Characterization of the Essential Gene glmM Encoding Phosphoglucosamine Mutase in Escherichia coli *

Two different approaches to identify the gene encod- ing the phosphoglucosamine mutase in Escherichia coli were used: (i) the purification to near homogeneity of this enzyme from a wild type strain and the determina-tion of its N-terminal amino acid sequence; (ii) the search in data bases of an E. coli protein of unknown function showing sequence similarities with other hex- osephosphate mutase activities. Both investigations revealed the same open reading frame named yhbF located within the leuU-dacB region at 69.5 min on the chromosome (Dallas, W. S., Dev, I. K., and Ray, P. H. (1993) J. Bacteriol. 175, 7743–7744). The predicted 445-residue protein with a calculated mass of 47.5 kDa contained in particular a short region GIVISASHNP with high similarity to the putative active site of hexosephosphate mu- tases. In vitro assays showed that the overexpression of this gene in E. coli cells led to a significant overproduc- tion (from 15- to 50-fold) of phosphoglucosamine mutase activity. A hexose 1,6-diphosphate-dependent phospho- rylation of the enzyme, which probably involves the serine residue at position 102, is apparently required for its catalytic action. As expected, the inactivation of this gene, which is essential for bacterial growth, led to the progressive depletion of the pools of precursors located downstream from glucosamine 1-phosphate in the path- way for peptidoglycan synthesis

In vitro assays showed that the overexpression of this gene in E. coli cells led to a significant overproduction (from 15-to 50-fold) of phosphoglucosamine mutase activity. A hexose 1,6-diphosphate-dependent phosphorylation of the enzyme, which probably involves the serine residue at position 102, is apparently required for its catalytic action. As expected, the inactivation of this gene, which is essential for bacterial growth, led to the progressive depletion of the pools of precursors located downstream from glucosamine 1-phosphate in the pathway for peptidoglycan synthesis. This was followed by various alterations of cell shape and finally cells were lysed when their peptidoglycan content decreased to a critical value corresponding to about 60% of its normal level. The gene for this enzyme, which is essential for peptidoglycan and lipopolysaccharide biosyntheses, has been designated glmM.
UDP-GlcNAc 1 is one of the main cytoplasmic precursors of bacterial cell-wall peptidoglycan (1,2). In Escherichia coli and other Gram-negative bacteria, it is also the precursor for outer membrane lipopolysaccharide, as well as for the synthesis of the enterobacterial common antigen (3,4). The genes and enzymes involved in the steps located downstream from this branchpoint (shown schematically in Fig. 1) in the different pathways have in most cases been characterized and studied in detail (3)(4)(5)(6)(7)(8)(9)(10). In comparison, the metabolic route leading to the formation of UDP-GlcNAc has been poorly investigated. Considering that it was a potential site for the regulation of the flow of metabolites going through the peptidoglycan and lipopolysaccharide pathways, the latter reaction sequence has recently been investigated in more detail (11,12). Four successive steps are required for the synthesis of UDP-GlcNAc from fructose 6-phosphate (13,14) (Fig. 1). The first of these reactions is catalyzed by the L-glutamine:D-fructose-6-phosphate amidotransferase (also named glucosamine-6-P synthase) (15). Mutants altered in this activity are characterized by an auxotrophy for GlcN or GlcNAc (16,17), and the corresponding glmS gene has been located at 84 min on the E. coli map (15,18). The subsequent steps from GlcN-6-P to UDP-GlcNAc are via GlcN-1-P (Fig. 1). We have previously shown that the gene of unknown function preceding glmS on the E. coli chromosome codes in fact for the GlcNAc-1-P uridyltransferase, and the gene for this final step leading to UDP-GlcNAc ( Fig. 1) was named glmU (11). More recently, we demonstrated that the glmU gene product also catalyzes the preceding step of acetylation of GlcN-1-P and thus appears as a bifunctional enzyme catalyzing two subsequent steps in the same pathway (12) (Fig.  1). Here we describe the partial purification and some properties of the phosphoglucosamine mutase which catalyzes the interconversion of GlcN-6-P and GlcN-1-P isomers, as well as the identification of the corresponding glmM gene on the chromosome of E. coli.

MATERIALS AND METHODS
Bacterial Strains and Plasmids-E. coli strain JM83 (ara ⌬ [lac-proAB] rpsL thi ⌽80 dlacZ ⌬ M15) (19) was used as the host strain for plasmids and for the purification of the glmM gene product. Strain W1485 pgm⌬::tet and plasmid pML14 carrying the E. coli pgm (phosphoglucomutase) gene (20) were kindly provided by S. Slater. pJP900, a pUC9-derived plasmid carrying the 2.4-kb XhoII fragment of bacteriophage with the p R promoter inserted into its BamHI site, was obtained from J. Plá (CSIC, Universidad Autonoma de Madrid, Spain). Cloning vector pUC18 and the Kan r cartridge originating from the pUC4K plasmid were purchased from Pharmacia Biotech Inc. Plasmid pMAK705 bearing a thermosensitive replicon was obtained from S. R. Kushner (21).
Growth Conditions-Unless otherwise noted, 2YT was used as a rich medium for growing cells (22). Growth was monitored at 600 nm with a spectrophotometer (model 240, Gilford Instrument Laboratories, Inc., Oberlin, Ohio). For strains carrying drug resistance genes, antibiotics were used at the following concentrations (in g⅐ml Ϫ1 ): ampicillin (100), kanamycin (30), and chloramphenicol (25).
General DNA Techniques and E. coli Cell Transformation-Smalland large-scale plasmid isolations were carried out by the alkaline lysis method (23). Standard procedures for endonuclease digestions, ligation, filling-in of 5Ј-protruding ends by using the Klenow fragment of DNA polymerase I, and agarose electrophoresis were used (23,24). E. coli cells were made competent for transformation with plasmid DNA by the method of Dagert and Ehrlich (25).
Construction of Plasmids-The 17-kb chromosomal fragment from E. coli which is carried by phage 14F11 (clone 521) from the bank of Kohara et al. (26) was used as the starting material for the construction of the different plasmids described in Fig. 2. The pMLD90 plasmid was constructed by inserting the 3.1-kb BamHI fragment carrying the glmM gene (yhbF) into the BamHI site of the pUC18 vector (with glmM in the opposite orientation as compared to the lac promoter from the vector). The pMLD96 plasmid was constructed by inserting the 2.1-kb PstI fragment from pMLD90 into the corresponding site of the pUC18 vector (with glmM and lac promoter in the same orientation). An internal AccI-ClaI deletion in pMLD96 produced the plasmid pMLD99 (Fig. 2). For the expression of glmM under the control of the p R promoter, plasmid pMLD100 was constructed by inserting the 2.4-kb XhoII fragment (carrying the structural gene cI857 encoding a thermosensitive form of the cI repressor, the strong p R promoter, and the ribosome-binding site of cro) (27) into the unique BamHI site of pMLD99. The disruption of the glmM gene was obtained by inserting the 1.28-kb HincII kan cartridge originating from pUC4K into the unique BstEII site of plasmid pMLD90 lying within the glmM gene coding sequence, generating plasmid pMLD94 (Fig. 2). The pGMMkan plasmid was constructed by inserting the 4.5-kb BamHI fragment from pMLD94 (harboring the disrupted glmM gene) into the BamHI site of pMAK705 (21). The pGMM plasmid was isolated as described below during the excision process that followed integration of the pGMMkan plasmid at the chromosomal glmM locus.
Disruption of the Chromosomal glmM Gene-The wild-type chromosomal copy of glmM was replaced by a disrupted one by following the procedure of Hamilton et al. (21) which uses pMAK705, a plasmid bearing a thermosensitive replicon. pGMMkan, a pMAK705 derivative carrying the disrupted glmM gene, was transformed into JM83. Integration of the plasmid into the chromosome was selected for by plating cells at 44°C on 2YT-chloramphenicol. Several clones were picked up, and the integration of pGMMkan at the glmM locus was verified by Southern analysis. Excision of the plasmid from the chromosome was then achieved as follows. Cointegrants were grown at 30°C in 2YTchloramphenicol for at least 30 doubling times and subsequently plated at 30°C. Individual clones were then screened for sensitivity to chloramphenicol at 44°C, which is indicative of plasmid excision. The structures of the excised plasmids from several clones were determined by DNA restriction analysis. One clone containing a plasmid bearing the wild-type glmM gene (called pGMM) was chosen. Finally, the replacement in that strain (named GPM83) of the chromosomal glmM copy by the inactivated one was verified by Southern analysis (data not shown).
Pool Levels of Peptidoglycan Precursors-Cells of GPM83 (1-liter cultures) were grown exponentially at 30°C in 2YT medium. At a cell density of 5 ϫ 10 4 ⅐ml Ϫ1 , the temperature of the culture was either maintained at 30°C or increased to 43°C. Incubation was continued until the optical density of the culture at 43°C reached a plateau value of 0.5-0.7 about 5 to 6 h later. At this time, cells were rapidly chilled to 0°C and harvested in the cold. Control cultures were made with strain JM83 which carries an intact chromosomal copy of glmM. The extraction of peptidoglycan nucleotide precursors as well as the analytical procedure used for their quantitation were as described previously (6,7).

Isolation of Sacculi and Quantitation of Peptidoglycan-Exponen-
tial-phase cells of JM83 and GPM83 were grown at 30°C or first at 30°C and then 43°C as described above. Harvested cells were washed with a cold 0.85% NaCl solution and centrifuged again. Bacteria were then rapidly suspended by vigorous stirring in 40 ml of a hot (95 to 100°C) aqueous 4% SDS solution for 30 min. After standing overnight at room temperature, the suspensions were centrifuged for 30 min at 200,000 ϫ g, and the pellets were washed several times with water. Final suspensions made in 5 ml of water were homogenized by brief sonication. Aliquots were hydrolyzed and analyzed as described previously, and the peptidoglycan content of the sacculi was expressed in terms of its DAP content (6,28).
Preparation of Crude Enzyme-Cells of JM83 harboring the different plasmids described above (excepting pMLD100) were grown exponentially at 37°C in 2YT-ampicillin medium (1-liter cultures). Cells were harvested in the cold when the optical density of the cultures reached 0.7. A different protocol was used with strain JM83(pMLD100): cells were grown first at 30°C and at an absorbance of 0.01 (5 ϫ 10 6 cells⅐ml Ϫ1 ), cultures were shifted to 42°C for 3 h during which time the absorbance reached a value of 0.7. Strains JM83 and GPM83 were grown at 30°C or 43°C as described above for the extraction of peptidoglycan precursors. In all cases, harvested cells were washed with 40 ml of cold 0.02 M potassium phosphate buffer (pH 7) containing 1 mM ␤-mercaptoethanol. The wet cell pellet was suspended in 9 ml of the same buffer and disrupted by sonication (Sonicator 150; T. S. Ultrasons, Annemasse, France) for 10 min with cooling. The resulting suspension was centrifuged at 4°C for 30 min at 200,000 ϫ g. The supernatant was dialyzed overnight at 4°C against 100 volumes of the same phosphate buffer, and the resulting solution (10 ml, 10 to 12 mg of protein⅐ml Ϫ1 ) designated as crude enzyme was stored at Ϫ20°C. SDS-PAGE analysis of proteins was performed as described previously using 13% polyacrylamide gels (29). Protein concentrations were determined by the method of Lowry, using bovine serum albumin as standard (30).
Assay for Phosphoglucosamine Mutase Activity-A coupled assay was routinely used in which the GlcN-1-P synthesized from GlcN-6-P by the mutase was quantitatively converted to UDP-GlcNAc in the presence of purified bifunctional GlmU enzyme (12). The standard assay mixture contained 50 mM Tris-HCl buffer (pH 8.0), 3 mM MgCl 2 , 1 mM GlcN-6-P, 0.4 mM [ 14 C]acetyl-CoA (700 Bq), 10 mM UTP, 0.7 mM Glc-1,6-diP, pure GlmU enzyme (1 g), and enzyme (0.1 to 10 g of protein, depending on overexpression or purification factors) in a final volume of 100 l. Mixtures were incubated at 37°C for 30 min, and reactions were terminated by the addition of 10 l of acetic acid. Reaction products were separated by high-voltage electrophoresis on Whatman No. 3MM filter paper in 2% formic acid (pH 1.9) for 90 min at 40 V/cm using an LT36 apparatus (Savant Instruments, Hicksville, NY). The only two radioactive spots (acetyl-CoA and UDP-GlcNAc) were located by overnight autoradiography using type R2 films (3 M, St. Paul, MN) or with a radioactivity scanner (model Multi-Tracermaster LB285, Berthold France, Elancourt, France). They were cut out and counted in an Intertechnique SL 30 liquid scintillation spectrophotometer with a solvent system consisting of 2 ml of water and 13 ml of Aqualyte mixture (J. T. Baker Chemicals, Deventer, Netherlands). 1 unit of enzyme activity was defined as the amount which catalyzed the synthesis of 1 mol of GlcN-1-P in 1 min.
Purification of GlmM from Strain JM83-A large culture (3 ϫ 18 liters) of strain JM83 was performed at 37°C in a fermenter. Cells were harvested at the end of the exponential phase and disrupted by sonication as described above. The resulting soluble protein extract (5 g of protein, 240 units of phosphoglucosamine mutase activity, 198 ml) was loaded onto a column (5 ϫ 25 cm) of DEAE-Trisacryl-M (BioSepra, Villeneuve-la-Garenne, France) pre-equilibrated with buffer A (20 mM potassium phosphate buffer, pH 7.4, containing 5 mM MgCl 2 , 2.5 mM ␤-mercaptoethanol, and 5% (v/v) glycerol). The elution was run at a flow rate of 2 ml⅐min Ϫ1 , first with 120 ml of buffer A, and then with a linear gradient (1 liter) of NaCl (0 to 160 mM) in buffer A. Fractions (12 ml) were collected and assayed for phosphoglucosamine mutase activity. The most active fractions which eluted in 70 mM NaCl were pooled (96 units, 420 mg of protein) and concentrated to 30 ml by ultrafiltration through PM10 membranes (Amicon). The buffer was exchanged with buffer B (5 mM potassium phosphate buffer, pH 7.4, containing the same additives as buffer A) by repeated concentration and dilution using PM10 membranes. The protein solution was loaded onto a column (5 ϫ 17 cm) of hydroxylapatite-Ultrogel (BioSepra) pre-equilibrated with buffer B. The elution was run at a flow rate of 2 ml⅐min Ϫ1 , first with 150 ml of buffer B, and then with a linear gradient (800 ml) of potassium phosphate (5 mM to 200 mM). 10-ml fractions were collected and assayed for phosphoglucosamine mutase activity. The most active fractions which eluted in 150 mM potassium phosphate were pooled (50 units, 41 mg of protein) and concentrated as above to about 5 ml. This solution was loaded onto a column (1 ϫ 50 cm) of Ultrogel-AcA 44 (BioSepra) equilibrated with buffer C (as buffer A but 100 mM in potassium phosphate), and the elution was run at a flow rate of 0.1 ml⅐min Ϫ1 . Active fractions (48 units, 32 mg of protein) were pooled, concentrated as above to 5 ml, and dialyzed against buffer A. This solution was loaded onto a column (2.5 ϫ 13 cm) of carboxymethylcellulose CM32 equilibrated with buffer D (buffer A at pH 6.15). The elution was run at a flow rate of 1 ml⅐min Ϫ1 , first with 60 ml buffer D, and then with a linear gradient (400 ml) of NaCl (0 to 200 mM) in buffer D. Under these conditions, the phosphoglucosamine mutase eluted in 10 mM NaCl, slightly later than the remaining contaminant proteins. Most active fractions were pooled (20 units, 7.7 mg of protein) and dialyzed against buffer A.
Amino Acid Composition and Peptide Sequencing-Quantitative amino acid analyses were performed with an analyzer (model LC 2000; Biotronik, Frankfurt, FRG) after hydrolysis of samples in 6 M HCl at 95°C for 16 h.
For NH 2 -terminal amino acid analysis, 25 g of partially purified phosphoglucosamine mutase was resolved by SDS-PAGE and electroblotted to an Immobilon-P membrane (Millipore). The membrane was stained with Ponceau Red, and the major protein band (corresponding to the enzyme) was excised. Microsequencing was performed with a model 473A protein sequencer (Applied Biosystems) under standard conditions involving the detection of phenylthiohydantoin-derivatives.

RESULTS
Purification and Amino Acid Sequencing of Phosphoglucosamine Mutase-Starting from a 54-liter culture of wild-type strain JM83 which yielded 5 g of total soluble proteins, the phosphoglucosamine mutase activity was purified approximately 150-fold to a final specific activity of 7.2 mol⅐min Ϫ1 ⅐mg Ϫ1 (Table I). The purest fractions eluted from the last column contained a major protein species migrating after denaturating polyacrylamide gel electrophoresis as a protein of 45-50 kDa (Fig. 3). It was the only one coeluting perfectly in all purification steps with the phosphoglucosamine mutase activity. The N-terminal amino acid sequence of this protein species was thus determined after electroblotting on Immobilon-P and appeared as follows: SNRKYFGTXXI.
Search for the glmM Gene Using Data Bases-The alignment of this sequence with those of proteins from the Swissprot and Genpro libraries revealed that it matched perfectly with the N-terminal sequence MSNRKYFGTDGI predicted for the product of an E. coli open reading frame of unknown function named yhbF (31) or mrsA (GenBank accession number U01376) which was located near folP at 69.5 min on the chromosome. Several discrepancies were observed when comparing the two reported sequences of this gene. Using plasmid pMLD96 (Fig. 2) as the template and appropriate synthetic oligonucleotides, the complete 1335-bp coding sequence was thus verified again and appeared similar to that published first (yhbF). Taking into account the post-translational removal of the N-terminal methionine (revealed by protein sequencing), the mature product of this gene is theoretically a 444-amino acid protein with a molecular mass of 47,380.
In fact, this gene was identified at the same time by searching in data bases for a protein exhibiting sequence similarities with other known hexosephosphate mutases. The sequence of the phosphomannomutase from E. coli (the cpsG gene product) was used in the search which yielded all phosphoglucomutase and phosphomannomutase sequences characterized to date, as well as the YhbF putative peptide (26% identity in 433 overlaps). The alignment of YhbF with the same libraries gave as best scores two other proteins of unknown function containing 445 and 463 amino acids, respectively: UreC from Helicobacter pylori (44.1% identity on 425 residues) (32) and a putative protein from Mycobacterium leprae (Genpro accession number U00020.PE23, 44% identity on 440 residues). All these sequences contain in particular a short region previously characterized as the putative active site of hexosephosphate mutases. This motif which is described as (GA)(LIVM)X (LIVM)(ST)(PGA)S*HXPX 4 (GN) in the PROSITE data base (PS00710) appeared as GIVISAS*HNPFYDNG in YhbF, where S* represents the active-site serine residue. It was generally assumed that the serine residue was phosphorylated during the catalytic action of the hexosephosphate mutases, a reaction requiring the presence of the corresponding hexose 1,6-diphosphate (33)(34)(35). This serine residue was located at position 102 in the sequence of YhbF. All these data clearly showed that the a The purest fraction eluted from this column of CM32 and used for electroblotting and protein sequencing contained 0.5 mg of protein with a specific activity of 7.2 units ⅐ mg Ϫ1 , corresponding to a purification factor of 150.
Levels of Phosphoglucosamine Mutase in E. coli Cells-The specific activity of phosphoglucosamine mutase in strains overexpressing the yhbF gene was estimated. A significant overproduction (up to 18-fold) in cells carrying plasmids with the yhbF gene expressed under the control of the lac promoter was observed (Table II). As shown in Fig. 4, this was accompanied by the accumulation in the cell content of a protein species of about 47 kDa, a value in agreement with that (47,380) calculated from the DNA sequence (31). A more efficient overproducing plasmid pMLD100 with the yhbF gene expressed under the control of the strong p R promoter was constructed. When cells of JM83(pMLD100) were grown exponentially at 30°C and then shifted to 42°C, an initial increase in the growth rate was observed, but growth progressively slowed down and the culture finally reached a plateau value. As soon as 1 h after the temperature shift, a large accumulation of the 47K protein was observed, which further increased to finally account for more than 20 -30% of the cell proteins after a 3-h incubation period (Fig. 4). It comigrated with the protein overproduced from the other plasmid construction as well as with the phosphoglucosamine mutase purified from wild-type strain JM83 (Fig. 4). A typical fractionation procedure of cell extracts showed that the highly overproduced protein was mainly found in the soluble fraction, but that significant amounts (10 to 20%) remained associated with the particulate fraction. This finding was certainly due to the formation of aggregates at this high level of protein expression, as confirmed by the presence of multiple inclusion bodies in induced cells when observed by optical microscopy. Finally, we showed that after 3 h of growth at 42°C, JM83(pMLD100) cells contained about 50-fold more phosphoglucosamine mutase activity than cells carrying the pUC18 control vector (Table II). All these different results taken together supported the initial proposal that the yhbF gene coded for the phosphoglucosamine mutase. This gene was thus named glmM (for glucosamine mutase) following the nomenclature previously adopted for the glmS and glmU genes from the same pathway (11,17).
Hexose 1,6-Diphosphate-dependent Phosphorylation of Phosphoglucosamine Mutase-By similarity with the reaction mechanism of other well characterized hexose-phosphate mu-tases (33)(34)(35)(36)(37), the phosphoglucosamine mutase was assumed to be active only in a phosphorylated form. As explained before, the serine residue at position 102 was the putative phosphorylated site and GlcN-1,6-diP was expected to be the phosphorylating agent. Since the latter compound was not commercially available, we used instead Glc-1,6-diP in our enzymatic assays, as generally was done with hexosephosphate mutases other than the phosphoglucomutase species (35,37). The apparent GlmM activity that could be detected in the extract from wild-type strain JM83 was very low when Glc-1,6-diP was omitted from the reaction mixture (Table III). This basal level was greatly enhanced (up to 20-fold) by increasing the concentration of this compound (Table III), according to a hyperbolic saturation curve from which a constant of 150 M could be calculated (data not shown). Interestingly, the overproduction of the enzyme activity was no longer detected in crude extracts from strains JM83(pMLD96) and JM83(pMLD100) when assays were performed in the absence of Glc-1,6-diP. Under these conditions, the apparent activity of GlmM was similar or curiously lower in these extracts than that detected in the wildtype extract (Table III). The addition of a saturating concentration of Glc-1,6-diP was accompanied by a very large increase (500-or 2000-fold) of the activity, which was consistent with the relative amounts of GlmM known to be present in these various extracts (see above). These different results first indicated that only a small part of the enzyme extracted from the cell content was present in the active phosphorylated form (5%, as judged from results obtained with the extract of strain JM83(pUC18)). Furthermore, the total amount of phosphorylated enzyme was clearly not increased in strains overproducing this protein to high levels, suggesting that the level of enzyme phosphorylation could be tightly regulated.
Inactivation of the Chromosomal glmM Gene and Its Effect on Bacterial Growth-To inactivate the glmM gene on the E. coli chromosome, we used the procedure described by Hamilton et al. (21) that is particularly well adapted for the disruption of essential genes, as recently shown for the construction of mutants altered in the glmU and murI genes which are essential for peptidoglycan synthesis (11,38). First, the glmM gene coding sequence carried by the pMLD90 plasmid was disrupted by inserting at the unique BstEII site the 1.28-kb kanamycin resistance gene from pUC4K, generating plasmid pMLD94 (Fig. 2). The complete insert from the pMLD94 plasmid was inserted into the pMAK705 vector which bears a thermosensitive replicon. The resulting plasmid, pGMMkan, was used to construct strain GPM83 (JM83 glmM::kan [pGMM]) having the inactivated glmM gene on the chromosome and the wildtype allele on the pMAK705 vector. At the restrictive temperature for plasmid replication (43 to 44°C), GPM83 failed to grow on 2YT plates, indicating that this strain with a disrupted chromosomal copy of the glmM gene was viable only in the presence of a plasmid carrying the wild-type gene. The failure to transduce by phage P1 the glmM::kan marker from GPM83 to other E. coli K12 strains further confirmed that glmM was an essential gene.
Since the pGMM plasmid bears a thermosensitive replicon, the effects of the specific inactivation of the glmM gene were observed by shifting exponentially growing cells of GPM83 and JM83 from 30°C to 43°C. Both strains showed an identical growth rate and cell morphology when grown at 30°C. However, after 5 to 6 h at 43°C, the growth rate of GPM83 rapidly slowed down and cells apparently entered a stationary phase at a lower cell mass (Fig. 5). In addition, GPM83 cells progressively changed from rods to greatly enlarged ovoids when observed by phase-contrast microscopy, whereas the morphology of the parental strain was unaltered (data not shown). Cells finally lysed after prolonged incubation at the restrictive temperature, as judged by a progressive decrease of turbidity of the culture (Fig. 5) and the presence of many ghosts within the cell population. This was consistent with the involvement of the glmM gene product in the biosynthesis of a cell-envelope component. The fact that these different effects were observed only after a few hours was explained by the time required for the progressive dilution or inactivation of the low-copy number pGMM plasmid and of the functional GlmM enzyme molecules present at the time of the temperature shift. As previously observed with a glmU mutant, these morphological changes were amplified when using a growth medium deprived of NaCl, and the precocious stationary phase which characterized GPM83 cells grown at 43°C was no longer observed when either 2% NaCl or 20% sucrose was added to the growth medium.
Biochemical Effects of the glmM Mutation-The effects of the mutation on bacterial growth were clearly associated with the depletion of the phosphoglucosamine mutase activity. GPM83 cells effectively contained 50 fold less phosphoglucosamine mutase activity than the parental strain after growth for 5 to 6 h at the restrictive temperature (Table II). As shown in Table IV, an arrest in the de novo synthesis of functional GlmM enzyme molecules rapidly led to the depletion of the pools of the two main nucleotide precursors UDP-GlcNAc and UDP-MurNAcpentapeptide (Fig. 1). The pools of the other intermediates from UDP-GlcNAc-enolpyruvate to UDP-MurNAc-tripeptide, which are always detected at a much lower intracellular concentration (28), were also depleted (data not shown). This finding confirmed that the mutational block was in one of the early steps located upstream from UDP-GlcNAc which are required for both peptidoglycan and lipopolysaccharide biosyntheses (Fig. 1). As a result, the synthesis of both cell-envelope components was inhibited and GPM83 cells finally were lysed when their peptidoglycan content was decreased to a critical value

TABLE III
Effect of glucose 1,6-diphosphate on phosphoglucosamine mutase activity Strains JM83(pUC18) and JM83(pMLD96) were grown at 37°C, and strain JM83(pMLD100) was grown first at 30°C and then for 3 h at 42°C as described under "Materials and Methods." The corresponding crude protein extracts were assayed for phosphoglucosamine mutase activity in the presence of varying concentrations of glucose 1,6-diphosphate.

Strain
Glucose 1,6diphosphate representing about 60% of its normal level (Table IV). The highly reduced peptidoglycan content determined in the mutant cells most probably represented the lowest physiological value compatible with cell integrity.
We failed to detect any significant accumulation of GlcN-6-P (the substrate of GlmM) in GPM83 cells grown at the restrictive temperature (data not shown). This unexpected finding probably resulted from its permanent and rapid conversion by the GlcN-6-P deaminase (the nagB gene product) back to fructose 6-phosphate ( Fig. 1) (14, 39).
Capability of Different Plasmids to Complement the glmM::kan Mutation-The different pUC18-derived plasmids which carry the glmM gene were tested for their capability to complement the thermosensitivity of strain GPM83. Interestingly, only those carrying inserts with glmM in the same orientation as the lac vector promoter (pMLD96, pMLD99) could restore the growth of GPM83 at 43°C. The pMLD90 plasmid carrying the BamHI fragment where glmM and most (800 bp) of the preceding folP gene were inserted in the orientation opposite to that of the lac promoter (Fig. 2) failed to complement the mutation. As described in Table II, no overproduction of phosphoglucosamine mutase could be detected in cells carrying the pMLD90 plasmid. These results implied that transcription of the chromosomal glmM gene may occur from a promoter located far upstream from its initiation codon, suggesting that glmM is cotranscribed with the proximal folP gene.
Owing to the fact that the phosphoglucomutase and phosphoglucosamine mutase activities catalyzed similar reactions and used substrates which only differ by the presence of the amino group at position 2 of the sugar, a plasmid pML14 (20) carrying the pgm gene from E. coli was also tested but it failed to restore growth of GPM83 cells at 43°C. Reciprocally, the pMLD96 plasmid was assayed for complementation of a strain deficient in phosphoglucomutase activity. W1485 pgm⌬::tet (20) appeared as pink colonies of good size on Mac Conkeygalactose plates, when transformed with this plasmid. As compared to the large red colonies obtained with the pML14 plasmid and to the very small white colonies obtained with the control vector pUC18, this result could be interpreted as a partial complementation. It suggested that GlmM could catalyze at least to some extent the interconversion of the glucosephosphate isomers. This side activity of phosphoglucosamine mutase was not investigated further.
As described above, two proteins of unknown function from Helicobacter pylori and Mycobacterium leprae showing more than 40% sequence identity with GlmM were found in data bases. It was tempting to speculate that these proteins were also phosphoglucosamine mutases. The observation that a plasmid carrying the ureC gene from H. pylori (pILL594 in Ref. 32) fully complemented the GPM83 mutant apparently confirmed this hypothesis. DISCUSSION Recently, we showed that both glucosamine-1-phosphate acetyltransferase and N-acetylglucosamine-1-phosphate uri-dyltransferase activities from E. coli were carried by the glmU gene product which thus acted as a bifunctional enzyme catalyzing two contiguous steps in this pathway (11,12). The actual characterization of a phosphoglucosamine mutase activity in crude extracts of E. coli and the demonstration that it is essential for growth is a final confirmation that the steps leading from GlcN-6-P to UDP-GlcNAc are via GlcN-1-P in bacteria (Fig. 1). This finding is consistent with the previous demonstration that exogenously supplied GlcNAc had to be deacetylated before it could be incorporated into cell walls (13). It suggests that any isomerase converting GlcNAc-6-P to GlcNAc-1-P has insignificant activity and that the only (or major) flux goes via GlcN-1-P which has to be reacetylated before formation of the nucleotide. Interestingly, a different reaction sequence is encountered in yeast where synthesis of GlcNAc-1-P from GlcN-6-P occurs by the successive actions of glucosamine-6-P acetylase and N-acetylglucosamine-phosphate mutase activities (37,40).
The glmM gene encoding phosphoglucosamine mutase was the last gene of the pathway for UDP-GlcNAc synthesis to be identified. Evidence is here provided that it corresponded to the previously sequenced open reading frame yhbF located at 69.5 min on the E. coli map. glmM is thus not linked to the related glmS and glmU genes previously identified at 84 min (11,18) and which are probably cotranscribed (39,41). The presence of other genes involved in peptidoglycan metabolism in the vicinity of glmM is noteworthy: in particular, the dacB and murZ genes encoding penicillin-binding protein 4 (42) and phosphoenolpyruvate:UDP-GlcNAc enoylpyruvyl transferase (43), respectively (Fig. 2). However, it is clear that these genes of related function do not belong to a cluster of tightly packed genes as observed for the mur genes in the 2-min region (8). The experiments of complementation described in this work and the lack of an obvious promoter sequence on the DNA upstream of glmM (31,44) suggested that this gene could be cotranscribed with the proximal folP gene. This is quite surprising when considering that the function of the folP gene product (dihydropteroate synthase involved in tetrahydrofolic acid synthesis) has no apparent relationship with cell-envelope metabolism (44). In fact, the same was observed in H. pylori where the gene for phosphoglucosamine mutase appears inserted within the urease operon, the reason for its initial designation as ureC (32). The transcription in this chromosomal region and a possible regulation of the glmM gene expression now has to be examined.
The effects of inactivating the glmM gene on the growth and cell morphology of E. coli were reminiscent of those observed with a glmU mutant altered in the next step from the same pathway (11,12). Cells progressively lost their rod shape to become greatly enlarged ovoids and growth stopped early at a lower cell density. Interestingly, this was not followed by an abrupt decrease of culture absorbance indicative of cell lysis, as generally observed with mutants defective in peptidoglycan synthesis (8,9,38,45). This was most probably due to the fact that the glmM mutation not only affects peptidoglycan but also IV Peptidoglycan content and pool levels of its main precursors in the parental and glmM mutant strains Cells were grown exponentially in 2YT medium at 30°C, or first at 30°C and then for 5 h at 43°C (the time at which the growth rate of the mutant strain began to decrease). Cells were harvested, and the peptidoglycan and its precursors were extracted and quantified as detailed under "Materials and Methods." The peptidoglycan content of sacculi was expressed in terms of its DAP content.

Strain
Growth conditions UDP-GlcNAc UDP-MurNAc-pentapeptide Peptidoglycan°C nmol/g bacteria (dry wt)   JM83  30  975  605  8600  JM83  43  740  700  9050  GPM83  30  850  700  9200  GPM83  43  45  95  5500 lipopolysaccharide synthesis. Effectively, thermosensitive mutants altered in the lipopolysaccharide pathway and in particular in the essential steps leading from UDP-GlcNAc to lipid A are characterized by an arrest of growth at the restrictive temperature (4,5). The effects observed here with the glmM mutation are probably those expected from a simultaneous depletion of both cell-envelope components.
In the present paper, the first characterization and purification to near-homogeneity of a bacterial phosphoglucosamine mutase is described. The final preparation had a specific activity of 7.2 units⅐mg of protein Ϫ1 whereas that of the crude extract from a wild-type strain was approximately 0.05 (Table  I). Assuming that there was no loss of activity in this crude extract from 4 ϫ 10 13 cells (5 g of protein) and that the purified enzyme contained only active GlmM molecules, a copy number of about 10,000 per cell could be estimated for this enzyme in a plasmid-free parental strain. This significant cellular abundance (0.5 to 1% of cell proteins) has certainly facilitated the successful purification in milligram quantities of this protein from a wild-type strain. It also explains why a 50-fold overproduction factor is enough to make it represent more than 20 -30% of total cell proteins.
The amino acid sequence of the phosphoglucosamine mutase contains the characteristic signature of hexosephosphate mutases. This motif includes the putative serine residue (S102) whose phosphorylation is a prerequisite for enzyme activity. By similarity with other mutase activities and in particular the well-characterized phosphoglucomutase species (33)(34)(35), the reaction catalyzed by GlmM is thought to proceed in two subsequent steps as follows: GlcN-1,6-diP which appears as an intermediate in the catalytic process could be also considered as the compound required for the initial activation (phosphorylation) of the enzyme. However, this is not yet clearly established as Glc-1,6-diP itself could efficiently phosphorylate GlmM. An extension of this work will be to characterize a putative enzyme involved in the specific synthesis of GlcN-1,6-diP in E. coli.
The fact that the phosphoglucosamine mutase is active only in a phosphorylated form is of great interest when considering the regulation of the flow of metabolites in this pathway. When assayed in the absence of added hexose-1,6-diphosphate, the apparent GlmM activity that could be measured after cell extraction theoretically reflects the total amount of phosphorylated enzyme present in E. coli growing cells. As described in this work, this basal activity in a wild-type strain was enhanced about 20-fold in the presence of saturating concentrations of Glc-1,6-diP, suggesting that most of the GlmM molecules were dephosphorylated and thus inactive in vivo. From the specific activity of 7.2 units⅐mg Ϫ1 determined for the purified enzyme in the presence of a saturating concentration of Glc-1,6-diP and assuming a molecular weight of 47,380, a turnover number of 340 min Ϫ1 could be calculated. Under the in vitro conditions used, the phosphoglucosamine mutase can catalyze the formation of approximately 10 8 molecules of GlcN-1-P in each cell during a 30-min generation time (10,000 ϫ 340 ϫ 30). Even if a 50% turnover of peptidoglycan material is taken into consideration (46), this value is much higher than that required for the formation of the average peptidoglycan content of exponentially growing cells previously estimated in the range from 3.5 to 5.5 ϫ 10 6 , depending on growth conditions (6,28). The requirements for GlcN-1-P (UDP-GlcNAc) molecules of the lipopolysaccharide pathway have not been precisely determined but seem to be more or less equivalent (4,47). This relative excess of enzyme is consistent with the observation that only a small number of enzyme molecules are apparently phosphorylated and thus active in vivo. It was also noteworthy that this basal activity was clearly not increased in strains overproducing as much as 50-fold the GlmM protein, a result indicating that the total amount of phosphorylated enzyme present in cells was unchanged and thus probably tightly regulated. Any specific regulation of the activity of this enzyme which catalyzes the first step in this reaction sequence could adjust in some way the synthesis of GlcN-1-P molecules to the specific requirements of the peptidoglycan and lipopolysaccharide pathways. The extent of enzyme phosphorylation could therefore be an important factor in the control of enzyme activity which should be investigated in detail now. Taking advantage of the plasmids constructed in this work, the enzyme is now being purified in large amounts for more precise investigations of its kinetic parameters and structure.