Biosynthesis of nucleotide-activated D - glycero -D - manno -heptose*

The glycan chain repeats of the S-layer glycoprotein of Aneurinibacillus thermoaerophilus DSM 10155 contain D - glycero -D - manno -heptose, which has also been described as constituent of lipopolysaccharide cores of Gram-negative bacteria. The four genes required for biosynthesis of the nucleotide-activated form GDP-D - glycero -D - manno -heptose were cloned, sequenced and overexpressed in Escherichia coli , and the corresponding enzymes GmhA, GmhB, GmhC and GmhD were purified to homogeneity. The isomerase GmhA catalyzed the conversion of D -sedoheptulose 7-phosphate to D - glycero -D - manno -heptose 7-phosphate, and the phosphokinase GmhB added a phosphate group to form D - glycero -D - manno -heptose 1,7-bisphosphate. The phosphatase GmhC removed the phosphate in the C-7 position, and the intermediate D - glycero - a -D - manno -heptose 1-phosphate was eventually activated with GTP by the pyrophosphorylase GmhD to yield the final product GDP-D - glycero - a -D - manno -heptose. The intermediate and end products were analyzed by high-performance liquid chromatography. Nuclear magnetic resonance spectroscopy was used to confirm the structure of these substances. This is the first report of the biosynthesis of GDP-D - glycero- a -D - manno -heptose in Gram-positive organisms. In addition, we propose a pathway for biosynthesis of the nucleotide-activated form of L - glycero -D - manno -heptose.


INTRODUCTION
The cell surface of many archaea and bacteria is composed of crystalline two-dimensional protein arrays, termed S-layers 1 (1). Frequently, the S-layer proteins are glycosylated (2). The S-layer glycoprotein of the Gram-positive bacterium Aneurinibacillus thermoaerophilus DSM 10155, a member of the Bacillus/Clostridium group, is composed of disaccharide repeating units of α-L-rhamnose and D-β-D-heptose units (3). So far, this is the only Gram-positive bacterium, where D,D-heptose has been described as constituent of a cellular component. The principal architecture of S-layer glycoproteins resembles that of the LPS of Gram-negative bacteria (4). Both glycoconjugates exhibit a tripartite structural organization, where, in general, conserved core regions connect a glycan chain, composed of identical repeating units, either with the S-layer polypeptide or the lipid A of LPS. It has been proposed that comparable pathways are used for the biosynthesis of these similar glycoconjugates (2). Recently we were able to verify this proposal by investigation of the biosynthesis of nucleotide-activated D-rhamnose (5) and L-rhamnose (Graninger, M., Kneidinger, B., Bruno, K., Messner, P., unpublished) in A. thermoaerophilus strains L420-91 T and DSM 10155, respectively. L,D-heptose is a common constituent of the LPS inner core of enteric and non-enteric bacteria (6). D,D-heptose, on the other hand, has been described as component of the outer core region of LPS. Numerous bacteria, most importantly the pathogens Proteus vulgaris R110/1959 (7), Haemophilus ducreyi (8), Klebsiella pneumoniae ssp. pneumoniae R20 (O1 -:K20 -) (9) and Helicobacter pylori AF1 and 007 (10) have been shown to contain both types of heptose in LPS: L,D-heptose in the inner core and D,D-heptose in the outer core. In Yersinia enterocolitica Ye75R both D,D-heptose and L,D-heptose are part of the inner core (11). It has to be noted, however, that only part of the existing literature on D,D-heptose and L,D-heptose is listed above, and coverage is confined to those references where complete structural studies have been performed. L-Rhamnose, the second sugar of the repeating unit of the Slayer glycan of A. thermoaerophilus DSM 10155 (3), is usually a constituent of the outer core region as well as of the O-antigen of LPS (6,12).
The inner core oligosaccharide of most LPS-containing bacteria consists of 3-deoxy-D-manno-oct-2-ulosonic acid and L,D-heptose units. The outer core, however, is composed predominantly of hexoses and hexosamines (13), but D,D-heptose has also been reported recently (see above). Impairment of 3-deoxy-D-manno-oct-2-ulosonic acid biosynthesis usually results in nonviable cells (6). On the other hand, mutants having a defect in the biosynthesis of L,D-heptose are viable, albeit with certain characteristics, referred to as the "deep rough" phenotype in Escherichia coli and Salmonella typhimurium. Transduction by the P1 bacteriophage and F-plasmid conjugation are impaired (14,15), and the mutants show increased sensitivity to detergents, bile salts and hydrophobic antibiotics (16). Reduction of virulence has been reported for Haemophilus influenzae (17) and S. typhimurium (18).
Crystallization of the biosynthesis enzymes could lead to the development of enzyme inhibitors via molecular modeling. These inhibitors can support novel antibiotic therapies to circumvent drug-resistance.
The biosynthesis of the nucleotide-activated precursor of 3-deoxy-D-manno-oct-2-ulosonic acid, namely CMP-3-deoxy-D-manno-oct-2-ulosonic acid, has been described previously (13). In contrast, the complete biosynthesis of the nucleotide-activated form of D/L,D-heptose has not yet been elucidated. Eidels and Osborn (19) have proposed a four-step pathway for the synthesis of NDP-L,D-heptose: i. D-sedoheptulose 7-phosphate is converted to D,D-heptose 7-phosphate by phosphoheptose isomerase; ii. a mutase catalyzes the second step to form D,D-heptose 1-phosphate. iii. this intermediate product is activated to NDP-D,D-heptose by the action of NDP-heptose synthetase, and iv. in the last step an epimerase catalyzes the formation of the final product NDP-L,D-heptose. The nucleotide-activated sugar GDP-D,D-heptose has been described in baker's yeast (20), and both ADP-D,D-heptose and ADP-L,D-heptose have been isolated from Shigella sonnei and Salmonella minnesota mutants (21,22). Recently, it has been shown that heptosyltransferases I and II from E. coli accept ADP-L-β-D-heptose as substrate, and the D-β-D isomer is also accepted, albeit with tenfold reduced efficiency (23).
The first step in heptose biosynthesis was described in S. typhimurium (24), and the phosphoheptose isomerase gene gmhA from E. coli was cloned and sequenced (25). So far, no gene product, which catalyzes the proposed mutase step has been described. The RfaE protein, also referred to as ADP-heptose synthetase is proposed to be involved in the nucleotide-activating step. This enzyme has been shown to consist of two separate domains in E. coli (26). Domain I displays homology to members of the ribokinase family, whereas domain II is homologous to various cytidyltransferases. Thus, RfaE is supposed to be a bifunctional enzyme involved in the biosynthesis of D,D-heptose 1-phosphate as well as in the nucleotide-activating reaction step. The rfaD gene product has been proposed to carry out the last step in the biosynthesis of the L,D-form, namely the epimerization to yield NDP-L,D-heptose (27). Recently, RfaD has been crystallized and characterized (28). So far, functional studies have only been performed for the isomerization reaction (24,25) and for the epimerization step (29). Enzymes that might carry out the mutase reaction, or an alternative phosphatase have not been described.
In this report we show for the first time the overall biosynthesis of a nucleotide-activated form of D,D-heptose, namely GDP-D-α-D-heptose in Gram-positive bacteria and propose a general reaction pathway for the biosynthesis of activated L,D-heptose in Gram-negative bacteria. In addition, we present a modified nomenclature scheme for genes involved in heptose biosynthesis.

EXPERIMENTAL PROCEDURES
Materials: ATP, GTP, D-sedoheptulose 7-phosphate and dithiothreitol were obtained from Sigma (Sigma Aldrich Fluka GmbH, Wien, Austria). D-glycero-α-D-manno-heptose 1-phosphate and D-glycero-β-D-manno-heptose 1-phosphate were kindly pro by A. Zamyatina     Cloning of the heptose genes: In order to clone the genes involved in S-layer glycoprotein glycan biosynthesis, particularly the dTDP-L-rhamnose genes, an alignment of 14 RmlA and putative RmlA protein sequences, available in the NCBI database, was carried out using Multalign (34). The highly conserved six amino acid stretch LGDNIF/Y was used to design the degenerate probe 5´-YTI GGI GAY AAY ATH TT-3´ (where I is inosine, H is A/C/T and Y is C/T). This oligonucleotide was 3´-end-labeled with digoxigenin, and Southern hybridization experiments of completely digested chromosomal DNA of A. thermoaerophilus DSM 10155 yielded specific signals at 35 °C. A 4.5 kb BglII-fragment was isolated by preparative agarose gel electrophoresis and cloned into the plasmid pBCKS, linearized with the endonuclease BamHI. The corresponding construct pRML was sequenced.
Plasmid construction: Oligonucleotide primer sequences are given in Table I.
Primers for the amplification of DNA fragments containing either the gmhA (primers GMHAA1 and GMHAA2), the kinase (primers GMHBA1 and GMHBA2) or the pyrophosphorylase (primers GMHDA1 and GMHDA2) gene were designed with attB1 or attB2 sites for the insertion into the GATEWAY donor vector pDONR201 by homologous recombination. The PCR products were cloned into pDONR201 and the resulting plasmids gGMHA1, gGMHB1 and gGMHD1 were used to transfer the gene sequences into pDEST15  Enzyme assays: Ten nmol of D-sedoheptulose 7-phosphate or α-D,D-heptose 1-phosphate were used for enzyme assays. 50 nmol α-D-glucose 1,6-bisphosphate was used in a negative control assay to test the specificity of D,D-heptose 1,7-bisphosphate phosphatase. The assay buffer TH8 contained 20 mM Tris-HCl, pH 8.0 and 10 mM MgCl 2 . Appropriate amounts of enzymes were added, and after 45 min incubation at 37 °C the samples were analyzed by HPAEC.

Enzymatic synthesis of D,D-heptose phosphates and GDP-D-α-D-heptose:
One µmol D-sedoheptulose 7-phosphate was incubated with appropriate amounts of GmhA in TH8 buffer at 37 °C until equilibrium was reached. The enzyme was removed by ultrafiltration using Ultrafree-MC 10000 ultrafiltration cartridges. D,D-heptose 7-phosphate was separated from D-sedoheptulose 7-phosphate by preparative HPAEC on a CarboPac PA-1. The sample was desalted using a Carbohydrate Membrane Desalter (CMD; Dionex), lyophilized and investigated by NMR spectroscopy. D,D-heptose 7-phosphate, purified in the first reaction step (approximately 150 nmol), was incubated with 250 nmol ATP and the kinase in 20 mM triethylammonium bicarbonate buffer (pH 8.5; 10 mM MgCl 2 ) at 37 °C. After removal of the enzyme, the pH was adjusted to 3.0 by addition of Dowex-50 ion exchange resin, the sample was lyophilized and D-α-D-heptose 1,7-bisphosphate was analyzed by NMR spectroscopy. 300 nmol D-sedoheptulose 7-phosphate were converted to D-α-D-heptose 1-phosphate, using appropriate amounts of GmhA, the kinase, the phosphatase and 600 nmol ATP in TH8 buffer at 37 °C.

RESULTS
Cloning of the heptose operon: The genes coding for enzymes involved in the biosynthesis of a nucleotide-activated sugar precursor generally are clustered within the corresponding gene cluster for a particular bacterial polysaccharide (38). For the genes in the dTDP-L-rhamnose biosynthetic pathway this proposal was verified in Gram-negative as well as in Gram-positive organisms (39, 40, Graninger, M., Kneidinger, B., Bruno, K., Messner, P., unpublished). To clone the genes involved in S-layer glycoprotein glycan biosynthesis of A. thermoaerophilus DSM 10155, a strategy was applied using a degenerate probe derived from RmlA protein sequences (Fig. 1). Characterization of the rml genes will be published elsewhere. The insert of the cloned construct pRML was sequenced and contained the rmlA gene, incomplete at the 3´-end, and four additional ORFs upstream of rmlA. One of these open reading frames codes for a homologue of GmhA, the first enzyme involved in the biosynthesis of nucleotide-activated heptose. The fourth ORF was incomplete, and its 5´-end was sequenced using a recently described method (41), revealing two additional ORFs. Again, the 5´-end of the most upstream gene was missing.
Sequence analyses: The physical map of the sequenced 6,652 bp DNA fragment is depicted in Fig. 1 (42,43). An alignment of the amino acid sequences encoded by these hypothetical ORFs was performed using Multalign (34, see Fig. 2). The six hypothetical proteins showed remarkable homology, and thus, are expected to carry out the same catalytic reaction (Table II) (Fig. 5C). Final characterization of D-α-D-heptose 1-phosphate was performed by NMR analysis, which displayed a single 31 P signal δ 2.53 coupled to the anomeric proton (δ 5.32). Furthermore, the spectral characteristics were in good agreement with the synthetic compound as well as with published data (45). To test the specificity of the phosphatase α-D-glucose 1,6-bisphosphate was used as a substrate for this enzyme. Similar to the kinase negative control the peak pattern did not change. Whereas the first reaction step, catalyzed by GmhA, is an equilibrium reaction, the next two reactions appear to proceed quantitatively.

Activation of D-α-D-heptose using GTP:
The final step in the biosynthetic pathway of nucleotide-activated D,D-heptose is the transfer of an NDP residue to D,D-heptose in the C-1 position. Guanosine and adenosine had previously been described as the activating nucleotide (20)(21)(22). To test the proposal, which nucleotide is actually used in the Grampositive organism A. thermoaerophilus DSM 10155, D-α-D-heptose 1-phosphate was incubated with the putative pyrophosphorylase and ATP or GTP, respectively. The reaction mixtures were analyzed by HPAEC using a CarboPac PA-1 column. No new peak appeared, when ATP was used in the reaction (Fig. 6A). However, when the guanosine nucleotide was used, the GTP peak decreased, and a new peak with a retention time slower than GMP (13.3 min), but faster than GDP (37 min), was detected (Fig. 6B). The retention time of 25.9 min was similar to that of GDP-D-mannose (28.4 min). The reaction was also carried out with CTP, dTTP and UTP, but no additional peaks could be detected (data not shown).
Thus, guanosine diphosphate is the activating nucleotide of D-α-D-heptose in the biosynthesis of the S-layer glycoprotein glycan of A. thermoaerophilus DSM 10155. The end product of the pathway was fully characterized using NMR analysis. The structure of the sugar nucleoside diphosphate, GDP-D-glycero-α-D-manno-heptopyranose was unambiguously confirmed by the NMR data and by comparison with both anomers of synthetic ADP-D-glycero-D-manno-heptopyranose (Fig. 7). The proton and carbon NMR signals of position 8 of the guanine unit were observed at δ 8.09 and δ 138.6, respectively. In addition, the 1 H NMR signals of the β-D-ribofuranosyl residues were clearly separated from those of the heptose, which allowed the complete assignment of all proton signals.
Furthermore, HMQC experiments allowed a partial detection of the connected carbons (Table III) determination of the pathway, as has the inability to identify a gene encoding an enzyme catalyzing the mutase step, proposed in the biosynthetic pathway of Eidels and Osborn (19).
In addition, until now, genetic complementation of heptose-deficient mutants did not identify an enzyme with phosphatase homology. In the Gram-positive bacterium A. thermoaerophilus DSM 10155, however, four enzymes, the genes of which are part of the same operon, are sufficient to synthesize GDP-D-α-D-heptose from D-sedoheptulose 7-phosphate. This operon extends in the upstream direction, but the whole cluster has not been sequenced yet.
Upstream of the four heptose genes are hypothetical ORFs, that encode putative glycosyl transferases (Fig. 1), likely involved in S-layer glycan synthesis. and Haemophilus influenzae Rd. Accession numbers of putative heptose biosynthesis genes of the completely sequenced bacterial strains are given in Table II. For brevity of this work, only one of the sequenced strains of Helicobacter pylori (26695) and Neisseria meningitidis (MC58) was taken into account, but the genes are found in the genomes of additional strains.
Judging from the high homology throughout the whole protein sequence, all of these enzymes are expected to carry out the same catalytic reaction (Fig. 2). For the conversion of the anomeric mixture of D,D-heptose 7-phosphate to D,D-heptose 1-phosphate it seems reasonable that in Gram-positive and Gram-negative organisms either the αor the β-anomer is taken up selectively by a phosphokinase (Fig. 8). The subsequent removal of the 7-phosphate group is expected to be performed by a phosphatase. Such phosphatases can also be found in Pseudomonas aeruginosa PAO1 and Neisseria meningitidis MC58 and Z2491, however, with lower homologies (Table II). To date, all heptose-containing bacteria, that have been completely sequenced, contain a gene encoding a protein homologous to the phosphatase GmhC of A. thermoaerophilus. Mycobacterium tuberculosis H37Rv also has a gene homologous to this phosphatase despite the fact that there is currently no report of heptose-containing saccharides in this organism. Homologues of the first three enzymes of the heptose pathway are part of a single operon, whereas the last enzyme can also be found elsewhere on the chromosome in this strain.
As far as currently known from Gram-negative bacteria (except N. meningitidis, see below), enzymes for heptose biosynthesis appear to be encoded by four genes, one of which encodes a bifunctional protein (26). This bifunctional enzyme putatively catalyzes two  (46). Additional glycoconjugates are known to occur in the cell envelope of this organism (47). Although there was no indication for the presence of heptose in these glycoconjugates two separate sets of genes would potentially facilitate for incorporation of D,D-heptose or L,D-heptose into the corresponding polysaccharide structure.
The biosynthetic pathway of heptose seems to diverge during the kinase step (Fig. 8). For the phosphatase step, which does not involve the anomeric phosphate group, the homology to the corresponding genes is high (Fig. 2). The D,D-heptose 1-phosphate guanosyltransferase accepts D-α-D-heptose 1-phosphate, whereas the D,D-heptose 1-phosphate adenosyltransferase would accept the putative D-β-D-heptose 1-phosphate.

The kinase GmhB from
Recently, a general nomenclature for bacterial polysaccharide synthesis has been introduced (48). The three-letter gene name gmh (glycero-manno-heptose) is used for the heptose biosynthesis genes. As we have shown in A. thermoaerophilus, there are four steps to get GDP-activated D,D-heptose from D-sedoheptulose 7-phosphate. Thus, the suffixes A-D are sufficient to describe this pathway (Fig. 8). In Gram-negative bacteria, however, suffixes A-E would be required to identify the five steps including the epimerization step to make L,D-heptose. gmhA has been assigned to sedoheptulose 7-phosphate isomerase and gmhD has been proposed for the ADP-L,D-heptose epimerase gene. Since the latter enzyme catalyzes the fifth step in the heptose pathway of Gram-negative bacteria, we suggest this gene to be named gmhE. gmhB with subscript α or β might be assigned to the gene encoding D,D-heptose 7-phosphate kinase, and the proposed name for the gene encoding the D,D-heptose 1,7-bisphosphate phosphatase should be gmhC. Finally, gmhD with the subscript NDP (e.g. gmhD ADP or gmhD GDP ) might be assigned to the gene encoding the enzyme carrying out the D,D-heptose 1-phosphate nucleosyltransferase step. As described above, a bifunctional enzyme, currently named ADP-heptose synthetase, putatively catalyzes the addition of a phosphate group to D,D-heptose 7-phosphate and the activation of D,D-heptose 1-phosphate. In our scheme these two reactions are catalyzed by gmhB and gmhD, respectively. Therefore, gmhBD would be an appropriate name for the gene encoding this bifunctional enzyme (Fig. 8).
In this study we have shown biosynthesis of GDP-D-α-D-heptose in a Gram-positive bacterium. To test our proposal for the heptose pathway, the enzymes gmhBD and gmhC from E. coli are currently under investigation. As there are no genetic tools presently available for the Gram-positive bacterium A. thermoaerophilus, knock-out experiments with the D,D-heptose 1,7-bisphosphate phosphatase gene gmhC will be performed in E. coli K-12 to assess its involvement in heptose biosynthesis.