Biosynthesis of dTDP-3-acetamido-3,6-dideoxy-alpha-D-galactose in Aneurinibacillus thermoaerophilus L420-91T.

The glycan chain of the S-layer protein of Aneurinibacillus thermoaerophilus L420-91T (DSM 10154) consists of d-rhamnose and 3-acetamido-3,6-dideoxy-d-galactose (d-Fucp3NAc). Thymidine diphosphate-activated d-Fucp3NAc serves as precursor for the assembly of structural polysaccharides in Gram-positive and Gram-negative organisms. The biosynthesis of dTDP-3-acetamido-3,6-dideoxy-alpha-d-galactose (dTDP-d-Fucp3NAc) involves five enzymes. The first two steps of the reaction are catalyzed by enzymes that are part of the well studied dTDP-l-rhamnose biosynthetic pathway, namely d-glucose-1-phosphate thymidyltransferase (RmlA) and dTDP-d-glucose-4,6-dehydratase (RmlB). The enzymes catalyzing the last three synthesis reactions have not been characterized biochemically so far. These steps include an isomerase, a transaminase, and a transacetylase. We identified all five genes involved by chromosome walking in the Gram-positive organism A. thermoaerophilus L420-91T and overexpressed the three new enzymes heterologously in Escherichia coli. The activities of these enzymes were monitored by reverse phase high performance liquid chromatography, and the intermediate products formed were characterized by 1H and 13C nuclear magnetic resonance spectroscopy analysis. Alignment of the newly identified proteins with known sequences revealed that the elucidated pathway in this Gram-positive organism may also be valid in the biosynthesis of the O-antigen of lipopolysaccharides of Gram-negative organisms. The key enzyme in the biosynthesis of dTDP-d-Fucp3NAc has been identified as an isomerase, which converts the 4-keto educt into the 3-keto product, with concomitant epimerization at C-4 to produce a 6-deoxy-d-xylo configuration. This is the first report of the functional characterization of the biosynthesis of dTDP-d-Fucp3NAc and description of a novel type of isomerase capable of synthesizing dTDP-6-deoxy-d-xylohex-3-ulose from dTDP-6-deoxy-d-xylohex-4-ulose.

Glycosylated and non-glycosylated S-layer 1 proteins represent the outermost cell envelope components of organisms of the domains Bacteria and Archaea (1,2). Structurally, S-layer glycoproteins from Gram-positive bacteria resemble O-antigens of LPS of Gram-negative organisms (3). Usually both macromolecules consist of a glycan chain containing identical repeating units, a core saccharide, and a linkage region, which is bound either to the S-layer protein or, as in LPS, to lipid A. Whereas the biosynthesis of LPS is well characterized (4), the exact mechanisms involved in biosynthesis of glycans of bacterial S-layer glycoproteins are still unknown. There is, however, preliminary evidence that both systems follow comparable pathways. For formation of the glycan chains in the cytoplasm in both systems nucleotide-activated sugars serve as precursors in the glycosylation reaction (5). It is proposed that the pre-assembled glycan chains are transported across the cytoplasmic membrane in a lipid-bound state, either linked to undecaprenol (LPS) (4) or dolichol (glycoproteins) (6). Although experimentally not yet determined, final transfer of the glycan chains to the nascent S-layer polypeptide is assumed to take place on the outer side of the cytoplasmic membrane, comparable with archaeal S-layers (7).
The repeating unit of the glycan chain of Aneurinibacillus thermoaerophilus L420-91 T (8,9), a member of the Bacillus/ Clostridium group, consists of the sugars D-rhamnose and 3-acetamido-3,6-dideoxy-D-galactose (3-N-acetylfucosamine (D-Fucp3NAc)). The backbone of each repeating unit is built by four D-rhamnose units, and two of them are substituted by ␣-1,2-linked D-Fucp3NAc units (10,11). Both constituting sugars are well known components of the LPS of Gram-negative bacteria (12)(13)(14). Derivatives of D-Fucp3N have also been found in the LPS of clinical strains of Vibrio cholerae (15) and Acinetobacter baumannii (16), where the acetyl group is replaced by modified amino acids and acyl residues, respectively.
Recently, we were able to characterize the biosynthesis of the GDP-D-rhamnose, one of the sugar components of the glycan chain of A. thermoaerophilus L420-91 T (17). The other constituent, D-Fucp3NAc, was first described in cell-free extracts of Xanthomonas campestris by Volk and Ashwell (18). They proposed that D-glucose will eventually be converted into dTDP-D-Fucp3NAc by amidation with L-glutamate and pyridoxal 5-phosphate (PLP) as co-factor followed by acetylation with acetyl-CoA (18). This proposal was further substantiated by Shibaev (19), who proposed a model for the biosynthesis of nucleotide-activated D-Fucp3NAc. Two of the enzymes presum-ably involved in dTDP-D-Fucp3NAc biosynthesis were first identified in the dTDP-L-rhamnose pathway (for review, see Ref. 20). These are glucose-1-phosphate thymidyltransferase (RmlA) and dTDP-D-glucose-4,6-dehydratase (RmlB), which catalyze the reaction of glucose 1-phosphate and thymidine triphosphate to dTDP-D-glucose and in the second step to dTDP-6-deoxy-D-xylohex-4-ulose. This latter compound serves as substrate for a series of reactions, including the formation of dTDP-D-Fucp3NAc (21). Activities of RmlA and RmlB in crude extracts of A. thermoaerophilus L420-91 T indicated a possible involvement of these two enzymes in the biosynthesis of dTDP-D-Fucp3NAc (22). Recently, the gene cluster responsible for LPS biosynthesis of X. campestris has been identified (14). Downstream of the dTDP-L-rhamnose genes the authors described a cluster of 15 genes to be involved in the biosynthesis of the LPS glycan chain of X. campestris. Two genes of the cluster, namely wxcM and wxcK, were assumed to code for enzymes that synthesize dTDP-D-Fucp3NAc. WxcM was proposed to be bifunctional and to catalyze both the isomerization and the transacetylation reaction. The second ORF, WxcK, was assumed to be responsible for the transamination reaction. So far none of the described genes has been functionally expressed nor have the gene products been biochemically characterized.
This report is the first functional characterization of the biosynthesis pathway of dTDP-D-Fucp3NAc in the S-layer glycoprotein glycan of the Gram-positive thermophile A. thermoaerophilus L420-91 T . We could show that five enzymes, including a novel type of isomerase, are involved in the biosynthesis of this nucleotide-activated precursor of D-Fucp3NAc.
DNA Manipulations, Polymerase Chain Reaction, and Gene Identification-All standard DNA manipulation and transformation procedures were performed according to the methods described by Sambrook and Russell (23) or according to the protocols recommended by the specific manufacturers. The gene cluster was identified by chromosome walking as described previously (24). Briefly, restriction cuts of chromosomal DNA of A. thermoaerophilus L420-91 T were ligated into pBluescript II SK(ϩ) from Stratagene (Amsterdam, The Netherlands), and PCR was performed with forward primers of the known sequences and reverse primers of the plasmid. Primers were designed with attB1 and attB2 sites for the insertion into the Gateway TM vector pDonr201 (Invitrogen). E. coli DH5␣ was transformed with the vectors by electroporation with a Gene Pulser apparatus (Bio-Rad). The vectors were prepared with a Miniprep kit of Qiagen (Hilden, Germany) and sequenced by AGOWA GmbH (Berlin, Germany). PCR was performed with a PCR Sprint thermocylcer from Hybaid (Ashford, UK). The nucleotide sequences were analyzed by the sequence analysis program OMIGA TM (Accelrys, Unterhaching-Mü nchen, Germany). Homology searches and sequence aligments were performed with the BLAST tool at NCBI (25) and Multialin (26), respectively.
PCR fragments flanked by the attB1 and attB2 sites can be directionally recombined into pDonr201 in one simple step and subsequently selected for. The positive clones obtained were used to insert the genes into the expression vectors pDest15 for GST fusions or pDest17 for (His) 6 fusions. For plasmid propagation E. coli DH5␣ was used. The expression vectors gFdtA, gFdtB, and gFdtC were finally transformed by electroporation into E. coli BL21(DE3). The exact protocol for these procedures was obtained directly from the manufacturer. All plasmid constructs were sequenced for sequence verification.
Overexpression and Purification of Recombinant Proteins-Cells carrying the plasmids gFdtA, gFdtB, and gFdtC were grown in 0.7 liters of culture volumes of Luria Bertani medium containing 50 g/ml ampicillin to an optical density at 600 nm of 0.6. Expression was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h. The cells were harvested, washed in 20 mM phosphate buffer, pH 7.4, and resuspended in 10 ml of the same buffer. Cells were lysed on ice by ultrasonication, and cell debris was removed by ultracentrifugation at 331,000 ϫ g. Protein purification was performed as recommended by the manufacturer by applying the supernatant either to a HiTrap chelating column using Ni 2ϩ as the metal ion or a GSTrap column, dependent on the tag used. After dialysis against 20 mM Tris/HCl buffer, pH 7.7, the enzymes were further purified by anion exchange chromatography on a Mono Q HR 5/5 column with 0.5 M KCl as the eluent. Fractions containing the enzymes, as determined by SDS-PAGE (27), were concentrated with centrifugal filter devices Ultrafree-MC (cutoff 10,000 Da; Millipore, Vienna, Austria) and stored at 4°C or Ϫ20°C. In both cases 50% glycerol and 1 mM 1,4-dithio-DL-threitol were used as stabilizers. Protein concentrations were determined according to Bradford (28) using the Bio-Rad protein assay (Bio-Rad).
Enzyme Assays and Substrate Synthesis-All standard assays for the determination of enzyme functions contained either 50 nmol of dTDP-6-deoxy-D-xylohex-4-ulose or dTDP-D-Fucp3N and were performed in 50 mM K 2 HPO 4 buffer, pH 7.4, containing 5 mM MgCl 2 . The assay for the conversion of dTDP-6-deoxy-D-xylohex-4-ulose to dTDP-D-Fucp3N contained 5 nmol of FdtB, 5 nmol PLP, and 50 nmol of L-glutamate and was performed with and without the addition of FdtA. To test the function of the transacetylase 5 nmol of FdtC was incubated with 50 nmol of acetyl-CoA and 50 nmol of dTDP-D-Fucp3N. All assays were performed at 37°C and analyzed by reverse phase (RP)-HPLC with 0.5 M NaH 2 PO 4 solution buffer, pH 6.0, as the mobile phase at a flow rate of 0.6 ml/min (29). The eluate was monitored with a UV detector at 254 nm. Approximately 50 mg of dTDP-6-deoxy-D-xylohex-4-ulose were synthesized as described previously (30). To synthesize dTDP-D-Fucp3N, 20 mol of dTDP-6-deoxy-D-xylohex-4-ulose, 1 mol of PLP, 50 mol of L-glutamate, 200 nmol of FdtA, and 200 nmol of FdtB were incubated in 5 ml of 50 mM K 2 HPO 4 buffer, pH 7.4, and 5 mM MgCl 2 at 37°C for 1 h. The volume was reduced by vacuum evaporation to approximately 1 ml, and the material was purified by RP-HPLC with 0.2 M triethylammonium acetate buffer, pH 6.0, as the mobile phase. To remove the HPLC buffer the solution containing dTDP-D-Fucp3N was frozen and lyophilized. An overall yield of 80% of dTDP-D-Fucp3N could be achieved.
Kinetic Measurements-To determine the kinetic constants of FdtC the reaction between the free sulfhydryl group of SH-CoA, a reaction product during the transacetylation, and Ellman's reagent (5,5Ј-dithiobis(2-nitrobenzoic acid) (31) was measured continuously as the increase in absorbance at 412 nm on a Hitachi U-2010 (Maidenhead, UK) spectrophotometer. The resulting enzymatic activities were calculated with a molar absorption coefficient of 14,150 M Ϫ1 cm Ϫ1 according to the instructions of the manufacturer. A standard assay was performed in a 50 mM potassium phosphate buffer, pH 7.4, containing 5 mM MgCl 2 , 200 M 5,5Ј-dithio-bis(2-nitrobenzoic acid, 25-200 M dTDP-D-Fucp3N, 600 M acetyl-CoA, and 3 nM FdtC. To determine kinetic constants for acetyl-CoA the same assay was used, but the acetyl-CoA concentration was varied from 25 to 200 M, and the dTDP-D-Fucp3N concentration was kept constant at 600 M. To obtain the apparent Michaelis constant, K m , kinetic data were fitted directly to the Michaelis-Menten equation by using the program Sigma Plot (SPSS Science, Chicago, IL). V max and the final enzyme concentration E were used to determine the turnover number k cat of the substrates employed (30).
NMR Measurements-Spectra were recorded at 300 K at 300.13 MHz for 1 H and at 75.47 MHz for 13 C with a Bruker AVANCE 300 spectrometer equipped with a 5-mm quadrupole nuclear probe head with z gradients. Data acquisition and processing were performed with the standard XWINNMR software (Bruker). 1 H spectra were referenced to 2,2-dimethyl-2-silapentane-5-sulfonic acid (␦ ϭ 0), 13 C spectra were referenced externally to 1,4-dioxane (␦ ϭ 67.40), and 31 P spectra were dTDP-D-Fucp3NAc Biosynthesis recorded at 121.50 MHz and referenced externally to H 3 PO 4 (␦ ϭ 0). 1 H/ 13 C and 1 H/ 31 P heteronuclear multiple quantum coherence spectra and 1 H/ 13 C heteronuclear multiple bond correlation spectra were recorded in the phase-sensitive mode using time-proportional phase increments and pulsed field gradients for coherence selection. Spectra resulted from a 256 ϫ 4096 data matrix, with 1800 or 800 scans per t1 value, respectively. For in situ experiments of the isomerase reaction, two samples containing dTDP-6-deoxy-D-xylohex-4-ulose (9 mg) in 20 mM K 2 HPO 4 buffer (in 0.6 ml of D 2 O, pH 7.4) were simultaneously prepared and kept at 310 K. After adding FdtA (0.113 mg) and mixing in the NMR vial, NMR measurements were immediately started and recorded at distinct time intervals. To disprove a bifunctional activity of FdtB, the same experiment was performed with the transaminase. Briefly, 0.1 mg of FdtB was incubated with dTDP-6-deoxy-D-xylohex-4ulose (9 mg) in phosphate buffer, pH 7.4, at 310 K.

Identification of the Genes Involved in the Biosynthesis of dTDP-D-Fucp3NAc-Previously
we reported cloning and characterization of the two genes that are responsible for the formation of GDP-D-rhamnose from GDP-D-mannose in A. thermoaerophilus L420-91 T (17). Assuming that the genes for the biosynthesis of the nucleotide-activated precursors of the glycan chain are clustered, we used chromosome walking to locate the genomic region for the biosynthesis of dTDP-D-Fucp3NAc (24). Approximately 9 kilobases downstream of the rmd gene, which encodes an enzyme involved in the formation of GDP-Drhamnose (17), two overlapping PCR fragments (2 and 3.5 kilobases) were obtained. Subsequent sequencing of both fragments revealed a genomic region containing seven ORFs ( Fig.  1; GenBank TM accession number AY205257). Using BLAST searches, the first ORF of that region, consisting of 139 amino acids, showed homologies to a putative bifunctional enzyme of X. campestris, namely WxcM. Its C terminus is predicted to code for an isomerase and the N terminus for a transacetylase (14). The putative isomerase presumably responsible for the conversion of dTDP-6-deoxy-D-xylohex-4-ulose into nucleotidelinked hex-3-ulose isomer showed in addition to homologies to the X. campestris gene (AF204145), homologies to genes from Campylobacter jejuni (AF343914), Thermosynechococcus elongatus (AP005373), Listonella anguillarum (AF025396), Leptospira interrogans (AF316500), Leptospira borgpetersenii (AF078135), Streptomyces fradiae (U08223), Bacteroides fragilis (AF125164) and E. coli O91 (AY035396). With the exception of X. campestris, the presence of D-Fucp3N (either acetylated or acylated) has not been reported in the LPS or lipooligosaccharides of the Gram-negative representatives. S. fradiae, however, is a Gram-positive organism without LPS in the cell envelope. Interestingly, for E. coli O91 as well as for S. fradiae, the gluco epimer of D-Fucp3N, 3-acetamido-3,6-dideoxy-D-glucose (D-Quip3N), was found to be part of either the repeating units of LPS O-antigens or of the antibiotic tylosin (32). The second ORF (192 amino acids) was predicted to code for an acyl-/acetyltransferase. A BLAST search revealed high homologies to several acyl-and acetyltransferases that are involved in the metabolism of nucleotide-activated sugar precursors. The third ORF (363 amino acids) showed high similarities to genes of the DegT/DnrJ/EryC1/StrS family of aminotransferases. Highest homologies were achieved with sequences from organisms that possess genes similar to the putative isomerase FdtA (Table I). The presence of all three enzymes in these organisms would thus indicate the possible occurrence of modified D-Fucp3N residues in the cell envelope or metabolite glycan structures. Downstream of the transaminase two unidentified ORFs showed weak similarities to glycosyl transferases and a putative integral membrane protein. The role of these two genes was not the subject of this study. Sequence alignment of the last two ORFs of the gene cluster showed high similarity to glucose-1-phosphate thymidyltransferase (RmlA) and dTDP-D-glucose dehydratase (RmlB). These two enzymes are known to be involved in the well characterized biosynthesis of nucleotide-activated L-rhamnose of LPS (20,33) as well as of the S-layer glycoprotein of Gram-positive organisms (22). They synthesize the precursors for the biosynthesis of e.g. dTDP-D-fucose (34) and dTDP-6-deoxy-Ltalose (35). None of the encoded proteins possesses membrane-spanning domains or signal peptides for secretion, implying that the biosynthesis of this nucleotide-activated sugar occurs in the cytoplasm.
For the newly identified genes we suggest the following names: fdtA for the putative isomerase, fdtB for the transaminase, and fdtC for the transacetylase. In this designation f codes for fucosamine, d for the D configuration and t (ϭ three) indicates the substituted C-3 atom of the sugar ring. Based on the results from our work in A. thermoaerophilus L420-91 T (22) and also from previous studies on X. campestris (14,18,36), we propose a general biosynthesis pathway for D-Fucp3NAc as depicted in Fig. 2.
Overexpression and Purification of FdtA, FdtB, and FdtC-Before overexpression and purification of the enzymes the sequences of the corresponding plasmids containing the open reading frames of interest were verified by nucleotide sequencing. To confirm the function proposed for the three enzymes involved in the biosynthesis of dTDP-D-Fucp3NAc they were heterologously overexpressed in E. coli and purified by affinity chromatography and anion exchange chromatography. To overcome problems with protein stability, FdtA and FdtC were expressed with a GST tag and purified using GSTrap columns. In addition, FdtA was also expressed as N-terminal (His) 6tagged fusion protein as was FdtB. Both recombinant proteins could be highly purified on a HiTrap chelating column (Ni 2ϩ ). To improve the purity of the enzymes an anion exchange chromatography step on a Mono Q column was applied subsequently. Expression and purification was monitored by SDS-PAGE and showed that the apparent molecular mass of all proteins was in agreement with the values deduced from the nucleotide sequences (Fig. 3). To increase the stability, the enzymes were concentrated by ultrafiltration before storage at 4°C or Ϫ20°C and supplemented with 50% glycerol. The concentrated enzyme preparations were stable at 4°C for several months with only a marginal decrease of their enzyme activity (data not shown). dTDP-D-Fucp3NAc Biosynthesis profile identical to that described in a recent work (35). The isomerase product dTDP-6-deoxy-D-xylohex-3-ulose revealed an identical elution profile, and thus, distinction from the 4-keto educt was impossible (Fig. 4B). The activity of FdtA was monitored by a simultaneous incubation with dTDP-6-deoxy-Dxylohex-4-ulose with FdtB and the required co-substrates, as described under "Experimental Procedures." The resulting product, dTDP-D-Fucp3N, was detectable by HPLC analysis (Fig. 4C). The activities of FdtA as GST-and His-tagged pro-teins were also examined, and both fusion proteins showed comparable activities.
The direct conversion of dTDP-6-deoxy-D-xylohex-4-hexulose into the 3-keto product by the action of FdtA was studied by NMR spectroscopy. Before the interpretation of the NMR data, a full spectral assignment of the starting material was performed. It matched favorably with the literature data (37) and   Fig. 5 conversion into the 3-keto form occurred rapidly; however, a full assignment could not be achieved due to the small signal intensity and severe signal overlap of the remaining substrate. Comparison of the data with a non-enzymatic control reaction and assignment of signals separated from the bulk region allowed a partial assignment of the spectra. Whereas in the non-catalyzed reaction no significant conversion was seen after 2 h, the NMR spectrum of the enzymatic reaction performed at 37°C already indicated after 5 min the formation of a new component (Fig. 5). The anomeric proton of this reaction product was observed at lower field (5.86 ppm) with a 3 J H-1,P coupling constant of 7.0 Hz and a 3 J H-1,H-2 coupling constant of 4.5 Hz. The signal of H-2 at 4.93 ppm gave a coupling to the anomeric phosphate but no further spin-spin coupling to position 3, which in conjunction with the observed downfield shift would be consistent with the presence of a neighboring keto group at C-3. Moreover, an additional 6-deoxy signal was observed at 1.31 ppm ( 3 J H-6,H-5 6.5 Hz), which was shown by COSY and heteronuclear multiple quantum coherence experiments to be correlated to H-5/C-5 signals at 4.46 ppm/66.6 ppm, respectively, indicating a change of configuration at the neighboring C-4 carbon (relative to dTDP-glucose). These data are also distinctly different from those published for dTDP-6-deoxy-D-ribohex-3-ulose, obtained from a reversed transamination reaction (38) and by a non-enzymatic conversion of substrate (37). The combined evidence would, thus, be in agreement with an inversion of configuration at C-4. Based on the integration values of the anomeric protons of the intermediates, the proportion of the 3-keto product reached a maximum of ϳ10 -11% after 1 h and then slowly decreased due to formation of additional products that were not further analyzed.
Among several amino group donors tested, including L-glutamine, L-alanine, and L-aspartate, only L-glutamate resulted in a turnover. The product of the amination reaction was monitored by RP-HPLC (Fig. 4C) and was subsequently purified as described under "Experimental Procedures." To obtain better insight into the isomerization reaction and the specific activity of FdtB the 4-keto substrate was incubated without the addition of FdtA. Incubation of the reaction mixture at pH 7.4 in the absence of co-factors/co-substrates was monitored by NMR spectroscopy. The spectra recorded in a time course of 24 h did not show differences to a non-enzymatic control reaction (compare with Fig. 5A). Incubation of 4-keto substrate with FdtB and the co-factors yielded over a time span of 36 h ϳ30% conversion to dTDP-D-Fucp3N.
To test the substrate specificity of the transaminase further, the isomerase reaction was conducted for 7 h before the addition of FdtB, which would also allow the onset of nonspecific isomerization reactions. Work-up of the reaction mixture and NMR analysis revealed a single product, dTDP-D-Fucp3N, similar to the experiments where both enzymes had been added to the substrate simultaneously.
The structure of the product generated by the action of the enzymes FdtA/FdtB could be fully assigned on the basis of the NMR spectroscopic data (Table II, substance V). Thus, for the 3-amino-3-deoxy-derivative dTDP-D-Fucp3N, the signal at 53.4 ppm was due to a nitrogen-bearing carbon atom being correlated to a H-3 signal, displaying a large trans-diaxial coupling constant 3 J H-2,H-3 (11.1 Hz) and a small value for the coupling constant to H-4 ( 3 J H-3,H-4 3.2 Hz). This is consistent with a 3-amino-3,6-dideoxy-D-xylo configuration. In addition, the 13  dTDP-D-Fucp3NAc Biosynthesis activated D-Fucp3NAc. The commonly used acetyl donor of these reactions acetyl-CoA was also used in this experiment. Isolation of the final product was achieved by RP-HPLC (Fig.  4D), as described under "Experimental Procedures," and this product was also characterized by NMR spectroscopy (Table II, substance VI). The product of the N-acetylation reaction displayed signals of an N-acetyl group at 2.02 ppm/22.8 ppm for the 1 H and 13 C signals of the methyl group and signals of the pyranose unit that were in good agreement with the data of D-Fucp3NAc units observed in the native polysaccharide (10). Furthermore, an 1 H/ 31 P-correlated spectrum confirmed the presence of the diphosphoryl unit in dTDP-D-Fucp3NAc.
Because of the fact that only couple assays were possible for the isomerase and transaminase reactions, no kinetic constants were determined. However, reaction kinetics were analyzed for the transacetylation reaction. For the determination of the kinetic constants of FdtC with acetyl-CoA and dTDP-D-Fucp3N Ellman's reagent was used. Both substrates exhibited comparable affinity to FdtC. The Michaelis-Menten constants K m were 66.7 (Ϯ15.9) and 61.0 (Ϯ4.4) M for dTDP-D-Fucp3N and acetyl-CoA, respectively. The turnover numbers, k cat , were 2.3 (Ϯ 0.2) and 3.1 (Ϯ 0.1) s Ϫ1 .
General Properties of the Enzymes-All purified enzymes showed maximum activities between 50 and 55°C, although some activity could also be detected at 30°C. FdtB was also shown catalyze the reverse reaction using ␣-ketoglutarate and dTDP-D-Fucp3N as the substrate instead of L-glutamate and dTDP-6-deoxy-D-xylohex-3-ulose. By adding FdtA to the reaction mixture the equilibrium shifted toward dTDP-6-deoxy-Dxylohex-4-ulose. The reverse reaction was monitored by HPLC (data not shown). However, a reverse reaction catalyzed by FdtC could not be detected. DISCUSSION D-Fucp3NAc is a sugar component of S-layer glycoproteins (10) and LPS of several Gram-negative organisms. It was first described in X. campestris in 1963 (18). Based on this work Shibaev (19) postulated that the formation of D-Fucp3NAc involves two well characterized enzymes from the biosynthetic pathway of either nucleotide-activated L-rhamnose (40,41), dTDP-D-fucose (34), dTDP-6-deoxy-L-talose (35), or dTDP-4acetamido-4,6-dideoxy-D-glucose (42), namely RmlA and RmlB. These two enzymes catalyze the formation of dTDP-D-glucose and dTDP-6-deoxy-D-xylohex-4-ulose, respectively. The latter compound represents the precursor for the sugar biosynthesis pathway of D-Fucp3NAc.
In a previous report the gene cluster and the enzymes involved in the biosynthesis of GDP-D-rhamnose in A. thermoaerophilus L420-91 T have been identified and characterized biochemically (17). Chromosome walking revealed the presence of five open reading frames downstream of the GDP-D-rhamnose operon, presumably responsible for the biosynthesis of dTDP-D-Fucp3NAc. Two of the ORFs showed high similarities to rmlA and rmlB, and two others were homologous to the protein families of transaminases and transacetylases. The fifth gene in the cluster showed homology with a putative enzyme, which is assumed to catalyze potentially an isomerase reaction (14). To characterize the enzymes involved in the biosynthetic pathway they were expressed in E. coli and purified.
FdtA catalyzes the isomerization step in the biosynthesis pathway of dTDP-D-Fucp3NAc. In this reaction the 4-keto intermediate is isomerized to the 3-keto product, with D-xylo configuration. Several attempts to isolate the reaction product

dTDP-D-Fucp3NAc Biosynthesis
of the isomerase reaction dTDP-6-deoxy-D-xylohex-3-ulose by HPLC were unsuccessful. Thus, to confirm the function of this enzyme, on-line NMR measurements were applied. These experiments showed that FdtA catalyzes the rapid formation of dTDP-6-deoxy-D-xylohex-3-ulose in small amounts. This finding points toward the importance of a tightly controlled product formation in organisms such as X. campestris, which besides dTDP-D-Fucp3NAc also synthesizes dTDP-L-rhamnose from the same source of precursor. Here we provide the first functional description of a specific isomerase that catalyzes the transition of dTDP-6-deoxy-D-xylohex-4-ulose into dTDP-6deoxy-D-xylohex-3-ulose. FdtA shares no homologies with known isomerase families, and no common motifs could be detected. Because the enzyme lacks metal ions and oxido reduction co-factor binding motifs, the isomerization reaction catalyzed by FdtA may act by a novel mechanism. We identified gene homologs of fdtA in several Gram-negative organisms, although their specific functions are not known yet. Because of the importance of the isomerase reaction step, the presence of FdtA may imply that these organisms are able to synthesize dTDP-D-Fucp3N (14) or dTDP-3-acetamido-3,6-dideoxy-D-glucose derivatives (43). Almost all gene clusters described for these organisms also contain in the vicinity of fdtA either transaminases, transacetylases, or rml genes (Table I). Multiple sequence alignment of FdtA with 11 sequences found in the data base yielded remarkable homologies, indicating the presence of potential reaction or binding domains (Fig. 6). Further investigations will be required to identify the reaction centers and the amino acids involved in the conversion reaction. Because of the inability to transform A. thermoaerophilus L420-91 T genetically so far, we were not able to create an fdtA-deficient mutant for complementation experiments. Previous attempts to create knock-out mutants of wxcM, the fdtA-like open reading frame in X. campestris, were also not successful (14).
FdtA occurs in A. thermoaerophilus L420-91 T as a single functional enzyme, whereas in X. campestris it is part of a bifunctional enzyme also catalyzing the acetylation reaction (14). The presence of bifunctional enzymes in these pathways seems to be a common characteristic of Gram-negative organisms, whereas in Gram-positive organisms the use of single functional genes is preferred (44). Recently, Perelle et al. (43) have described a new enzyme, WbsB, which presumably is involved in the biosynthesis of the acylated quinovosamine residue ([R]-3-hydroxybutyramido-3-acetamido-3,6-dideoxy-Dglucose) occurring in the O-antigen of E. coli O91, which shows high homology to FdtA (Fig. 6). These data imply that the transition of the keto group from C-4 to C-3 of the sugar ring may follow a similar pathway, albeit resulting in the formation of the 6-deoxy-D-ribo configuration.
FdtB is a monofunctional and highly specific enzyme that catalyzes the transamination of dTDP-6-deoxy-D-xylohex-3-ulose to dTDP-D-Fucp3N. It does not possess any isomerase activity; however, it may convert catalytic amounts of non-enzymatically formed 3-keto substrate present in the reaction mixture. Previously it had been suggested that the formation of dTDP-6-deoxy-D-ribohex-3-ulose could be achieved chemically via enolization without the addition of an isomerase (37). For an efficient turnover of the substrate, however, the presence of FdtA is indispensable. Our experiments indicated that incubation of dTDP-6-deoxy-D-xylohex-4-ulose with FdtB and the corresponding co-substrates at pH 7.4 leads over time only to dTDP-D-Fucp3N.
FdtB turned out to be a member of the DegT/DnrJ/EryC1/ StrS family of aminotransferases that contains among others 3-amino-5-hydrobenzoic acid synthase, an enzyme required for the biosynthesis of ansamycin antibiotics. Crystal structure analysis of this enzyme revealed the presence of a PLP binding site, which is conserved among the aspartate aminotransferase structural family and is also present in FdtB. This specific site is characterized by a serine residue that is followed by three to five residues before the active site lysine (SX 3-5 K) (45). In the BLAST search the amino acid sequence around the PLP binding site revealed a highly conserved region, obviously a common feature of aminotransferases, which use nucleotide-activated 6-deoxy sugars as substrate. Besides the DNA sequences described in Table I, similar transferases may also be involved in the biosynthesis of several antibiotics that contain C-3aminated 6-deoxy sugars, such as oleandomycin (46), daunomycin (47), erythromycin (48), and tylosin (32). The formation of C-3 aminated 6-deoxy sugars also requires the presence of an isomerase-like enzyme, which until today was found only in the dTDP-D-Fucp3NAc Biosynthesis biosynthetic pathway of tylosin of S. fradiae (49).
The nucleotide-activated form of D-Fucp3N may also play an important role in the formation of acetylated sugar precursors as described in this study but also in other acylated carbohydrates. In several Gram-negative organisms the corresponding LPS was found to contain D-Fucp3N acylated with e.g. (R,R)-3-hydroxy-3-methyl-5-oxoproline such as in V. cholerae O5 (15) or L-2-acetoxypropionamido (50) and D-3-hydroxybutyryl groups (16) in A. baumannii. The biosynthesis of these modified D-Fucp3N units may, therefore, have similar if not identical pathways. However, because of the lack of genetic data and functional characterization of these pathways this assumption remains to be further substantiated.
FdtC catalyzes the transfer of an acetyl group to dTDP-D-Fucp3N. It is the last enzyme in this reaction sequence and contains a hexapeptide repeat sequence motif, a common feature of a number of bacterial and plant acetyltransferases. This motif is characterized by imperfect, tandem-repeated copies of the six amino acids (L/I/V)(G/A/E/D)X 2 (S/T/A/V)X (51). Recently, x-ray crystal structure analysis of N-acetylglucosamine 1-phosphate uridyltransferase, a bifunctional enzyme whose C-terminal domain is responsible for the CoA-dependent acetylation of glucose 1-phosphate in E. coli, revealed that the lefthanded parallel ␤-helix formed by the hexapeptide repeat sequence provides the pocket for the binding of the sugar substrate and catalyzes the transfer of the acetyl group (52). Only for this last step in the biosynthesis of dTDP-D-Fucp3NAc were we able to get kinetic data. The low turnover number obtained may be caused by temperatures below the optimum temperature of that enzyme. These data are in agreement with previous kinetic studies on the biosynthesis of dTDP-L-rhamnose (22). Because of the fact that we were not able to purify the 3-keto substrate for the transamination reaction, reliable kinetic data could neither be obtained for the isomerase nor the transaminase reaction.
In summary, in this work we were able to elucidate the biosynthetic pathway of dTDP-D-Fucp3NAc. This precursor is involved in the synthesis of glycan chains of glycosylated Slayer proteins as well as of LPS O-antigens. The biosynthesis involves, in addition to a transaminase and a transacetylase, a highly specific isomerase. Sequence comparison of the involved genes suggests that the pathway may also be valid in Gramnegative organisms, such as in X. campestris (14). A. thermoaerophilus L420-91 T is a moderately thermophilic organism with optimal growth temperatures in the range of 50 -55°C. In this context it is interesting to note that enzymes from thermophilic organisms, such as the Rml enzymes from A. thermoaerophilus DSM 10155, have shown a higher thermal stability than similar enzymes from mesophilic organisms (22). Therefore, exploration of their potential biotechnological application is definitely worthwhile.