The Gene glvA of Bacillus subtilis 168 Encodes a Metal-requiring, NAD(H)-dependent 6-Phospho-α-glucosidase

The gene glvA (formerlyglv-1) from Bacillus subtilis has been cloned and expressed in Escherichia coli. The purified protein GlvA (449 residues, M r = 50,513) is a unique 6-phosphoryl-O-α-d-glucopyranosyl:phosphoglucohydrolase (6-phospho-α-glucosidase) that requires both NAD(H) and divalent metal (Mn2+, Fe2+, Co2+, or Ni2+) for activity. 6-Phospho-α-glucosidase (EC3.2.1.122) from B. subtilis cross-reacts with polyclonal antibody to maltose 6-phosphate hydrolase from Fusobacterium mortiferum, and the two proteins exhibit amino acid sequence identity of 73%. Estimates for the M r of GlvA determined by SDS-polyacrylamide gel electrophoresis (51,000) and electrospray-mass spectroscopy (50,510) were in excellent agreement with the molecular weight of 50,513 deduced from the amino acid sequence. The sequence of the first 37 residues from the N terminus determined by automated analysis agreed precisely with that predicted by translation of glvA. The chromogenic and fluorogenic substrates, p-nitrophenyl-α-d-glucopyranoside 6-phosphate and 4-methylumbelliferyl-α-d-glucopyranoside 6-phosphate were used for the discontinuous assay and in situ detection of enzyme activity, respectively. Site-directed mutagenesis shows that three acidic residues, Asp41, Glu111, and Glu359, are required for GlvA activity. Asp41 is located at the C terminus of a βαβ fold that may constitute the dinucleotide binding domain of the protein. Glu111 and Glu359 may function as the catalytic acid (proton donor) and nucleophile (base), respectively, during hydrolysis of 6-phospho-α-glucoside substrates including maltose 6-phosphate and trehalose 6-phosphate. In metal-free buffer, GlvA exists as an inactive dimer, but in the presence of Mn2+ ion, these species associate to form the NAD(H)-dependent catalytically active tetramer. By comparative sequence alignment with its homologs, the novel 6-phospho-α-glucosidase from B. subtilis can be assigned to the nine-member family 4 of the glycosylhydrolase superfamily.

In 1996, as participants in the Bacillus genome project (33), Sekiguchi and co-workers determined the nucleotide sequence of a 12.4-kb fragment of DNA near the 76°region of the Bacillus subtilis chromosome (34). Following translation of the ten open reading frames within this fragment, a search of the protein data bases revealed that the deduced amino acid sequences of open reading frames glv-1 and glv-2 exhibited similarity to 6-phospho-␤-glucosidase and the IIC domain (GlvC) of the arbutin (PEP-PTS) of E. coli, respectively. It seemed likely, as suggested by Yamamoto et al. (34), that the products of glv-1 and glv-2 might participate in the transport and dissimilation of ␤-glucosides in the Gram-positive spore-forming organism. Contemporaneous with the Bacillus studies in Japan, our pro-gram at the National Institutes of Health was directed toward the purification and characterization of a metal-dependent maltose 6-phosphate hydrolase (MalH) from the anaerobic pathogen Fusobacterium mortiferum (35). This novel enzyme, together with an inducible maltose PEP-PTS permits growth of F. mortiferum on a wide variety of ␣-glucosides including maltose, ␣-methyl glucoside, trehalose, palatinose, and turanose (36). The 6-phospho-␣-glucosidase gene (malH) has recently been cloned, sequenced, and expressed in E. coli (37). Remarkably, the deduced amino acid sequence of MalH exhibited 73% identity (88% similarity) to that deduced from the nucleotide sequence of glv-1 in B. subtilis. These findings raised doubts concerning the initial classification and catalytic activity of the polypeptide encoded by glv-1 (34).
In a collaborative program, we have addressed and resolved these issues. This communication describes the cloning, expression, and site-directed mutagenesis of glv-1 (now designated glvA (33)) from B. subtilis. Purification and characterization of the novel 6-phospho-␣-glucosidase (GlvA) encoded by glvA was facilitated by the availability of the natural substrate for the enzyme (maltose 6-phosphate) and by chemical synthesis of the chromogenic and fluorogenic analogs pNP␣Glc6P and 4MU␣Glc6P, respectively. In contrast to other phosphoglycosylhydrolases, GlvA from B. subtilis exhibits specific requirements for both divalent metal (Mn 2ϩ , Fe 2ϩ , Co 2ϩ , or Ni 2ϩ ) and NAD(H) for activity. Furthermore, by sequence similarity and conservation of functionally important acidic residues, GlvA can be assigned to the nine-member family 4 of the glycosylhydrolase superfamily (31,37).

Materials and Reagents
PD-10 gel filtration columns, isoelectric focusing standards, Ampholine PAG plates (pH 3.5-9.0), DEAE-Sephacel, and phenyl-Sepharose CL-4B were purchased from Amersham Pharmacia Biotech. Ultrogel AcA 44 and TrisAcryl DEAE-M were supplied by Sepracor. Enzymes, nucleotides, cofactors, and trehalose 6-phosphate were obtained from Sigma. Trimethylphosphate, phosphorus oxychloride, and cyclohexylamine were obtained from Aldrich. Pressure concentration cells and Diaflo PM-10 ultrafiltration membranes were from Amicon Corp. [U-14 C]Maltose 6-phosphate was prepared enzymatically by PEP-dependent phosphorylation of the disaccharide by the maltose PEP-PTS in permeabilized cells of F. mortiferum. The radiolabeled disaccharide phosphate was purified by ion exchange chromatography, Ba 2ϩ and ethanol precipitation, and finally by paper chromatography (36). Chemical syntheses of pNP␣Glc6P, pNP␣Man6P, pNP␣Gal6P and 4MU␣Glc6P were initiated with the commercially available non-phosphorylated glycosides (Sigma) using the procedure of Wilson and Fox (24). Selective phosphorylation at the C-6 hydroxyl group of the nonreducing glucopyranose was achieved by use of a mixture of phosphorus oxychloride in trimethylphosphate containing small amounts of water. The phosphorylated derivatives were obtained as white, crystalline cyclohexylamine salts in 25-30% yield.
Bacterial Strains, Plasmids, and Culture Conditions B. subtilis 168 and E. coli strains JM109 and XL1-Blue were grown in Luria-Bertani (LB) medium at 37°C as described previously (38). When required, ampicillin was included in the medium at a final concentration of 50 g/ml. E. coli plasmids pUC119 (Takara Shuzo Co., Kyoto) and a high expression vector pKP1500 (39) were used to construct a plasmid (pKPglv-1) containing glvA. Site-directed mutagenesis was carried out by the Pfu polymerase method (QuickChange sitedirected mutagenesis kit, Stratagene). The desired mutations and the primers used to effect these changes are described in the text (Table IV).

Primers and Sequence Analysis
For the amplification of the gene glvA (also glv-1 (33,34)), two primers were synthesized: forward primer G1PEF, 5Ј-GCCGGAAT-TCATGAAGAAAAAATCATTCTCAA-3Ј (the glvA sequence is italicized and the EcoRI site is underlined) and reverse primer G1PBR, 5Ј-GCGCGGATCCCTGATTGATCAGTTCTTCG-3Ј (the sequence complimentary to the downstream region of glvA is italicized and the BamHI site is underlined). PCR amplification was performed with the Gene-Amp PCR Core kit (Perkin Elmer) using 0.1 g of B. subtilis 168 genomic DNA as template, 10 l of 10 ϫ reaction buffer, 8 l of 25 mM MgCl 2 , 2 l each of 10 mM dNTP, 30 pmol of each primer, and 0.5 unit of Taq polymerase in a total volume of 100 l. The annealing temperature was 56°C. The PCR product was analyzed by agarose gel electrophoresis and subsequently digested with EcoRI and BamHI. The 0.5-kb EcoRI and 0.95-kb EcoRI-BamHI restriction fragments corresponding to the 5Ј and 3Ј regions of glvA, respectively, were purified with GeneClean II kit (BIO 101). The fragments were ligated into the corresponding sites of pUC119 and used for transformation of E. coli JM109. Nucleotide sequences of the fragments inserted into the recombinant plasmids (designated pUC119EE and pUC119EB, respectively) were determined by the following procedure. Transformed cells were boiled in water, and the sample was transferred to the PCR reaction mixture containing universal forward and reverse primers of pUC plasmids (M13 primers M4 and RV; Takara), Taq polymerase, and dNTP. After amplification, the primers were removed by Microcon (Amicon). Sequences were determined using a Taq Dye Primer Cycle Sequencing kit (Perkin Elmer) and a 373A DNA sequencer from Applied Biosystems. After confirmation of the sequences, the 0.95-kb EcoRI-BamHI fragment from pUC119EB was purified and ligated to the EcoRI and BamHI sites of pKP1500 to form pKPGEB. Transformants of E. coli JM109 containing the recombinant plasmid were boiled in water and transferred to the PCR reaction mixture containing primers G1PEF and G1PBR, Taq polymerase, and dNTP. After amplification and transformation, those colonies that contained a 1.45-kb DNA restriction fragment were considered to harbor a plasmid (pKPglv-1) containing the insert in the correct orientation. For confirmation, pKPglv-1 plasmid DNA was isolated from the transformant, digested with EcoRI or HindIII, and subjected to agarose gel electrophoresis. The size of the restriction fragments corresponded to those expected from the sequence of glvA, thereby confirming the presence of the complete gene in pKPglv-1.

Growth of Cells and Preparation of Cell Extract
E. coli JM109 (pKPglv-1) was grown at 37°C in LB medium containing ampicillin (100 g/ml). Cells were harvested by centrifugation (13,000 ϫ g for 10 min at 5°C) and washed by resuspension and centrifugation from 25 mM Tris-HCl buffer (pH 7.5) containing 1 mM MnSO 4 (TM buffer). The washed cell pellet (ϳ80 g) was resuspended in 120 ml of TM buffer, and the cells were disrupted (at 0°C) by 2 ϫ 1.5-min periods of sonic oscillation with a Branson model 350 sonifier operating at ϳ75% of maximum power.

Purification of 6-Phospho-␣-glucosidase
The enzyme was purified by conventional low pressure chromatography, and all procedures were performed at 4°C in a cold room. Column flow rates were maintained by a P-1 peristaltic pump interfaced with a Frac-100 collector. Protein in column eluents was monitored at 280 nm by a UV-1 optical control unit connected to a single channel chart recorder (all instrumentation from Pharmacia Biotech).
Step 1: Preparation of High Speed Supernatant Fluid (HSS)-The sonicated preparation was centrifuged (25,000 ϫ g for 30 min at 5°C) to remove intact cells and cell debris. The supernatant was collected and centrifuged at 180,000 ϫ g for 2 h at 5°C. The clarified HSS was transferred to dialysis sacs (molecular weight cut-off, 6000 -8000) and dialyzed overnight against 4 liters of TM buffer.
Step 2: DEAE-TrisAcryl-M (Anion Exchange) Chromatography-Dialyzed HSS (ϳ120 ml) was transferred (0.8 ml/min) to a column (2.6 ϫ 16 cm) of DEAE-TrisAcryl-M previously equilibrated with TM buffer. The column was washed to remove nonadsorbed material, and 6-phospho-␣-glucosidase was eluted with 800 ml of a linear, increasing concentration gradient of NaCl (0 -300 mM) in TM buffer. Fractions of 10 ml were collected, and 10 l of each fraction was tested for enzyme activity by the formation of a yellow color in microtiter wells containing 100 l of the standard pNP␣Glc6P reaction mixture. Fractions 22-27 (inclusive) were pooled and concentrated to 25 ml by pressure filtration (Amicon PM-10 membrane, 35 psi). Ammonium sulfate was then added slowly, and with gentle stirring, to a final concentration of 0.75 M.
Step 3: Phenyl-Sepharose CL-4B (Hydrophobic) Chromatography-The solution from step 2 was transferred (0.4 ml/min) to a 2.6 ϫ 16-cm column of phenyl-Sepharose CL-4B equilibrated with TM buffer containing 0.75 M (NH 4 ) 2 SO 4 . The column was washed with equilibration buffer to remove material that did not bind, and then 600 ml of a decreasing, linear gradient of (NH 4 ) 2 SO 4 (300 -0 mM) in TM buffer was passed through the column. Fractions of 10 ml were collected, and enzyme was recovered in a broad protein peak comprising fractions 25-45. These fractions were pooled and concentrated to 7.5 ml.
Step 4: Ultrogel AcA-44 (Molecular Sieve) Chromatography-Approx-imately 2.5 ml of the preparation from step 3 was applied at a flow rate of 0.15 ml/min to a column (1.6 ϫ 94 cm) of Ultrogel AcA-44 previously equilibrated with TM buffer containing 0.1 M NaCI. Fractions of 2.15 ml were collected, and maximum levels of enzyme activity were found in fractions 49 -53, inclusive. These fractions were concentrated to 2 ml, and aliquots were either frozen directly in dry ice or glycerol was added to a final concentration of 10% prior to storage of the enzyme at Ϫ20°C.

Assay of Enzyme Activity
The chromogenic analog pNP␣Glc6P was used as substrate in the discontinuous assay for 6-phospho-␣-glucosidase activity. The 2-ml reaction mixture (at 37°C) contained, when required, 50 mM Tris-HCl buffer (pH 7.5), 1 mM pNP␣Glc6P, 0.5 mM MnS0 4 , and 0.1 mM NAD(H). After addition of the enzyme preparation, samples of 0.25 ml were removed at intervals of 0.5, 1, 1.5, 2, 2.5, and 3 min and immediately injected into 0.75 ml of 0.5 M Na 2 CO 3 solution. The A 400 nm of the yellow solution was measured, and the amount of pNP was calculated by assuming a molar extinction coefficient for the p-nitrophenoxide anion ⑀ ϭ 18,300 M Ϫ1 cm Ϫ1 . One unit of 6-phospho-␣-glucosidase activity is the amount of enzyme that catalyzes the formation of 1 mol of pNP per min at 37°C.

Electrophoresis Procedures
Native gel electrophoresis and SDS-PAGE were carried out in the Novex XCell Mini-Cell system according to manufacturer's instructions. Electrophoresis of proteins under nonreducing (native) conditions was performed in Tris-glycine (4 -20%) gels, from Novex, with Tris-glycine (pH 8.3) running buffer. For SDS-PAGE experiments, Novex NuPage (4 -12%) Bis-Tris gels and MES-SDS running buffer (pH 7.3) were used together with Novex Mark 12 wide range or Bio-Rad low range molecular weight protein standards. In Western blots, proteins were transferred to nitrocellulose membranes using NuPage transfer buffer (pH 7.2) and SeeBlue prestained standards. Immunodetection of 6-phospho-␣-glucosidase with polyclonal antibody to maltose 6-phosphate hydrolase was as described previously (35).

Analytical Methods
The concentrations of glucose and Glc6P were determined enzymatically in an NADP ϩ -coupled assay that contained (in 1 ml) 0.1 M potassium phosphate (pH 7) buffer, 1 mM MgCl 2 , 5 mM ATP, 1 mM NADP ϩ , and 2 units each of Glc6P dehydrogenase (EC 1.1.1.49) and hexokinase (EC 2.7.1.1). Formation of NADPH was followed in a Beckman DU 70 recording spectrophotometer, and a molar extinction coefficient ⑀ ϭ 6,220 M Ϫ1 cm Ϫ1 was assumed for calculation of NADPH produced (i.e., equivalent to glucose or Glc6P formed). The Pharmacia Biotech Multiphor flat-bed electrophoresis unit and precast Ampholine PAG plates (pH range 3.5-9.5) were used for electrofocusing experiments as described previously (40). Protein concentrations were routinely determined by the BCA protein assay kit (Pierce). Chromatographic procedures are described in a previous report (35). The N-terminal amino acid sequence of 6-phospho-␣-glucosidase was determined by automated Edman degradation in a 494A Procise sequenator (Applied Biosystems, Inc.) using pulse liquid chemistry. PTH derivatives were identified by on-line high pressure liquid chromatography. The mass of purified 6-phospho-␣-glucosidase was determined by electrospray in an HP1100 mass spectrometer. The enzyme sample was dissolved in 0.05% trifluoroacetic acid, injected onto a Zorbax 300SB-C3 narrow bore (2.1 ϫ 150 mm) high pressure liquid chromatography column, and eluted with a gradient of 5% acetic acid to 100% acetonitrile. A mass range from m/z 700 to 2000 was scanned every 4 s, and the protein mass (50,510) was obtained by deconvolution from the only peak that eluted from the column.

Analytical Ultracentrifugation
Analytical ultracentrifugation experiments were conducted at 20°C in a Beckman Optima model XL-I instrument equipped with a fourplace An-Ti rotor. Specific absorption coefficients were determined by a combination of absorbance and interference optics. Absorbance readings on protein samples were measured in a Perkin Elmer model 320 double-beam spectrophotometer at 20°C. Protein concentrations (mg/ ml) were determined on the same solutions using the analytical ultracentrifuge in an interference mode as a differential refractometer (41) and an experimentally determined value of 3.191 Ϯ 0.005 fringes (mg/ ml) Ϫ1 for the same instrument. 2  6-Phospho-␣-glucosidase from B. subtilis windows was used with initial volumes of 130 l of protein solution and 410 l of dialysate buffer. After temperature equilibration at 20.0°C at 3000 rpm, the rotor speed was increased to 10,000 rpm to initiate boundary formation and to 20,000 rpm (in 2000-rpm steps) until protein and solvent-side menisci were matched (total time ϳ10 min) and then decelerated to 3000 rpm; 10 interference scans were then recorded at 2-min intervals (42). For 10 scans, the difference between the fringes in the plateau and solvent sides of the boundary (100 data points in both radial positions) was constant and within 0.1% accuracy during the period of data collection. The specific absorption coefficient for 6-phospho-␣-glucosidase was determined to be A 280 nm 1 cm ϭ 1.25 Ϯ 0.01 ml/mg (average of three independent measurements). This value is similar to that calculated from the amino acid composition of the protein (1.20 ml/mg). Sedimentation velocity experiments were conducted using charcoal-filled double-sector Epon 12-mm centerpieces. Enzyme solution (A 280 nm ϳ1.0) was loaded on the right (330 l per channel) with the reference buffer on the left (340 l per channel). After thermal equilibration at 3000 rpm, the rotor was accelerated to 40,000 rpm, and radial scans were collected at 280 nm (0.003-cm step size, 4-min intervals) with triple averaging in a continuous scan mode. For sedimentation equilibrium experiments, a 12-mm cell equipped with a carbonfilled 6-channel centerpiece and plane quartz windows was used. Enzyme solutions (A 280 nm from 0.17 to 0.35) were loaded on the right (100 l per channel) with the reference buffer on the left (110 l per channel). Radial scans at 10,000 rpm with 13 averages were made at 280 nm in 0.001-cm steps (step mode) after 18 and 20 h (equilibrium was reached by 16 h). Analysis of ultracentrifugation data was performed as described previously (43) with software from Beckman, Inc. and A. P. Minton (NIDDK, National Institutes of Health). The densities of dialysate buffers (20.0°C) were determined using an Anton Paar model DMA 58 densitometer. The partial specific volume for the protein ( ϭ 0.720 ml/g) was calculated from the amino acid composition (44).

RESULTS
Nucleotide Sequence of glvA-The nucleotide sequence of the gene glvA (previously called glv-1 (34)) is presented in Fig. 1. The 1347-base pair open reading frame begins with an ATG initiation codon at nucleotide position 15 and terminates with a TAA stop codon at position 1362. A putative ribosome binding site AAGGAGGT precedes the start codon. Translation of the codon sequence of glvA predicts a polypeptide of 449 residues of calculated M r ϭ 50,513 and theoretical pI ϭ 4.77. The 48.1 mol% (GϩC) base composition of glvA is somewhat higher than the average 43.5 mol% (GϩC) content of the B. subtilis chromosome (33).
Expression of GlvA-Cells of E. coli JM109 transformed with plasmid pKP glv-1 ( Fig. 2A) produced a polypeptide that crossreacted strongly with antibody prepared against purified MalH from F. mortiferum (Fig. 2B, lane 2). The estimated M r of this immunoreactive protein was of the size (ϳ51 kDa) expected for the product of glvA, and the protein was not detected in a Western blot of an extract prepared from plasmid-free E. coli JM 109 (Fig. 2B, lane 1). The cross-reactivity noted in Fig. 2B (lane 2) indicated epitopic and conformational similarity between GlvA and the MalH, and an extract of E. coli JM 109 (pKPglv-1) rapidly hydrolyzed pNP␣Glc6P to form the (yellow) p-nitrophenolate anion and Glc6P (specific activity 0.22 mol of pNP␣Glc6P hydrolyzed/mg protein/min). Under the same reaction conditions, there was no detectable hydrolysis of the nonphosphorylated analog pNP␣Glc, and GlvA was tentatively identified as a 6-phospho-␣-glucosidase.
Purification of GlvA-The activity of GlvA in an extract of E. coli JM109 (pKPglv-1) declined by Ͼ90% during overnight dialysis at 4°C in 25 mM Tris-HCl (pH 7.5) buffer, but there was little loss of activity when the buffer contained 1 mM Mn 2ϩ . For this reason, Mn 2ϩ ion was included in all buffers throughout the purification of the enzyme (Table I). The four-stage procedure yielded about 50 mg of pure 6-phospho-␣-glucosidase from 25 g of cells (wet weight). Although successful in terms of providing a significant amount of purified material, the procedure yielded a final preparation of extremely low activity. Indeed, the specific activity of the purified enzyme (0.36 units/ mg) was not significantly greater than that of the original high speed supernatant (0.22 units/mg). The instability of the 6-phospho-␣-glucosidase from B. subtilis (see below) was reminiscent of the behavior of MalH during its purification from F. mortiferum (35).
Homogeneity and Size-Analysis of denatured 6-phospho-␣glucosidase by SDS-PAGE (Fig. 3A, lane 4) revealed a single polypeptide of M r ϳ 51,000, and the molecular weight of the protein determined by electrospray-MS was 50,510. However,  since the enzyme emerged close to the void volume of the AcA-44 gel filtration column (exclusion limit about 200 kDa; Table I, step 4), it seems that in the native state 6-phospho-␣glucosidase exists as a tetrameric species. Enzyme preparations that had been heated (or unheated) in the presence or absence of dithiothreitol revealed a single ϳ51-kDa polypeptide by SDS-PAGE (Fig. 3B). The absence of higher molecular weight species thus precludes participation of disulfide bonds in formation of the oligomeric structure. The homogeneity of 6-phospho-␣-glucosidase was confirmed by the unambiguous data obtained by microsequence analysis of the protein. The first 37 residues from the N terminus agreed perfectly with the sequence deduced from glvA (MKKKSFSIVIAGGGSTFTPGIV-LMLLDHLEEFPIRKL). Surprisingly, electrophoresis of the purified enzyme under nondenaturing conditions revealed two polypeptides (Fig. 3A, inset). Both species cross-reacted with antibody to MalH, and both catalyzed the in situ hydrolysis of the fluorogenic substrate 4MU␣Glc6P upon addition of Mn 2ϩ and NAD ϩ (data not shown). Analytical electrofocusing of purified 6-phospho-␣-glucosidase (Fig. 3C) also revealed two species. The estimated pI values of 4.3 and 4.5 for the major (A) and minor (B) polypeptides, respectively, agreed quite well with the pI of 4.77 calculated from the deduced amino acid composition of GlvA.
Requirements for NAD(H) and Me 2ϩ for GlvA Activity-The extensive loss in enzyme activity incurred throughout the purification (Table I) could be attributed to either (a) irreversible inactivation due to conformational changes of the protein or (b) removal of a required cofactor or activator. Preliminary studies showed that incubation of 5-l (ϳ100 g) samples of 6-phospho-␣-glucosidase with 5, 10, and 15 l of an extract of plasmidfree E. coli JM109 (ϳ20 mg/ml) enhanced enzyme activity by 8-, 15-, and 19-fold, respectively. The extract of E. coli was then heated (2 min at 100°C), and precipitated material was removed by centrifugation. The clarified supernatant also activated the enzyme. Finally, a portion of E. coli extract was separated into high and low molecular weight components by passage through a PD-10 gel filtration column. The low molecular weight fraction produced Ͼ80% of the total increase in 6-phospho-␣-glucosidase activity. Collectively, these findings suggested that a small, relatively heat-stable effector was required for catalytic activity. A study of potential cofactors (Table II) identified NAD ϩ and, to a lesser degree, NADH as the enzyme activators. With this knowledge, a more thorough examination of the metal requirements of the enzyme was under-taken. For these experiments, 6-phospho-␣-glucosidase was first exhaustively dialyzed against 25 mM Tris-HCl (pH 7.5) buffer to remove endogenous Mn 2ϩ and other metal ion contaminants. Enzyme activity was then measured in the standard assay containing NAD ϩ , pNP␣Glc6P, and desired Me 2ϩ ion. In the presence of NAD ϩ alone, hydrolysis of substrate was barely discernible (Table II). However, addition of Co 2ϩ , Ni 2ϩ , Fe 2ϩ , or Mn 2ϩ to the assay elicited a 20 -100-fold increase in enzyme activity. These data established that both NAD(H) and divalent metal were prerequisites for GlvA activity. Inclusion of EDTA (2 mM) in these assays resulted in complete inactivation of the enzyme (data not shown).
Properties of 6-Phospho-␣-glucosidase-Enzyme activity was greatest between 35 and 38°C in buffer of pH range 7.5-8.0. Although 50 mM Tris-HCl (pH 7.5) was used routinely as the assay buffer, comparable activities (at pH 7.5 and same molarity) were measured in MES, Bicine, Tricine, HEPES, or imidazole buffers (data not shown). In the presence of 0.1 mM NAD ϩ and 1 mM Mn 2ϩ , it was found that the enzyme exhibited Michaelis-Menten saturation kinetics with pNP␣Glc6 as substrate (K m ϭ 0.09 mM, V max ϭ 2.2 mol hydrolyzed/mg protein/ min). The K m values for NAD ϩ and NADH were estimated to be No detectable activity a Assay conditions are described under "Experimental Procedures"; all assays contained 0.5 mM Mn 2ϩ . The enzyme was incubated (ϳ5 min) with 1 mM each compound prior to addition to the assay. b APAD, 3-acetylpyridine adenine dinucleotide; NMN, nicotinamide mononucleotide. c Assay conditions are described under "Experimental Procedures." The enzyme was incubated for 5 min in 50 l of 50 M Tris-HCl buffer (pH 7.5) containing 1 mM NAD ϩ and 2 mM of desired Me 2ϩ prior to addition to the 2-ml assay that contained 50 mM Tris-HCl buffer (pH 7.5), 1 mM pNP␣Glc6P, and 0.5 mM desired Me 2ϩ . 3.7 ϫ 10 Ϫ5 M and 2.0 ϫ 10 Ϫ4 M, respectively. Perhaps NAD ϩ is the preferred nucleotide for in vivo activation of the enzyme. The K m for Mn 2ϩ was determined to be 0.31 mM. Although pNP␣Glc6P was rapidly hydrolyzed by 6-phospho-␣-glucosidase, there was no detectable cleavage of nonphosphorylated nitrophenyl-glycosides including pNP␣Glc, pNP␣-galactopyranoside, and pNP␣-mannopyranoside. Enzyme discrimination with respect to the spatial orientation of hydroxyl groups in the glucopyranosyl moiety of pNP␣Glc6P is evident from the fact that neither the C-2 nor the C-4 epimer (pNP␣Man6P and pNP␣Gal6P, respectively) is a substrate for 6-phospho-␣-glucosidase. Maltose 6-phosphate and trehalose 6-phosphate are substrates for GlvA from B. subtilis, and hydrolysis of these O-␣ (1-4)-and O-␣ (1-1Ј)-linked disaccharide phosphates yielded equimolar amounts of Glc6P and glucose (data not shown). The rate of cleavage of [U-14 C]maltose 6-phosphate was increased ϳ10-fold in the presence of NAD ϩ , and chromatographic analysis confirmed formation of 14 C-labeled Glc6P and glucose as hydrolysis products (Fig. 4). Unexpectedly, the autoradiogram also revealed the presence of two other compounds (X and Y) in the reaction mixture, but the two compounds have not yet been identified. A fluorogenic analog of maltose 6-phosphate (4MU␣Glc6P) was also hydrolyzed by 6-phospho-␣-glucosidase to form Glc6P and the intensely fluorescent 4-methylumbelliferone (Fig. 5). These extremely sensitive microtiter assays provide visual confirmation for the requirements of phosphorylation of the substrate and divalent metal and dinucleotide specificity for GlvA-catalyzed hydrolysis.
Analytical Ultracentrifugation-Sedimentation velocity analysis was used to characterize 6-phospho-␣-glucosidase in solutions containing different concentrations of Mn 2ϩ ion. Time derivative analysis showed that in the absence of Mn 2ϩ , the enzyme exists as a single species (S 20,w ϳ 6.1 S) (Fig. 6A). The apo-enzyme showed a single component of M r ϭ 105,000 in sedimentation equilibrium experiments (Table III). This value approximates that calculated for a dimer from the amino acid sequence of GlvA (M r ϭ 101,026). In the presence of Mn 2ϩ , the major sedimenting component had a sedimentation coefficient of ϳ9.3 S, which is consistent with a tetrameric form. Small fractions of monomer (ϳ4 S) were present also (Fig. 6, B and C). Weight average molecular weight determinations of the enzyme in buffers containing Mn 2ϩ (Table III) were lower than that expected for a homotetramer (M r ϭ 202,052). Increasing the concentration of Mn 2ϩ from 1 to 5 mM did not lead to complete conversion of dimer to tetramer (Table III). Indeed, in the presence of 1 and 5 mM Mn 2ϩ , the best fits of data sets were obtained when molecular weights for two components were constrained at 100,000 and 200,000, respectively, using a model for nonassociating species. These analyses yielded estimates of ϳ25 and ϳ75% of dimer and tetramer, respectively. Thus, some dimer seems to be incompetent in associating to tetramer. Attempts to fit the sedimentation equilibrium to either self-associating or nonassociating models involving 1-2-4, 2-3-4, and 1-4 (where 1 denotes the M r ϭ 50,513 monomer) gave unsatisfactory fits.
Site-directed Mutagenesis of GlvA-Comparative alignment of the amino acid sequences of GlvA from B. subtilis, MalH of F. mortiferum, and other members of family 4 illustrates the extensive homology of these glycosyl hydrolases (Fig. 7). Significantly, three acidic residues at sequence positions 41, 111, and 359 of GlvA are conserved in all members of family 4, and site-directed mutagenesis revealed that these residues are essential for activity of GlvA. In experiments summarized in Table IV, residues Asp 41 , Glu 111 , and Glu 359 of GlvA were changed to either Gly or to the corresponding homolog (Glu or Asp). E. coli XL1-Blue was transformed with plasmids encoding glvA with the desired mutation, and Western blots verified the production by all transformants of an immunoreactive protein of the size (ϳ51 kDa) expected for full-length GlvA (data not shown). However, the level of GlvA activity in cell extracts of the transformants (Table IV) was Ͻ1% that of E. coli XL1-Blue (pKPglv-1) and was not significantly higher than that of an extract prepared from plasmidfree E. coli XL1-Blue (approximately 3 nmol of pNP␣Glc6P hydrolyzed/min/mg protein). DISCUSSION The gene glvA (formerly glv-1 (34)) of B. subtilis encodes a novel 6-phospho-␣-glucosidase. To our knowledge, this is the first report of a phosphoglycosylhydrolase that requires both dinucleotide (NAD(H)) and divalent metal (Mn 2ϩ , Fe 2ϩ , Co 2ϩ , or Ni 2ϩ ) for activity. Genes glvA and glvC (formerly glv-2) are part of an operon that facilitates the dissimilation of ␣-glucosides via a maltose (PEP-PTS) in this Gram-positive organism. We show here that the product of this PTS activity (maltose 6-phosphate) is hydrolyzed by 6-phospho-␣-glucosidase to yield Glc6P and glucose (Scheme 1). Surprisingly, two additional products were detected in the reaction mixture (Fig.  4, X and Y). The two compounds are not artifacts of the chromatographic procedure, and because the maltose 6-phosphate was of high purity, it is unlikely that the two compounds are derived from impurities in the substrate. The primary catalytic activity of GlvA is almost certainly the hydrolytic cleavage of the O-glycosyl linkage, but the formation of compounds X and Y may indicate a secondary catalytic function for GlvA.
When the sequences of two or more enzymes can be aligned over an entire domain, they are assigned to the same family, and on this basis some 50 families of glycosylhydrolases have now been described (31,32 (50 -52) have shown that for ␤-glucosidase from Agrobacterium faecalis two glutamyl residues (Glu 170 and Glu 358 ) function as acid catalyst (AH) and nucleophile/base (A Ϫ ), respectively. The  a Samples of GlvA were dialyzed overnight at 4°C against 25 mM Tris-HCl buffer, pH 7.5, containing 100 mM NaCl and the indicated component, and subsequently diluted with the dialysis buffer.
b From time-derivative analyses (see Fig. 6). c Weight average molecular weights for global fits of three protein concentrations were from sedimentation equilibrium analysis. d Calculated from sedimentation equilibrium analyses using a model for nonassociating species (see text).
Family 4 presently comprises a heterogeneous group of enzymes that includes ␣-galactosidase, ␣-glucosidase, putative 6-phospho-␤-glucosidase, and 6-phospho-␣-glucosidase (Fig. 7). The nine members of this family are all of bacterial origin, and, except for the truncated 6-phospho-␣-glucosidase of E. coli, the polypeptides are of comparable length (average 446 amino acids). Inspection of Fig. 7 reveals two acidic residues that are conserved in the motifs GQ(E/D)T(X)GPGG and VE(X) of all family 4 members. In the case of 6-phospho-␣-glucosidase from B. subtilis, these glutamyl residues (Glu 111 and Glu 359 ) are 248 residues apart, and their mutation, to either glycine or aspartic acid, yields immunoreactive but catalytically inactive protein.
By analogy with members of family 1, the two conserved acidic residues may also represent the catalytically functional residues in glycosylhydrolases of family 4. However, the question arises as to why the three enzymes of family 4 that have now been characterized (GlvA, MalH 3 (35) and ␣-galactosidase (65)) also require NAD(H) and Me 2ϩ ion for activity. Except for Mg 2ϩ -dependent ␤-galactosidase from E. coli, there are few reports of the activation of glycosylhydrolases by metal ions. For ␤-galactosidase, Mg 2ϩ does not seem to be an electrophilic catalyst; rather, this divalent cation elicits a conformational change within the protein such that the catalytic glutamyl residues assume the correct orientation at the active site (66 -68). Our ultracentrifugal analyses of GlvA show that the Mn 2ϩ ion promotes the association of dimers of 6-phospho-␣-glucosidase to the tetrameric state. Furthermore, removal of Mn 2ϩ from GlvA by chelation with EDTA causes dissociation of the active tetramer to the inactive dimeric form of the enzyme.
Although these results are evidence for a structural role for Mn 2ϩ , they do not preclude a catalytic function for this (and other) divalent metal ion(s).
The role of dinucleotide in activation of 6-phospho-␣-glucosidase cannot be deduced from our studies. The requirement for NAD(H) is clearly specific, and NAD(P)H, which differs only in the presence of a phosphoryl moiety at the 2Ј-OH of the adenine ribose ring, is not an acceptable substitute (Table II, Fig. 5). We have obtained no evidence for oxido/reduction of either NAD ϩ or NADH, and it seems unlikely that the dinucleotide serves as a reactant during substrate cleavage. Both NAD ϩ and NADH bind strongly (but noncovalently) to 6-phospho-␣-glucosidase, and computer-based analysis of the sequence of GlvA with its homologs reveals a putative dinucleotide-binding motif in all members of family 4 (Fig. 7). This domain comprises a ␤␣␤ unit containing a glycine-rich turn or loop between the first ␤ strand and the dinucleotide binding helix (69,70). Both ␤A and ␤B strands of this structural element consist almost entirely of hydrophobic amino acids, and the two glycine residues that follow ␤A (Gly 12 and Gly 14 in GlvA) are conserved in the sequences of all nine proteins. Although the third glycine residue in the G(X)G(X)(X)G fingerprint region (69,70) does not seem to be present, Gly 20 could assume this role in the B. subtilis enzyme. Indeed, the number of amino acids between the second and third glycyl residues of this triplet has been found to vary depending on the length of the loop between ␤A and ␣A (71). It is also noteworthy, that Ser 15 follows the second Gly residue in GlvA, and this pattern is found in all family 4 members. We suggest that this positionally conserved serine residue may hydrogen bond to the nicotinamide carboxyamide and nicotinamide phosphate oxygen atoms in a manner similar to the conserved asparagine residue of the glutamate dehydrogenase family (70). Further evidence in favor of the proposed nucleotide binding domain is the presence of a conserved aspartic acid residue (Asp 41 for GlvA) at the end of the ␤B strand in family 4 members. In many NAD ϩ and FAD-dependent enzymes, an acidic residue at this position in the ␤␣␤ fold is known to hydrogen bond to the 2Ј and 3Ј hydroxyl groups of the adenine ribose. For GlvA we find that site-directed mutagenesis of Asp 41 yields proteins (D41G and D41E) that are catalytically inactive (Table IV). These findings suggest a role for this conserved aspartyl residue in the binding of NAD(H) to 6-phospho-␣-glucosidase.
In 1971, Burstein and Kepes described both the instability of ␣-galactosidase from E. coli (65) and the unexpected require-3 J. Thompson, unpublished data.  6-Phospho-␣-glucosidase from B. subtilis ments of this hydrolase for NAD(H) and Mn 2ϩ ion for activity. A quarter of a century later, we report similar characteristics and requirements for GlvA from B. subtilis, MalH from F. mortiferum, 3 and CelF (6-phospho-␤-glucosidase) from E. coli. 4 The inherent instability can now be explained by the structural requirement of these enzymes for Mn 2ϩ , Co 2ϩ , Ni 2ϩ , or Fe 2ϩ and by progressive loss of an essential cofactor (NAD(H)) during chromatography. The requirements for dinucleotide and divalent ion from the transition series of elements may prove to be the salient characteristics that permit distinction of family 4 members from all others in the glycosylhydrolase superfamily.