Identification of the namH gene, encoding the hydroxylase responsible for the N-glycolylation of the mycobacterial peptidoglycan.

The peptidoglycan of most bacteria consists of a repeating disaccharide unit of beta-1,4-linked N-acetylmuramic acid and N-acetylglucosamine. However, the muramic acid moieties of the mycobacterial peptidoglycan are N-glycolylated, not N-acetylated. This is a rare modification seen only in the peptidoglycan of mycobacteria and five other closely related genera of bacteria. The N-glycolylation of sialic acids is a unique carbohydrate modification that has been studied extensively in eukaryotes. However, the significance of the N-glycolylation of bacterial peptidoglycan is unknown. The goal of this project was to identify the gene encoding the hydroxylase responsible for the N-glycolylation of the mycobacterial peptidoglycan. We developed a novel assay for the mycobacterial UDP-N-acetylmuramic acid hydroxylation reaction and demonstrated that Mycobacterium smegmatis has an enzyme activity that can convert UDP-N-acetylmuramic acid to UDP-N-glycolylmuramic acid. We identified the gene namH encoding the mycobacterial UDP-N-acetylmuramic acid hydroxylase by computer data base searching and motif comparisons with the eukaryotic enzymes responsible for the N-glycolyation of sialic acids. The namH gene is not essential for in vitro growth as we were successful in deleting the gene in M. smegmatis. The M. smegmatis mutant is devoid of UDP-N-acetylmuramic acid hydroxylase activity and synthesizes only N-acetylated muropeptide precursors. Furthermore, the mutant exhibits increased susceptibility to beta-lactam antibiotics and lysozyme. Our studies suggest that the N-glycolylation of mycobacterial peptidoglycan may play a role in lysozyme resistance or may contribute to the structural stability of the cell wall architecture.

In a manner similar to the outer membrane of Gram-negative bacteria, the outer portion of the cell envelope of mycobacteria is believed to be arranged as an asymmetric bilayer, with an outer layer composed of various lipids with short to medium length fatty acids, and an inner layer composed of long chain mycolic acids covalently attached to the arabinogalactan polysaccharide (1,2). The arabinogalactan is in turn covalently attached to the peptidoglycan via an essential rhamnose-Nacetylglucosamine linker, forming the mycolylarabinogalactanpeptidoglycan complex (mAGP) (3,4). The peptidoglycan ultimately serves to anchor the principle mass of the mycobacterial cell envelope.
Peptidoglycan is an essential component of the cell envelope of virtually all bacteria, providing both shape and structural integrity to the cell. Peptidoglycan fragments are also biologically active molecules with toxigenic and immunomodulatory properties (5)(6)(7)(8). In general, peptidoglycan is comprised of a ␤-1,4-linked polymer of N-acylated muramic acid and glucosamine carbohydrates with cross-linked peptides of varying composition attached to the muramyl moieties. In virtually all bacteria, both carbohydrates in the glycan chain are N-acetylated, however, in the mycobacteria the muramic acid moieties are N-glycolylated instead (9). Only five other genera of bacteria, Rhodococcus, Tsukamurella, Gordonia, Nocardia, and Micromonospora have N-glycolylmuramic acid in their peptidoglycan (10). These five genera are closely related to mycobacteria, and all six belong to the class Actinomycetales. Why N-glycolylated muramic acid is only produced by a small number of bacteria is unknown. N-glycolylation is a rare carbohydrate modification seen only for muramic acid in these actinobacteria and for neuraminic acids (sialic acids) in eukaryotes (11). The N-glycolylation of sialic acids in eukaryotes is known to confer differential biological activities upon the carbohydrates. However, it is not known what role N-glycolylated muramic acid plays in peptidoglycan biology, although it has been hypothesized to stabilize the cell wall through hydrogen bonding (2).
The initial discovery of N-glycolylated muramic acid (Mur-NGlyc) 1 in mycobacteria prompted the hypothesis that it was synthesized from an N-acetylated muramic acid (MurNAc) precursor by the action of a monooxygenase enzyme, known generically as a hydroxylase (12). These enzymes typically require a proton donor, usually NADH or NADPH, for the reduction of molecular oxygen to water prior to the insertion of an oxygen atom into the substrate and forming a hydroxyl group (13). Previous work with Mycobacterium phlei established that the production of N-glycolylmuramic acid in mycobacteria likely occurs by the action of an atmospheric oxygen-dependent hydroxylase that converts UDP-MurNAc to UDP-MurNGlyc (14). In addition, a soluble activity present in Nocardia asteroides capable of converting UDP-MurNAc to UDP-MurNGlyc in an NADPH-dependent fashion was demonstrated almost 30 years ago (15), but since then no UDP-MurNAc hydroxylase has been purified and no gene identified from any bacterial species.
Our laboratory studies the biosynthesis of the mycobacterial peptidoglycan and we are particularly interested in determining the significance of N-glycolylmuramic acid in the cell wall of these organisms. In this study, we report the development of a new assay to detect the conversion of UDP-MurNAc to UDP-MurNGlyc and the identification of the namH gene (N-acetylmuramic acid hydroxylase) encoding the mycobacterial UDP-MurNAc hydroxylase. We also show that while the namH gene is not essential for the survival of Mycobacterium smegmatis, a namH deletion mutant is hypersusceptible to ␤-lactam antibiotics and lysozyme.
Plasmid Construction-Plasmid DNA was purified using Qiagen columns (Qiagen Inc., Chatsworth, CA), and DNA fragments were isolated using GeneClean (Bio101, Vista, CA). Suicide plasmids pMP252 (⌬blaS1) and pMP354 (⌬namH1), were constructed in E. coli K-12 DH10B, whereas the M. smegmatis namH ϩ complementing plasmid pMP336 was constructed in the M. smegmatis ⌬namH1 mutant PM979. All plasmids were confirmed by restriction mapping and then sequencing by ACGT (Northbrook, IL). Detailed descriptions of the construction of the plasmids used in this study can be obtained from the corresponding author.
Construction of the M. smegmatis ⌬blaS1⌬namH1 Mutant PM979 -The M. smegmatis ⌬blaS1⌬namH1 mutant was constructed by allelic exchange using sacB as a counterselectable marker (20). This was accomplished in a two-step process beginning with the deletion of the M. smegmatis major ␤-lactamase blaS (18), resulting in strain PM965. Subsequent allelic exchange of the wild-type namH gene in PM965 was accomplished using pMP354, resulting in the double mutant PM979.
Preparation of M. smegmatis Whole Cell Extracts-M. smegmatis strains harboring plasmids pMV261.hyg or pMP336 (pMV261.hyg bearing namH ϩ )were inoculated into 100 ml 7H9 medium containing hy-gromycin and grown overnight at 37°C with shaking. The following morning, cells were harvested at 3,000 ϫ g at an OD 600 of ϳ0.6 -0.8, and cell pellets were washed once in ice-cold 50 mM Tris, pH 8.0. The washed cells were centrifuged, and pellets were resuspended in 3 ml of the same buffer containing DNase and RNase at a final concentration of 100 g/ml and protease inhibitor (Calbiochem, San Diego, CA). Cell suspensions were lysed in a French pressure cell (Aminco, Urbana, IL) three times at 15,000 psi, and cellular debris was pelleted at 10,000 ϫ g. The membrane fraction was obtained after centrifugation of the lysate at 100,000 ϫ g for 1 h at 4°C. The supernatant consisting of the soluble cytoplasmic fraction was saved on ice and the membrane pellet resuspended in 400 l of ice-cold 50 mM Tris, pH 8.0.
UDP-MurNAc Hydroxylase Assay-UDP-MurNAc hydroxylase reactions contained UDP-MurN[ 14 C]Ac (2 ϫ 10 9 Bq/mmol) (1 nmol), ␤-NADPH (250 nmol) in 50 mM Tris pH 8.0 buffer with 1 mM dithiothreitol. Mixtures were prewarmed to 37°C, and reactions were initiated by the addition of a freshly prepared M. smegmatis whole cell extract, soluble extract, or membrane fraction (2 mg of total protein/ reaction) for a final reaction volume of 300 l. Reactions were incubated at 37°C for 15 min at which time protein was removed from the reactions using an Amicon Centriplus YM-10 filter device (Millipore Corp., Bedford, MA) by centrifugation at 7,500 ϫ g for 1 h at 4°C. Following protein removal, 150 l of each sample was added to 150 l of 8 N HCl and hydrolyzed at 100°C for 4 h to liberate nucleotide sugar acyl groups (acetic or glycolic acid), then the mixture was neutralized by the addition of 150 l of 8 N NaOH. Components of the acid-hydrolyzed reaction mixtures were separated by anion exchange HPLC using a Waters 2695 separation module (Waters Corp., Milford, MA) and a Shodex KC-811 column (Waters Corp) equilibrated with 0.1% phosphoric acid. 50-l samples were injected into the column using an isocratic flow of 0.1% phosphoric acid at 1 ml/min. Separations were performed at 60°C for 15 min. Fractions were collected at 10-s intervals from 8 to 12.5 min post-injection based upon control separations of glycolic and acetic acid, which possessed retention times of 9.5 and 10.8 min, respectively. We routinely added unlabeled acetic acid and glycolic acid standards to the reaction mixtures prior to acid hydrolysis as internal controls. The unlabeled acids are detected by the absorbance at 210 nm. 100 l of each fraction was added to 5 ml of Ecolite ϩ scintillation fluid (ICN, Costa Mesa, CA) for detection of the labeled acids in a Beckman LS-1801 liquid scintillation counter. Radioactive counts were tallied under the organic acid peaks and activity expressed as pmol of UDP-MurNGlyc produced/min/mg protein.
Purification and Analysis of UDP-acylmuramyl Pentapeptides-Cells for muropeptide preparation were harvested from mid-exponential phase cultures (OD 600 of 0.4 -0.8) that were quickly chilled, the cells washed in ice-cold phosphate-buffered saline, pH 7, pelleted, and frozen at Ϫ20°C. Ten grams of cell pellet were suspended in ice-cold 50 mM MOPS buffer, pH 8, and subjected to probe sonication on ice. The cell lysate was centrifuged at 27,000 ϫ g and the supernatant transferred to a new tube to which trichloroacetic acid was added at a concentration of 10%. The mixture was stirred on ice and centrifuged at 15,000 ϫ g and 4°C. The clear supernatant was transferred to a new tube, and the trichloroacetic acid removed by three extractions with diethyl ether. The resulting solution was dried on a rotary evaporator and reconstituted in water and loaded on a Sephadex G-25 (116 ϫ 2.5 cm) column equilibrated with 75 mM ammonium acetate (pH 5). The column was calibrated with authentic UDP-acylmuramyl pentapeptide prepared using recombinant E. coli enzymes (4). Fractions containing UDPacylmuramyl pentapeptide were pooled and lyophilized to remove ammonium acetate. These partially purified nucleotide-linked precursors were resuspended in 2 M trifluoroacetic acid and incubated at 60°C for 1 h to remove the UDP moiety. The resulting muropeptides were further purified by size exclusion chromatography on a Superdex peptide 10/300 GL column, equilibrated and eluted with 30% acetonitrile containing 0.1% trifluoroacetic acid. An aliquot of the muropeptide-containing fraction was applied to a Hypersil ODS C 18 column connected to an 1100 HPLC system (Agilent Technologies. Palo Alto, CA). The eluent was directly introduced into a LCQ Duo electrospray mass spectrometer (Finnigan-Theromquest, San Jose, CA), and the muropeptides were analyzed by mass spectrometry (MS).
Acylated muramic acid was analyzed by GC-MS after treatment of purified muropeptides with mutanolysin to remove the peptides from the muramic acid. The reactions were deproteinated by ethanol precipitation and the supernatant dried under vacuum. The samples were resuspended in 3 N methanolic-HCl, heated to 80°C for 1 h, cooled to room temperature, dried under a stream of N 2 , and Tri-Sil reagent was added. The mixtures were heated at 70°C for 20 min., cooled to room temperature and the reagent evaporated under a stream of N 2 . The derivatized products were dissolved in a small volume of hexane and analyzed in a ThermoQuest Trace GC 2000 gas chromatograph linked to a Finnigan Polaris mass detector.
M. smegmatis Antibiotic and Lysozyme Susceptibility Assays-Determination of antibiotic susceptibility was done by disk diffusion assays, while lysozyme susceptibilities were assayed by minimal inhibitory concentration tests using the macrodilution method as previously described (18). Minimum inhibitory concentrations (MIC) of lysozyme were noted as the lowest concentration at which no microbial growth could be observed or a noticeable growth defect (granular clumping) had occurred.
M. smegmatis Lysozyme Growth Curves-M. smegmatis strains were grown to saturation in 7H9 broth and then diluted 10-fold into fresh 7H9 medium and incubated at 37°C. Optical density readings were recorded at 30-min intervals. Upon reaching an OD 600 of ϳ0.4, the cultures were split into two, and freshly prepared lysozyme was added to a final concentration of 20 g/ml to one of each of the split cultures of each strain. Readings were continued at 30-min intervals until each culture reached an OD 600 of ϳ1.0.

Development of the UDP-MurNAc Hydroxylase Assay-UDP-
MurNAc hydroxylase activity was first described in the actinobacterial species N. asteroides nearly 30 years ago using an assay that monitored the NADPH-dependent conversion of radiolabeled UDP-MurNAc to UDP-MurNGlyc (15). The substrate and product were detected using paper chromatography with comparison to chemically synthesized standards. Unfortunately, this assay was difficult for us to reproduce without UDP-MurNAc and UDP-MurNGlyc for comparison, as neither of these compounds is commercially available. Therefore, we decided to develop a new, HPLC-based assay, for the detection of the mycobacterial UDP-MurNAc hydroxylase activity, using UDP-MurNAc synthesized in our laboratory from commercially available UDP-GlcNAc.
Our initial attempts to assay the hydroxylase using UDP-MurNAc or radiolabeled UDP-MurNAc synthesized from UDP-GlcNAc in which the hexosamine backbone was 14 C-labeled, were unsuccessful (data not shown). In each case, we were not able to unambiguously determine if UDP-MurNGlyc was produced in our assays, as we had no authentic UDP-MurNGlyc available for comparison. To overcome this problem, we used UDP-MurNAc in which the acetyl group of the nucleotide sugar was 14 C-labeled in order to facilitate the identification of UDP-MurNGlyc ("Experimental Procedures"). The detection method is based upon the observation that acid hydrolysis of N-acylated hexosamine sugars liberates the acyl group as a free acid. Therefore, acid hydrolysis of the hydroxylase assay reaction mixtures would release the labeled acyl groups of UDP-MurNAc and UDP-MurNGlyc as [ 14 C]acetic acid and [ 14 C]glycolic acid, respectively. These acids are easily distinguished from each other by their separation profiles in anion exchange HPLC.
A schematic representation of our mycobacterial UDP-Mur-NAc hydroxylase assay is shown in Fig. 1. Completed reaction mixtures were clarified using an Amicon filtration device, and then subjected to acid hydrolysis. The components of the hydrolysate were separated by anion exchange HPLC for 15 min, and the fractions containing the 14 C organic acids were detected by scintillation counting. The radioactive peaks were compared with our established retention times for unlabeled glycolic acid (9.5 min) and unlabeled acetic acid (10.8 min) standards. Because the acid hydrolysis is essentially 100% efficient in the removal of acyl groups from nucleotide sugars (data not shown) the amount of [ 14 C]glycolic acid detected after acid hydrolysis of the reaction components corresponds to the amount of UDP-MurNGlyc produced.
The UDP-MurNAc hydroxylase activity in reactions containing 2 mg of protein of a whole cell extract of wild-type M. smegmatis strain PM965 was 23 pmol of UDP-MurNGlyc/ min/mg protein (Table I). The activity was present in a soluble fraction prepared from the cell extract but was not present in the membrane fraction (Table I). The addition of a 10-fold excess of unlabeled UDP-MurNAc to reactions with whole cell extracts decreased the amount of labeled product by a factor of ten, However, the same result was not seen when a 10-fold excess of unlabeled UDP-GlcNAc was added to the reactions ( Table I).
Identification of the Bacterial UDP-MurNAc Hydroxylase Genes-Only two carbohydrate N-glycolylation reactions have been demonstrated in biology: the N-glycolylation of CMP-Nacetylneuraminic acid in many eukaryotic organisms and the N-glycolylation of UDP-N-acetylmuramic acid in a limited number of bacteria. We used the protein sequence of the mouse CMP-N-acetylneuraminic acid hydroxylase in a BLAST (23) search (www.ncbi.nlm.nih.gov/BLAST/) for a homolog in the sequenced and annotated Mycobacterium tuberculosis H37Rv genome (24). The best match that we obtained was open reading frame Rv3818. Additional BLAST searches revealed the presence of Rv3818 homologs within the genomes of all the available mycobacterial species (www.ncbi.nlm.nih.gov/sutils/ genom_table.cgi) including M. smegmatis, Mycobacterium avium, and Mycobacterium leprae, in which the gene is designated an pseudogene in the annotated genome, (25) as well as homologs in the actinobacteria Nocardia farcinica and Rhodococcus sp. RHA1, both of which possess peptidoglycan containing N-glycolylmuramic acid. A comparison of these proteins is shown in Table II. Furthermore, Rv3818 homologs were not present within the genomes of the related actinobacteria Corynebacterium glutamicum and Streptomyces albus that do not produce N-glycolylmuramic acid (9).
We analyzed the Rv3818 homolog in M. smegmatis in greater detail for consideration as the mycobacterial UDP-MurNAc hydroxylase. Shown in Fig. 2, panel A is a ClustalW (26) alignment of the M. smegmatis UDP-MurNAc hydroxylase protein sequence with several eukaryotic CMP-N-acetylneuraminic acid hydroxylases and the homologs present in the genomes of N. farcinica and Rhodococcus sp. RHA1. To aid in sequence alignment only residues present within the protein cores were used in the ClustalW analysis. We omitted the N and C termini of the eukaryotic hydroxylases and C termini of the bacterial enzymes for the sake of clarity (see below). We found that the prokaryotic proteins are only 12% identical to the eukaryotic hydroxylases, although when strongly conserved residues are considered the similarity increases to 28%.
A conserved iron binding domain, presumably required for the metal-dependent acquisition of oxygen (13,27), is present in both the eukaryotic enzymes and the bacterial protein sequences ( Fig. 2A). Also, a Rieske iron-sulfur domain found in the N termini of the eukaryotic enzymes and involved with electron transfer (13,27), is located within the C termini of the bacterial proteins. The differential location of these [Fe-S] domains among the various proteins prompted us to remove the N and C termini of the proteins in the alignment in Fig. 2A  The namH Gene Is Not Essential to M. smegmatis-We tested the essentiality of the namH gene in M. smegmatis by allelic exchange. We constructed a sacB-based suicide plasmid, pMP354, bearing an unmarked, 500-bp deletion allele (⌬namH1). The deletion of the gene in wild-type M. smegmatis strain PM965 was done by a previously described counterselection protocol ("Experimental Procedures"). Southern blot analysis confirmed the exchange of the ⌬namH1 allele with that of wild type in the M. smegmatis genome of strain PM979 (data not shown).
The M. smegmatis ⌬namH1 Mutant Is Devoid of UDP-Mur-NAc Hydroxylase Activity-No UDP-MurNAc hydroxylase activity was observed in reactions containing 2 mg of protein of whole cell extracts prepared from the ⌬namH1 mutant (Table  I). Complementation of the mutant with a plasmid bearing the wild-type gene restored UDP-MurNAc hydroxylase activity to levels greater than that seen within the parental strain (Table  I). The increased hydroxylase activity of the complemented mutant compared with the parental strain can be attributed to the use of a multicopy plasmid containing a promoter that may be stronger than the native namH promoter.
The M. smegmatis ⌬namH1 Mutant Lacks N-Glycolylated Peptidoglycan Precursors-We isolated and analyzed the peptidoglycan precursor muropentapeptides from the wild-type parental strain PM965 and the ⌬namH1 mutant PM979 by LC-MS to determine the aclylation state of the muramyl moieties. The positive ion mass spectra of the muropeptides from PM965 (namH ϩ ) were dominated by ions having m/z values of 824.2 and 808.2 in an apparent ratio of 7:3 (Fig. 3).   removal of the peptides from the muramic acid with mutanolysin treatment and analysis of the muramic acid as described under "Experimental Procedures" (data not shown). The M. smegmatis ⌬namH1 Mutant Is Hypersusceptible to ␤-Lactam Antibiotics and Lysozyme-Although namH was determined to be a non-essential gene for M. smegmatis, we hypothesized that the lack of N-glycolylated muramic acid in the peptidoglycan would have a detrimental effect on cell wall biosynthesis or stability. This hypothesis was evaluated by testing susceptibility of the ⌬namH1 mutant to various antibiotics that interfere with cell envelope biosynthesis and to lysozyme.
We tested the sensitivity of the ⌬namH1 mutant to the ␤-lactam antibiotics amoxicillin and ampicillin, drugs that interfere with peptidoglycan assembly and cross-linking, by a disk diffusion assay. As shown in Table III, the mutant was hypersusceptible to ␤-lactams compared with the parental and namH ϩ complemented strains regardless of the media (rich or minimal) used for culture. However, the susceptibility of the mutant to amoxicillin (and to a lesser extent, ampicillin) was exacerbated when grown on rich medium (LB) as compared with minimal medium (7H9) ( Table III). We observed no difference in susceptibility between the ⌬namH1 mutant and the parental and complemented strains to the antimycobacterial drugs ethambutol and isoniazid that target the biosynthesis of mycolic acids and arabinan, two other components of the mycobacterial cell envelope (Table III).
In this work, we demonstrated hypersusceptibility of the ⌬namH1 mutant to penicillin type ␤-lactam antibiotics. In addition, a namH mutant of M. smegmatis is also hypersusceptible to the cephalosporin type ␤-lactam ceftriaxone. In a separate project, our laboratory isolated a namH::mariner transposon insertion mutant of M. smegmatis from a transposon mutagenesis search for mutants hypersusceptible to cephalosporins (28). The phenotype of this mutant is due to the transposon insertion in namH as a plasmid bearing a copy of the wild-type namH gene confers parental ceftriaxone susceptibility to the mutant (data not shown).

FIG. 2. Comparison of the M. smegmatis UDP-MurNAc hydroxylase with eukaryotic CMP-NeuNAc hydroxylases.
In panel A the sequences of several eukaryotic CMP-NeuNAc hydroxylases and actinobacterial UDP-MurNAc hydroxylases were used in a ClustalW alignment with the M. smegmatis UDP-MurNAc hydroxylase. The N and C termini of the eukaryotic sequences are not shown as well as the C termini of the bacterial sequences to facilitate alignment of the core hydroxylase sequences. Identical and strongly conserved residues are in bold (asterisks represent identical amino acids). Underlined sequences indicate the mononuclear iron binding site. Panel B, the C terminus of the M. smegmatis UDP-MurNAc hydroxylase is shown as an example representing the C termini of the prokaryotic proteins containing the Rieske iron-sulfur center. Underlined residues are required for coordinating the Fe-S cluster. Abbreviations and references: N. farcinica, nocardia.nih.go.jp/blast; R. sp. RHA1, Rhodococcus sp. RHA1 (www.bcgsc.ca/gc/rhodococcus); M. smegmatis, contig 3269, www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?id ϭ 246196); starfish, Asterias rubens (GenBank TM accession no. AJ308602); mouse, Mus musculus (GenBank TM accession no. AB061276); chimpanzee, Pan troglodytes (GenBank TM accession no. AF074481). Sequences were analyzed by BLAST (23) and CDART (35), at NCBI (www.ncbi.nlm.nih.gov/). A characteristic of organisms belonging to the genus Mycobacterium is an inherent resistance to lysozyme, a muramidase that cleaves the glycosidic bond between the sugar residues of the peptidoglycan chain. To see if the glycolylation state of the peptidoglycan affects lysozyme susceptibility, we determined the lysozyme MIC of the ⌬namH1 mutant grown in both minimal and rich media. As shown in Table IV, both the wild-type and complemented strains were able to tolerate lysozyme challenge to a greater extent than the mutant strain, although all strains were more susceptible growing in minimal medium compared with rich. The fold difference in lysozyme MIC between the various strains tested was never greater than 2-fold when grown in minimal medium. However, the difference in lysozyme MIC between the M. smegmatis ⌬namH1 mutant and controls strains was at least 8-fold when grown in rich medium (Table IV).
To further characterize the lysozyme sensitivity of the ⌬namH1 mutant, we studied the effect of lysozyme added to   mid-exponential phase cultures growing in minimal medium. The optical density of the cultures was recorded over an 8-h period following challenge with 20 g/ml lysozyme (Fig. 4). Consistent with the lysozyme MIC results, growth of the mutant was hindered following the addition of lysozyme to the media. The optical density of the wild type and the complemented cultures plateaued at 0.5 h after addition of lysozyme. This plateau lasted for 2 h, after which the density of the cultures increased. In contrast, the density of the ⌬namH1 culture decreased 1 h after the addition of lysozyme and the cells appeared to begin to aggregate. The culture eventually recovered 3.5 h later, but the density of the ⌬namH1 cultures never matched that of the controls. The viable counts of the ⌬namH1 cultures did not decrease throughout the growth curve following lysozyme addition (data not shown). The recovery and continuation of culture density during later time points of the growth curve can be attributed to the lability of the enzyme, as addition of fresh lysozyme at these later time points affected the growth of the mutant (data not shown).

DISCUSSION
The occurrence of N-glycolylated muramic acid in peptidoglycan is very rare and the purpose of this modification is unclear. The presence of this sugar in only closely related members of the class Actinomycetales suggests a correlation between the presence of N-glycolylmuramic acid and the type of cell envelope in the organism. All of the N-glycolylmuramic acid-containing bacteria also have a mycolylarabinogalactan-peptidoglycan type of cell envelope. In this study we have identified the namH gene encoding the enzyme responsible for the production N-glycolylated muramic acid in the mycobacterial peptodoglycan. This gene is found in all mycobacterial genomes sequenced to date, but it is a pseudogene in M. leprae. This was initially surprising, as the peptidoglycan of M. leprae has been reported to contain N-glycolylmuramic acid (29). A more recent study demonstrated that the peptidoglycan of M. leprae is not N-glycolyated (30), consistent with the identity of NamH as the enzyme responsible for this modification of the mycobacterial peptidoglycan. Given that the eukaryotic sialic acid hydroxylases require a cytochrome and cytochrome reductase for activity (11, 47, 48), there is the slight possibility that the namH gene encodes a protein that indirectly interacts with the peptidoglycan hydroxylase system in mycobacteria. To address this issue, we have made several attempts to express and purify NamH fusion proteins bearing N-or C-terminal polyhistidine or thioredoxin tags and then assay the purified protein with or without reconstitution using cell extracts. Unfortunately, the NamH protein does not appear to tolerate fusions, as all our fusion constructs failed to complement the namH deletion mutant. Likewise, attempts to overexpress the native protein in E. coli were unfruitful, since the protein is rather toxic to E. coli resulting in low protein expression or the generation of mutant namH genes (data not shown).
We show that namH is not essential for the growth of M. smegmatis, however, a namH deletion mutant is hypersusceptible to ␤-lactam antibiotics that interfere with peptidoglycan cross-linking and lysozyme. These two phenotypes could suggest that the N-glycolyl group does indeed serve to stabilize the cell wall via hydrogen bonds as previously suggested, and the loss of this group decreases the integrity of the wall such that ␤-lactam antibiotics and lysozyme have a greater ability to affect the peptidoglycan. However, there is very little information about the three-dimensional organization of this part of the mycobacterial cell wall. The rhamnose-N-acetylglucosamine linker that connects the arabinan polysaccharide to the peptidoglycan is attached to the C6 of muramic acid via a phosphoryl group (3). It may be that the phosphoryl groups are involved in hydrogen bonding to the N-glycolylmuramic acid on adjacent strands, but not all of the muramic acid groups are substituted with the linker. If there were a broad defect in cell envelope stability in the PG of the ⌬namH1 mutant, it would be expected to have a generalized effect upon other aspects of the cell envelope, presumably resulting in hypersusceptibility to other antibiotics (i.e. isoniazid, ethambutol) that target the biosynthesis of other cell envelope components. However, the mutant has wild-type susceptibility to these antibiotics.
We favor an alternative hypothesis that the two phenotypes of the ⌬namH1 mutant are the result of separate mechanisms. The increased susceptibility to ␤-lactams is due to problems with either the translocation of the peptidoglycan precursors across the cytoplasmic membrane, the polymerization of the glycan backbone, or the recognition of the precursors by the peptidases responsible for cross-linking of the peptidoglycan peptides. We propose that the recognition of the abnormal sugar by the transglycosylases is affected. This could influence peptide cross-linking, as transpeptidation reactions are generally linked to the polymerization of the glycan chain. The net result of this would be a decrease in the amount of peptidoglycan cross-linking and thus, hypersusceptibility to ␤-lactam antibiotics. The second part of this hypothesis posits that the increased susceptibility of the ⌬namH1 mutant to lysozyme is because the specific function of the N-glycolyl group is to block the enzyme from the ␤-1,4 bond between the muramyl and the N-acetylglucosaminyl residues in the glycan chain. It is known that O-acetylation at C6 of the muramic acid ring confers lysozyme resistance (31), as does the de-acetylation of N-acetylglucosamine (32). In this model, the replacement of the Nglycolyl group with N-acetyl on muramic acid in the ⌬namH1 mutant would result in a peptidoglycan with increased sensitivity to lysozyme.
The large differences in lysozyme sensitivity observed between the ⌬namH1 mutant and the parental strain during growth in rich versus minimal medium are interesting and ought to be explored in greater detail. All the strains were consistently more sensitive to lysozyme when grown in minimal medium. However, the fold difference in sensitivity between the parental strain and ⌬namH1 mutant were significantly more pronounced when grown in rich medium. Likewise, the mutant was more susceptible to the ␤-lactam antibiotic amoxicillin when grown in rich medium versus minimal medium. Whether or not these results are the consequence of metabolic changes under different growth conditions that affect the structure and permeability of the cell envelope remains to be determined.
Another question is whether this unique modification has a role in mycobacterial pathogenesis. Mammals possess a series of antibacterial enzymes (lysozymes, amidases) capable of cleaving bacterial peptidoglycan. Any modification of the peptidoglycan that confers resistance to these types of enzyme could have a role in pathogenesis. For Streptococcus pneumoniae, resistance to lysozyme is a virulence trait. This organism possesses a peptidoglycan de-acetylase that removes the N-acetyl group from N-acetylglucosamine, rendering the peptidoglycan lysozyme resistant (32). A S. pneumoniae mutant unable to de-acetylate its peptidoglycan is sensitive to lysozyme and is attenuated in a mouse model (33). The N-glycolylation of the mycobacterial PG could play a similar role in M. tuberculosis virulence.
Peptidoglycan fragments are pathogen-associated molecular patterns recognized by the innate immune system (8). It is possible that N-glycolylated muramic acid could function to modulate the innate immune response to mycobacterial peptidoglycan fragments released during a mycobacterial infection. Recent studies demonstrated that the peptidoglycan of M. tuberculosis is 40 times more effective than E. coli peptidoglycan at inhibiting the interferon-␥ induction of MHC Class II expression on macrophages (34). Given that the structures of the peptidoglycan of these two bacteria differ primarily on the basis of the acylation of the muramic acid residues, we hypothesize that the presence of the N-glycolyl group on muramic acid in the mycobacterial peptidoglycan could explain the potency of M. tuberculosis peptidoglycan in effecting macrophage function.