic activities in cell-free extracts and in vivo 13 C labeling experiments (Fig. 10). In R. marinus GDP mannose is used as the glycosyl donor for the synthesis of MG, in accordance with the general observation that GTP is the preferred substrate to carry mannose in an activated form in many reactions (42). From GDP mannose, the synthesis proceeds via a branched pathway: a single-step reaction converts GDP mannose and D-glycerate to MG by the action of mannosylglycerate synthase; in the other pathway, GDP mannose and D-3-P-glycerate react to produce the phosphorylated intermediate, mannosyl 3-phosphoglycerate, which is subsequently dephosphorylated by a phosphatase yielding MG.
A noticeable feature distinguishing the two enzymatic systems involved in the synthesis of MG is their dependence on salt; while the mannosylglycerate synthase reaction is saltindependent, addition of NaCl or KCl is required to achieve full activity of the mannosyl-3-phosphoglycerate synthase/phosphatase system. This salt dependence is actually a feature common to the biosynthetic pathways of other osmolytes such as, trehalose, glucosylglycerol, and galactosylglycerol, whose synthesis also occurs by two-step reactions, involving the concerted action of synthases and phosphatases (16,18,38). For instance, the synthesis of glucosylglycerol involves the condensation of ADP glucose and glycerol 3-phosphate to produce glucosylglycerol phosphate, which is subsequently dephosphorylated to yield glucosylglycerol (16). The enzymes involved in these biosynthetic pathways have not been purified or thoroughly characterized, but it has been shown that enzyme activities in crude extracts are strongly activated by potassium and sodium salts (16 -18, 38).  It is, therefore, intriguing that the salt-dependent enzymatic system for the synthesis of MG has an activity considerably lower (6 -14-fold) than that of the salt-independent system, especially if we take into consideration that the concentration of mannosylglycerate in R. marinus cells is markedly dependent on the salinity of the growth medium (5, 6). The present study does not allow an evaluation of the relative contributions of the two pathways in vivo; however, the in vivo 13 C NMR experiments showed synthesis of MG in R. marinus in response to salt upshock, an observation that seems to support the relevant contribution of the salt-dependent system, at least under the conditions examined. Also curious is the observation that the enzymatic activities involved in the synthesis of MG, as determined in cell extracts, were independent of the salt concentration of the growth medium, whereas salt-induced de novo synthesis of enzymes involved in the formation of several osmolytes has been reported (13,15,17,38). However, our results do not rule out the existence of a salt-dependent control mechanism at the level of gene expression (transcription or translation).
The reasons underlying the existence of two distinct pathways for the synthesis of MG in R. marinus remain unclear, but it is conceivable that these systems are differentially regulated by salt and temperature, the two physicochemical parameters that are known to influence the intracellular concentrations of MG in R. marinus (6). The two pathways could also have different metabolic functions, one being implicated in the synthesis of MG and the other in its catabolism, as observed in the synthesis of sucrose in plants, which employs two different pathways, both involving UDP glucose as glucosyl donor, but differing in the nature of the glycosyl acceptor: fructose-6-P or D-fructose. In this case it has been suggested that sucrose phosphate synthase is primarily involved in the synthesis of sucrose, whereas sucrose synthase is implicated in sucrose catabolism (43,44). However, mannosylglycerate synthase does not appear to be involved in catabolism of MG since the activity of the enzyme in the catalysis of the cleavage of MG was not detectable.
In R. marinus GDP mannose is formed via a conventional pathway involved in the synthesis of polymers containing mannose or mannose derivatives, as observed in the biosynthetic pathway of exopolysaccharides (45,46) and in colanic acid synthesis (47). The key glycolytic metabolite, glucose 6-phosphate is converted to mannose 6-phosphate by the action of phosphohexose isomerases, and mannose 6-phosphate is subsequently converted to mannose 1-phosphate by phosphomannomutase. The activated sugar, mannose 1-phosphate, and GTP are the substrates for mannose-1-phosphate guanylyltransferase catalyzing the formation of GDP mannose and pyrophosphate (Fig. 10). A marked difference exists in the magnitude of the specific activities of the enzymes involved in the formation of GDP mannose in R. marinus and those committed to the synthesis of MG, as generally observed when comparing primary metabolic trunks with specific branches.
In this study we choose to characterize biochemically and genetically the enzyme system that showed highest in vitro activity for the synthesis of mannosylglycerate, mannosylglycerate synthase. The enzyme has a high degree of specificity with regard to the substrates GDP mannose and D-glycerate. It is interesting to note that lactose synthase, one of the most extensively characterized glycosyltransferases, also uses a nonphosphorylated glycosyl acceptor. The enzyme catalyzes the transfer of galactose from UDP galactose to D-glucose, in the presence of ␣-lactalbumin, with a high degree of substrate specificity (48,49). However, this is not a general feature among glycosyltransferases.
The properties of the native and recombinant mannosylglycerate synthase are very similar, except for a lower thermal stability of the recombinant enzyme at 90°C, which may be accounted by slight differences in the folding of the protein produced in the heterologous host. Mannosylglycerate synthase shares a low degree of amino acid sequence homology with other sugar transferases and functionally related enzymes. Moreover, comparison of the amino acid sequence of mannosylglycerate with those in several data bases showed only very low similarities, although mannosylglycerate is synthesized by several hyperthermophilic archaea, such as Pyrococcus furiosus and P. horikoshii, whose complete genomic sequences are known.
In conclusion, we believe that the results reported in this study on the characterization of the biosynthesis of MG in the thermophilic bacterium R. marinus, as well as of the key enzyme, mannosylglycerate synthase, represent an important step toward the elucidation of MG synthesis and regulation in hyperthermophilic archaea sharing with R. marinus the ability to accumulate this osmolyte.