Oxidation and Transamination of the 3 -Position of UDP- N -Acetylglucosamine by Enzymes from Acidithiobacillus ferrooxidans Role in the formation of lipid A molecules with four amide - linked acyl chains

Lipid A, a major component of the outer membranes of Escherichia coli and other Gram-negative bacteria, is usually constructed around a b -1 ¢ ,6 linked glucosamine disaccharide backbone. However, in organisms like Acidithiobacillus ferrooxidans, Leptospira interrogans , Mesorhizobium loti , and Legionella pneumophila , one or both glucosamine residues are replaced with the sugar 2,3-diamino-2,3-dideoxy- D glucopyranose (GlcN3N). We now report the identification of two proteins, designated GnnA and GnnB, involved in the formation of the GlcN3N moiety. The genes encoding these proteins were recognized because of their location between lpxA and lpxB in A. ferrooxidans . Based upon their sequences, the 313 residue GnnA protein was proposed to catalyze the NAD + -dependent oxidation of the glucosamine 3-OH of UDP-GlcNAc, and the 369 residue GnnB protein the subsequent transamination to form UDP 2-acetamido-3-amino-2,3-dideoxy- a - D -glucopyranose (UDP-GlcNAc3N). Both gnnA and gnnB were cloned and expressed in E. coli using pET23c+. In the presence of L-glutamate and NAD + , both proteins were required for the conversion of [ a - 32 P]UDP-GlcNAc to a novel, less negatively charged sugar nucleotide shown to be [ a - 32 P]UDP-GlcNAc3N. The latter contained a free amine, as judged by modification with acetic anhydride. Using recombinant GnnA and GnnB, ~ 0.4 mg of the presumptive UDP-GlcNAc3N was synthesized. The product was purified and subjected to NMR analysis to confirm the replacement of the GlcNAc 3-OH group with an equatorial NH 2 . As shown in the accompanying papers, UDP-GlcNAc3N is selectively acylated by LpxAs of A.


Introduction
Many Gram-negative bacteria, including some human pathogens, synthesize lipid A molecules in which one or both glucosamine residues are replaced with the analogue 2,3-diamino-2,3-dideoxy-D-glucopyranose (GlcN3N) (Fig. 1) (1-7). Acidithiobacillus ferrooxidans (3) is a good model system for studying the origin of these lipid A variants, as genomic DNA sequences are available (www.tigr.org). A. ferrooxidans grows optimally at pH ~ 2 (8), and it is often found in acidic effluents of mines (9).
ferrooxidans and other bacteria that make GlcN3N-containing lipid A, implying that the well-characterized E. coli lipid A pathway (12) remains operative. These bioinformatic observations raise the possibility that lipid A molecules containing GlcN3N might be synthesized from the hypothetical sugar nucleotide UDP 2-acetamido-3-amino-2,3dideoxy-α-D-glucopyranose (UDP-GlcNAc3N), an analogue of UDP-GlcNAc in which the GlcNAc 3-OH group is replaced with NH 2 (Scheme 1). The selective utilization of this analogue by LpxA variants present in certain bacteria could account for the formation of the observed GlcN3N-containing lipid A species ( Fig. 1 and Scheme 1).
Based on what was known about mycaminose, it seemed plausible that the enzymatic formation of UDP-GlcNAc3N might involve the NAD + dependent oxidation of the GlcNAc 3-OH moiety of UDP-GlcNAc, followed by transamination (Scheme 1).
While inspecting the lpxA region of the A. ferrooxidans genome, we noticed that two genes, one encoding an oxidoreductase and the other a transaminase, are inserted between lpxA and lpxB (Scheme 1). We reasoned that these genes might be involved in GlcN3N formation, given that lpxA and lpxB are contiguous in E. coli (16)(17)(18) and other bacteria.
We have now cloned and purified the two proteins encoded by these A. ferrooxidans genes, and demonstrate that they can convert UDP-GlcNAc to UDP-GlcNAc3N in the The desired hybrid plasmid containing of the gnnB gene ligated into NdeI/HindIII-cut pET23c+ was designated pCS331.
The contiguous gnnA and gnnB genes were also cloned in tandem as a single PCR fragment into the same expression vector. To create this construct, PCR was performed using the above gnnA N-terminal primer and gnnB C-terminal primer. The successful ligation of this PCR product into pET23c+ cut with NdeI and BamHI generated the construct pCS355. All inserts were confirmed by DNA sequencing.
To construct plasmids expressing GnnA or GnnB with a C-terminal hexa-histidine tag, two additional primers were required. The gnnA His-tag C-terminal primer was 5′ - Construction of the GnnB mutant K190A -The Stratagene "Quikchange" doublestranded PCR mutagenesis kit was used to introduce the lysine to alanine substitution at position 190 of GnnB. This mutagenesis was carried out using pCS355 or pCS484 as template, and positive clones were designated pCS601 and pCS591 respectively. To prepare extracts of A. ferrooxidans 23270, a culture was grown in 1 L of modified 9K medium (21) at 30 °C with 215 rpm rotary shaking for 90 h. The culture was harvested by centrifugation at 9 100 mM HEPES, pH 8.0, were broken by two passages through a French pressure cell at 16,000 psi. The extract was centrifuged at 10,000 x g to remove large debris and residual inorganic precipitate. The protein concentration was determined using bicinchoninic acid (25) with bovine serum albumin as the standard. The sample was stored at -80 °C.

Preparation of A. ferrooxidans cell-free extracts -
Preparation of E. coli cell-free extracts -To prepare cell extracts of the overexpressing constructs for assays, cultures were grown at 30 °C in 50 ml of LB broth containing 80 µg/ml ampicillin, as described previously (26). When A 600 had reached 0.5, the cultures were shifted to 18 °C for 10 minutes, induced with 1 mM IPTG, and grown overnight. The cultures expressing GnnB were supplemented with pyridoxine.

Ni-NTA purification of the hexa-histidine tagged GnnA and GnnB proteins -
GnnA and GnnB were purified from extracts of 500 ml BL21(DE3)/pLysS/pCS462 and BL21(DE3)/pRP(CodonPlus)/pCS484 cultures, respectively, using Ni-NTA affinity chromatography (Qiagen). A 2 ml column of Ni-NTA resin was poured using 4 ml of the commercial 50% resin/30% ethanol slurry, and was washed with 20 ml of 100 mM HEPES, pH 7.5, containing 10% glycerol. The extract then was loaded onto the column at 1 ml/min, and the flow through was collected as a single fraction. The column was washed with 20 ml of 100 mM HEPES, pH 7.5, containing 10% glycerol, before stepwise elution with 10 ml each of 25 mM, 50 mM, 100 mM, 200 mM, and 400 mM imidazole in 100 mM HEPES, pH 7.5 containing 10% glycerol. The effluent was collected in 2 ml fractions, which were analyzed by 12% SDS-PAGE and by activity assays.
Assays were carried out at 30 °C, and product formation at various times was analyzed by spotting 1 µl portions of each reaction mixture onto a PEI-cellulose plate.
The spots were allowed to air dry. The plate was then soaked in anhydrous methanol for 10 minutes and dried again. The plate was then developed using 0.2 M guanidine HCl, and was analyzed with a PhosphorImager (Molecular Dynamics Storm 840 system) equipped with Molecular Dynamics ImageQuant software (26). Varian triple resonance probes (30).

Identification of the gnnA and gnnB genes in A. ferrooxidans -The gnnA and
gnnB genes are situated between lpxA and lpxB (Supplementary Table 1 ferrooxidans. In E. coli, the lpxA and lpxB genes are contiguous, and their transcription is coupled (17,31). The fact that the reading frames of lpxA and gnnA overlap by 4 base pairs in A. ferrooxidans suggests that gnnA and gnnB may be coordinately expressed with lpxA and lpxB in this system, and may encode novel enzymes of lipid A biosynthesis.
The predicted GnnA protein is homologous to NAD + -dependent dehydrogenases of the Gfo/Idh/MocA family (32). It might therefore function to oxidize the glucosamine an Rf close to that of [α-32 P]UDP-GlcN ( Fig. 2A, Lane 7). Compound A appears to contain a free amine group, as judged by its susceptibility to modification with acetic anhydride ( Fig. 2A, Lane 4). However, Compound A is not [α-32 P]UDP-GlcN, since it is converted by acetic anhydride to substance that migrates slightly faster than [α- 32  Purification and assay of GnnA and GnnB-To characterize the above reactions in more detail, C-terminal hexa-His tagged versions of GnnA and GnnB were constructed, and following expression in E. coli, these proteins were purified by Ni-NTA affinity chromatography. Both proteins, together with both NAD + and L-glutamate (Fig. 3), are required to support the formation of Compound A. FAD + and NADP + do not substitute for NAD + (data not shown). NADH and α-keto-glutarate inhibit the reaction (Fig. 4). Other amine donors, such as alanine and glutamine, can substitute for glutamate (data not shown), but product formation is slower. The rate of formation of Compound A by purified GnnA and GnnB is relatively constant between pH 6.0 and pH 10.0.
Transamination reactions generally require pyridoxal phosphate, but the coupled GnnA/GnnB reaction observed in crude extracts (Fig. 2) does not demonstrate any stimulation by added cofactor. When the purified GnnB protein is dialyzed in the absence of glycerol, however, pyridoxal phosphate must be added back to recover activity (data not shown).
The GnnB mutant K190A is inactive-The predicted N-terminal cofactor-binding domain of GnnB is homologous to the γ-subfamily of the aspartate aminotransferase class of pyridoxal phosphate-dependent enzymes (33), which includes certain cysteine metabolizing enzymes. X-ray crystallography has shown that the cofactor-binding residue at the active site of cystathionine ß-lyase is lysine 210 (34), which corresponds to lysine 190 in GnnB, the only absolutely conserved lysine in all GnnB orthologs. The K190A mutant of GnnB is completely inactive (data not shown).
Purification of Compound A-Several ion-exchange chromatography steps were necessary to purify mg quantities of Compound A for NMR characterization. The first DEAE cellulose column, run at pH 6, separated A from the substrate UDP-GlcNAc, but not from residual NAD + . Another DEAE cellulose column was therefore run at pH 8.5, above the pK of the free amine of Compound A. Under these conditions, A was well separated from NAD + and eluted at about the same salt concentration as UDP-GlcNAc, which had already removed by the first column. This two-step process generated a highly purified triethylammonium salt of Compound A. Excess triethylamine was removed using a Dowex AG-50W 8X column in the sodium form. Capillary electrophoresis was used to validate the purity (Fig. 5).

Evaluation of the Structure of Compound A by 1D and 2D 1 H NMR Spectroscopy-
Full 1D 1 H NMR spectra of the substrate, UDP-GlcNAc at pD 7.8, and of the putative numbering scheme is shown in Fig. 6. The results demonstrate the presence of the same uracil and ribose moieties in both compounds. The spectra of both compounds also contain the characteristic methyl singlet resonance near 2.1 ppm from the 2′′ N-acetyl group ( Table 2). Peaks from residual triethylamine are evident in the UDP-GlcNAc3N sample, indicated with x (Fig. 7, upper panel, and Supplementary Figs. 2 and 3).
COSY analysis of the UDP-GlcNAc3N sample at pD 4.0 ( Supplementary Fig. 4) gave similar results, and allowed better estimates of the coupling constants from the sharp and well resolved H-2′′, H-3′′ and H-4′′ multiplets ( Fig. 7 and Table 2). Comparison of the pD 4.0 and 7.8 data reveals that protonation of the 3-amino group shifted all the glucosamine protons downfield relative to their positions at pD 7.8, while the ribose and uracil proton shifts were unchanged ( Table 2). The small J 1",2" couplings (3. The uracil H-5 (≈ 5.97 ppm) correlates to the 105 ppm carbon signal. In addition, the uracil H-6 (7.96 ppm) correlates to a 144.2 ppm carbon signal, and the 2′′ N-acetyl methyl protons correlate to a carbon signal near 24.5 ppm (Fig. 9A and Table 3).

Discussion
A significant subset of evolutionarily diverse Gram-negative bacteria synthesize lipid A molecules in which one or both glucosamine residues are replaced with GlcN3N units (10)(11)(12). This substitution results in lipid A molecules containing three or four Nlinked hydroxyacyl chains (Fig. 1) Purified GnnA, GnnB, L-glutamate, and NAD + are sufficient to support the conversion of UDP-GlcNAc to UDP-GlcNAc3N (Fig. 3). The NMR studies in Figs. 7 to 9, and in Supplementary Figs. 1 to 5, demonstrated unequivocally that a nitrogen atom is present at the pyranose 3-position of UDP-GlcNAc3N, and that it is equatorial (Fig. 6).
This means that UDP-GlcNAc3N is essentially isosteric with UDP-GlcNAc, and may function as a substrate or inhibitor of many enzymes that utilize UDP-GlcNAc.
Introduction of the GlcN3N unit in place of GlcNAc at selected sites in complex glycoconjugates could prove useful for cell surface modification or affinity labeling (41).
The homology of GnnA and GnnB to proteins of known function suggested that GnnA might be a NAD + -dependent dehydrogenase and GnnB a pyridoxal phosphatedependent aminotransferase. The simplest scenario is that GnnA oxidizes the glucosamine 3-position of UDP-GlcNAc to a ketone moiety, which is subsequently transaminated by GnnB in the presence of excess L-glutamate to generate UDP-GlcNAc3N (Scheme 1). The striking inhibition of the coupled GnnA/GnnB system by NADH or α-ketoglutararate (Fig. 4) suggests that oxidation of the glucosamine 3position of UDP-GlcNAc might be thermodynamically unfavorable, which would be consistent with the reduction potentials of NAD + and typical ketones (42). The proposed keto-sugar intermediate (Fig. 6) was not detected by TLC or capillary electrophoresis (27,28) under our conditions, and purified GnnA was inactive in the absence of purified GnnB and excess L-glutamate, as judged by spectrophotometric detection of NADH formation at 340 nm (data not shown). To demonstrate the GnnA dependent formation of the proposed 3′′-ketone intermediate in the absence of added glutamate and GnnB (Scheme 1), it may be necessary to deplete the GnnA reaction system of NADH.
The lack of significant homology of GnnA to TylA2 is consistent with the idea that GnnA carries out a direct dehydrogenation of the 3′′-position of UDP-GlcNAc.
TylA2 is necessary for the modification of the pyranose 3-position of TDP-glucose, but it does so by first oxidizing the pyranose 4′′-position (43,44) and catalyzing the reductive elimination of the 6′′-OH group. This step is followed by the action of the TylM3 isomerase, which converts the 4′′-ketone to the 3′′-ketone (15,44). No ortholog of TylM3 is required for UDP-GlcNAc3N formation, given that purified GnnA and GnnB alone are sufficient (Fig. 3). We consider unlikely but cannot yet exclude the possibility that GnnA first catalyzes the oxidation of the glucosamine 4′′-position, which is then isomerized (perhaps by GnnA itself) to the 3′′-keto-sugar prior to transamination by GnnB.
All organisms known for certain to make lipid A substituted with GlcN3N units         Coupling of the 2′′ and 3′′ protons of UDP-GlcNAc3N to inferred carbon signals near 53 and 57 ppm respectively is diagnostic of nitrogen-linked carbon atoms. Relevant cross peaks are labeled as in Fig. 6. Tables   Table S1: Selected bacterial genomes that contain GnnA and GnnB.     The HMBC-NMR spectrum was recorded at 600MHz using ~ 1 mM UDP-GlcNAc3N in 99% D 2 0 at 25 °C. Relevant cross peaks are labeled as in Fig. 6.