Purification and characterization of blood group A-degrading isoforms of alpha-N-acetylgalactosaminidase from Ruminococcus torques strain IX-70.

To cleave blood group A immunodeterminants from erythrocytes (Hoskins, L. C., Larson, G., and Naff, G. B. (1995) Transfusion35, 813-821), we purified and characterized α-N-acetylgalactosaminidase (EC 3.2.1.49) activity from culture supernatants of the human fecal bacterium Ruminococcus torques strain IX-70. Three isoforms separated during hydrophobic interaction chromatography. Hydroxyapatite chromatography further resolved the most hydrophilic, isoform I, into isoforms IA and IB. The most hydrophobic, isoform III, differed from IA and IB by a more acidic pH optimum, greater heat resistance, greater sensitivity to alkylating agents, and anomalous retardation during gel filtration chromatography. Isoform IB differed from IA and III in N-terminal amino acid sequence and in sensitivity to EDTA inhibition. Each cleaved nonreducing α(1→3)-N-acetylgalactosamine residues from human blood group A and AB mucin glycoproteins, Forssman hapten, and blood group A lacto series glycolipids. The apparent molecular mass of denatured isoform subunits of IA, IB, and III-PII (158, 173, and 201 kDa, respectively) bore no integer relationship to the apparent molecular mass of the native isoforms (265, 417, and 530 kDa), but the latter bore a ratio of 1.96:3.09:3.93 to the weight-average apparent molecular mass of native IA (135 kDa), suggesting that the isoforms are multimers of a 135-kDa sequence. Isoforms IA and III-PII had an identical N-terminal amino acid sequence which showed homologies to the N-terminal sequence of sialidases produced by Bacteroides fragilis SBT3182, another commensal enteric bacterium.

Substrates-para-Nitrophenyl (pNP)-␣and ␤-glycopyranosides were purchased from Sigma and Toronto Research Biochemicals, Inc. Toronto, Canada. Hog gastric mucin was purified (29) from a commercial source (Lot 26303, ICN Biochemicals, Cleveland, OH) and contained 1.3 and 2.6 g/ml minimal hemagglutination inhibiting concentrations (MHIC) of blood group A and H antigen, respectively. Mucin glycoprotein subunits were purified from human small intestinal mucosal scrapings obtained at autopsy from a blood group A secretor and a blood group AB secretor. Homogenates of mucosal scrapings (20% w/v), prepared in 0.15 M NaCl containing 20 mM NaKHPO 4 buffer, pH 7.4, and 3 mM sodium azide, were incubated for 48 h at 37°C with 0.2 M 2-mercaptoethanol (MCE) and then boiled and centrifuged at 16,000 ϫ g. Ethanol was added to the supernatant at 0 -2°C to 60% v/v. The precipitate was dissolved in Pronase buffer and digested sequentially with ribonuclease, deoxyribonuclease, and Pronase at respective concentrations of 0.6 and 12 Kunitz units and 0.4 PUK units/g mucosal wet weight before being again boiled, centrifuged, and reprecipitated with ethanol to 60% v/v. The final precipitate was lyophilized after dissolving in and dialyzing against 1 mM ammonium acetate. The yield of group A subunits is as follows: 11 mg of glycoprotein/g of mucosal wet weight; composition, 41.8% w/v hexoses, 11.3% w/v protein, and 1.3 g/ml and 42 g/ml MHIC of A and H antigen, respectively. The yield of group AB subunits is as follows: 5.7 mg of glycoprotein/g of mucosal wet weight; composition, 28.1% w/v hexoses, 12.0% w/v protein, and 0.08 and 0.65 g/ml MHIC of A and B antigen, respectively, with no detectable H antigen (MHIC Ն333 g/mg).
Culture Medium-The anaerobic culture medium was "Medium 75" (29) modified by replacing 3 g/liter of the 15 g/liter Casamino acids (Difco) with 3 g/liter of a pancreatic digest of casein (Trypticase, BBL Microbiology Systems, Cockeysville, MD) and deleting tryptophan and hog gastric mucin supplements.

Methods
Glycosidase activities were measured in samples diluted to ensure that enzyme concentration was rate-limiting using pNP-glycopyranoside substrates in EAB at pH 6.0 and at 37°C as described previously (34). Sialidase was measured from the rate of cleavage of sialic acid from serum orosomucoid as described (34), and fucosidase was measured by the rate of cleavage of L-fucose from a purified human gastric mucin with a high H antigen titer using fucose dehydrogenase (Boehringer Mannheim) to measure the released fucose (35). One unit of enzyme activity was defined as that amount cleaving 1 mol of product/ min at pH 6.0 and 37°C.
One mM MgCl 2 and 0.1 mM Na 2 EDTA were included in incubations after we observed that these enhanced and stabilized enzyme activity in crude glycosidase mixtures. 0.1% v/v Triton X-100 enhanced several glycosidase activities by 30 -50% in crude preparations and was included in all enzyme assays using pNP-glycoside substrates. We included 1 mg/ml bovine serum albumin in assays of purified ␣-GalNAcase preparations at concentrations Յ1 g/ml since this improved the coefficient of variation (CV) among replicates from Ϯ40 to Ϯ6%.
Characterization of Substrate Specificities-Cleavage of GalNAc from blood group A and AB mucin glycoprotein subunits by purified enzyme fractions was measured in incubation mixtures containing 2.1 mg of mucin glycoprotein subunits and 1.3 milliunits of enzyme per ml of ESBA. Aliquots were removed at intervals, heated 3 min at 100°C, and cooled, and ethanol was added to 80% v/v. After 16 h at Ϫ20°C the aliquots were centrifuged. The glycoprotein precipitate was dissolved in 0.1 M NaCl and analyzed for residual hexose using anthrone (36) with D-galactose as standard, and for ABH blood group antigens by hemagglutination inhibition titration (15). The supernatants were flash evaporated to dryness, reconstituted in distilled water, and analyzed for N-acetylhexosamines using GalNAc as standard (37). Results were expressed as nmol of GalNAc released per h/mg of mucin. For further identification of saccharide cleavage products the reconstituted solutions were deionized by passage through AG 50W-X4 and AG 1-X8 ion exchange resins (Bio-Rad); the eluates were evaporated to dryness and reconstituted in a minimal amount of distilled water. They were applied along with saccharide standards to thin layer chromatography (TLC) plates (HPTLC Silica Gel 60 plates, no. 5641, Merck, Darmstadt, Federal Republic of Germany) which had been presoaked in 0.2 M borate buffer at pH 8.0 before being activated by heating 30 min at 180 -200°C. Chromatograms were developed using n-butanol/pyridine/ methanol/H 2 O (3:3:3:1). The saccharide spots were detected with cupric acetate reagent (38) and by examination under UV light at 365 nm.
Measurement of Glycolipid Substrate Specificities (G.L.)-Enzymatic hydrolysis of glycolipids was performed as described previously (31) at 37°C in EA buffer, pH 6.4, containing 7-10 mM sodium deoxycholate. Incubation mixtures contained 130 -1760 milliunits of isoform and 5-24 g of glycolipid substrate in a ratio of 7-88 milliunits of enzyme per g of glycolipid. Aliquots were removed at intervals from 1 to 24 h and desalted on 0.1 g of C 18 prepacked Bond Elut7 columns (Analytichem International, Harbor City, CA). For analysis by thin layer chromatography samples and glycolipid, standards were applied to TLC plates (no. 5641, Merck) that were developed with chloroform/methanol/water, 60:35:8 (v/v/v) and stained with anisaldehyde reagent (33).
Analytical Methods-Protein was assayed with the Folin-phenol reagent (39) or with fluorescamine (40) using human serum albumin as standard. Electrical conductivity of eluate fractions during column chromatography was measured on appropriately diluted samples with a conductimeter (Radiometer A/S, Copenhagen) and converted to NaCl or phosphate concentration by reference to conductivity curves of standard solutions containing NaCl or phosphate in ESBA. Polyacrylamide gel electrophoresis (PAGE) under nondenaturing conditions was performed by standard methods (41) using 3% polyacrylamide stacking and 5% running gels at pH 8.3. PAGE under denaturing conditions (SDS-PAGE) was performed at pH 7.0 by a modification (42) of the method of Weber and Osborne (43) and at pH 8.3 by the method of Laemmli (44,45). Gels were stained for protein with Coomassie Blue or Silver (44) and for carbohydrate with periodic acid-Schiff stain (46). Optical densitometry of gels stained with Coomassie Blue was performed using an optical scanner (model SciScan 5000, U. S. Biochemical Corp.) with image analysis software (Oberlin Scientific Corp. Os-Scan Image Analysis System).
Measurement of Enzyme Properties-Apparent molecular mass of denatured subunits was determined using SDS-PAGE at pH 7.0 by comparing the relative migration rates of enzyme samples with 40 -200-kDa protein standards (SDS-PAGE high range standards, Bio-Rad) (42,43). Apparent molecular mass of the purified native isoforms was determined by comparing their elution volumes with those of protein molecular weight standards (Pharmacia Kit 17-0441-1) during gel filtration in ESBA on a 1.6 ϫ 76-cm column of Sephacryl S-300 calibrated with blue dextran 2000 (Pharmacia) (V 0 ) and 0.4 M NaCl (V t ) by conductivity measurement (47). Apparent molecular mass of the purified native isoforms was also determined by nondenaturing PAGE from Ferguson plots with 4.5-10% gels by a modification (48) of the method of Bryan (49) using nondenatured, high molecular weight protein standards (Sigma MW-ND-500). The weight-average apparent molecular mass (wt-ave, m) of the native isoforms was calculated from optical densitometry scans of 5% PAGE gels using Equation 1: where OD i indicates the area under the optical density peak of the ith protein band, m i indicates the molecular mass of the ith protein band, and n indicates the number of protein bands with measured optical density peaks. Amino acid analysis and N-terminal sequence analysis of isoforms IA, IB, and III-PII were performed by Cheryl Lea Owens in the Molecular Biology Core Laboratory of Case Western Reserve University. For amino acid analysis samples were hydrolyzed in 6 N HCl, and the phenylisothiocyanate derivatives were analyzed (50) using a model 420A Amino Acid Analysis System (Perkin-Elmer). Results were the mean of duplicate analyses. To determine if the observed differences in amino acid content among the three isoforms were due solely to assay variability, we calculated the CV of the amino acid composition of a single peptide standard in 15 assays performed on 3 separate days. The average same day CV among 18 amino acids was Ϯ15.1% (range: 7.6 -36.3%), the greatest being with Cys (Ϯ36%), Met (Ϯ36%), Ile (Ϯ23%), and Val (Ϯ26%). N-terminal sequence analysis was performed using Edman degradation (51) with a Procise model 494 Protein Sequencer (Perkin-Elmer).
Kinetic Analysis-K m and V max were determined from Lineweaver-Burk plots of assays performed at 37°C in EAB buffered at pH 6.2 using pNP-␣-GalNAc as substrate.
pH activity was determined over the range 3.0 -9.0 in incubation mixtures containing 38 mM citrate-phosphate or Tris-HCl buffer and measuring the final pH of the incubation mixtures after incubation. pH stability was measured by incubating the enzymes in the same buffers for 4 h at 4°C and then dialyzing them overnight in ESBA at pH 6.0 before assaying. Thermal stability was measured by assaying samples that had been incubated at 50 and 55°C in ESBA for 0, 30, 60, and 120 min.
Effect of Added Reagents and Ions-Incubations with additives were performed in EAB buffer, pH 6.0, devoid of Triton X-100, magnesium and EDTA. Tests for EDTA inhibition were carried out by first dialyzing the purified enzyme preparations overnight against 0.1 M Na 2 EDTA buffered at pH 6.0 followed by dialysis against several changes of EAB devoid of Triton X-100, magnesium and EDTA before assaying for residual activity with and without added cations. Incubations with iodoacetamide, N-ethylmaleimide, p-hydroxymercuribenzoate, HgCl 2 , and CuSO 4 were performed at final concentrations of 1 and 10 mM for up to 24 h.
Statistical significance of the differences in amino acid composition and in the effect of ions on the activities of the three isoforms was determined using the Kruskal-Wallace analysis of variance by ranks test (52).
Purification of ␣-GalNAcase-In Step 1, 15-liter batch cultures of R. torques were grown in Medium 75 without glucose that had been prereduced by boiling prior to autoclave sterilization and cooled under a stream of N 2 . Following inoculation, the culture was stirred continuously at 37°C until bacterial growth, monitored by optical absorbance at 660 nm, had entered early stationary phase. Then a sterile solution of glucose in anaerobic Medium 75 was added at a rate of 1 g/h until an additional 4 g of glucose/liter of culture had been added. Incubation was continued for an additional 24 h and then the bacterial cells were removed by filtration through a 0.5-m Durapore filter in a Pellicon cassette (Millipore Corp. Bedford, MA). The cell-free filtrate was then concentrated 10-fold by ultrafiltration through a Pellicon "PT" cassette filter with a 10-kDa cutoff. All subsequent steps were performed at 4°C. For Step 2, fractionation was performed with (NH 4 ) 2 SO 4 . Solid (NH 4 ) 2 SO 4 was added to the concentrated filtrate to a final concentration of 1.25 M. After standing 16 -24 h the small amount of precipitate was removed by centrifugation and discarded. More (NH 4 ) 2 SO 4 was added to a final concentration of 3.5 M with respect to the initial volume. After standing 16 -24 h the precipitate was collected by centrifugation at 10,000 ϫ g and dissolved in and dialyzed against ESBA.
Step 3 was hydrophobic interaction column chromatography at high ionic strength. Solid NaCl was slowly added to the dialyzed (NH 4 ) 2 SO 4 fraction to a final concentration of 4.0 M, and Triton X-100 was added to a final concentration of 0.1% v/v. We added 0.1% v/v Triton X-100 to remove any lipid and lipophilic impurities as mixed micelles because preliminary studies using Sudan Black as a micellar marker dye showed that Triton X-100 remained tightly bound to phenyl-Sepharose throughout gradient elution. The fraction was then applied to a 2.6 ϫ 87-cm column of phenyl-Sepharose which had been pre-equilibrated with 11 column volumes of ESBA, 4.0 M NaCl. Following sample application the column was eluted at 22 ml/h with ESBA, 4.0 M NaCl until the eluate absorbance at 280 nm returned to base line. The column was then eluted with a two-step decreasing linear ionic strength gradient; the first step consisted of 6.5 column volumes of constantly stirred ESBA, 4.0 M NaCl in a mixing chamber to which an equal volume of ESBA, 2.0 M NaCl was added; the second step followed and consisted of 4 column volumes of ESBA, 2.05 M NaCl in the mixing chamber to which an equal volume of ESBA was added. Elution was completed with 2 column volumes of ESBA. Eluate fractions were assayed for enzymatic activities and for NaCl concentration by conductivity measurement.
␣-GalNAcase activity eluted in three peaks which were pooled as separate isoforms (Fig. 1, top). Isoform I eluted at 3.2 3 3.15 M NaCl, isoform II eluted at 2.9 3 2.8 M NaCl along with the single main ␤-galactosidase (␤-gal'ase) peak, and isoform III eluted over the range 1.0 3 0.12 M NaCl at the end of the elution gradients.
Step 4 was chromatography on hydroxyapatite (Fig. 2). Isoform I was applied to a 1.6 ϫ 10-cm column of hydroxyapatite (Bio-Rad HTP) pre-equilibrated in ESBA buffer. Unabsorbed, inactive protein and excess NaCl were removed by eluting the column with ESBA, and then the column was eluted into 10-ml fractions with an increasing phosphate concentration gradient using 30 column volumes of ESBA containing 40 mM NaKHPO 4 , pH 6.0, buffer in the mixing chamber and 30 column volumes of ESBA containing 200 mM NaKHPO 4 , pH 6.0, buffer in the reservoir. ␣-GalNAcase eluted in two successive peaks behind a small peak of ␤-Galase; these were pooled separately and labeled as isoforms IA and IB, respectively (Fig. 2).
Step 5 with isoforms IA and IB (Fig. 3) was gel filtration chromatography, as follows. Peaks IA and IB were further purified by gel filtration over a 1.6 ϫ 76-cm column of Sephacryl S-300 equilibrated in, and eluted with, ESBA. The final active fractions were pooled and concentrated by ultrafiltration on a YM-10 membrane (Amicon Corp., Cambridge, MA). Hydroxyapatite chromatography was performed with isoform III in a similar manner. ␣-GalNAcase in this fraction eluted in a broad peak that partly overlapped with ␤-Galase and smaller amounts of ␤-N-acetylglucosaminidase (␤-GlcNAcase). ␣-GalNAcase fractions with the least amount of ␤-Galase and ␤-GlcNAcase were pooled and applied to a 1.6 ϫ 23-cm column of DEAE-Sephadex that had been pre-equilibrated in ESBA. The column was washed with ESBA, and ␣-GalNAcase was eluted with a gradient consisting of 13 column volumes of ESB buffer containing 0.15 M NaCl in the mixing chamber and an equal volume of ESB containing 0.5 M NaCl in the reservoir. Fractions with peak ␣-GalNAcase activity were pooled, applied to a 1.6 ϫ 76-cm column of Sephacryl S-300 pre-equilibrated with ESBA, and eluted with ESBA. ␣-GalNAcase in fraction III differed from ␣-GalNAcase in IA and IB by displaying abnormal elution behavior from Sephacryl S-300 (Fig. 3); whereas half of the activity eluted as a discrete peak near the void volume along with traces of ␤-GlcNAc'ase, sialidase, and fucosidase (isoform III-PI), elution of the remainder was retarded because of apparent anomalous interaction with the gel bed; this portion eluted over two bed volumes as a long, slowly declining shoulder of the main peak (isoform III-PII). These were pooled separately and concentrated on a YM-10 membrane.

RESULTS
Purification of ␣-GalNAcase Isoforms-The key steps were hydrophobic interaction column chromatography on phenyl-Sepharose followed by column chromatography on hydroxyapatite (Table I). ␣-GalNAcase eluted from phenyl-Sepharose in three separate peaks of increasing hydrophobicity, isoforms I, II, and III (Fig. 1, top). The first and most hydrophilic, isoform I, eluted ahead of other glycosidases (Fig. 1, bottom) and comprised 9.1% of the total ␣-GalNAcase activity recovered in this step. Hydroxyapatite chromatography further resolved isoform I into two peaks, IA and IB, respectively, comprising 43 and 56% of the eluted activity (Fig. 2). The second peak, isoform II, eluted with ␤-gal'ase. When pooled, this fraction comprised only 3% of total recovered activity and was not investigated further. The third and most hydrophobic peak, isoform III, comprised 48% of total recovered activity and required anion exchange chromatography in addition to hydroxyapatite chromatography to separate it from other glycosidases. It displayed anomalous elution behavior during gel filtration chromatography on Sephacryl S-300; one-half, III-PI, eluted near the column void volume together with small amounts of other glycosidase activities, and the remainder, III-PII, eluted over two column volumes in nearly pure form (Fig. 3). ␣-GalNAcase recoveries in IA, IB, III-PI, and III-PII were 0.7, 1.9, 1.8, and 1.7%, respectively, totaling 6.1% of the initial culture filtrate activity purified 26 -121-fold (Table I). They contained traces of other activities that were less than 0.5% of ␣-GalNAcase activity (Table II). None of the ␣-GalNAcase isoforms possessed ␣-galactosidase activity which is present in purified lysosomal ␣-GalNAcase (10, 16 -18, 28, 53). Isoforms IA and IB contained no detectable sialidase or H-degrading ␣-fucosidase activity, but III-PII contained small amounts of both that comprised 0.02 and 0.1% of ␣-GalNAcase activity, respectively, and ␣-fucosidase in this fraction co-migrated with ␣-GalNAcase activity during PAGE. Properties of ␣-GalNAcase Isoforms: Homogeneity and Molecular Size Relationships (Fig. 4, Table III)-On denaturing SDS-PAGE IA and IB were a mixture of two subunit proteins with molecular mass of 158 and 172 kDa. By scanning densitometry IA was comprised of 69% of the 158-kDa band and 31% of the 172-kDa band, and IB was comprised of 97% of the 172-kDa band and 3% of the 158-kDa band. III-PII was a single subunit band at 205 kDa, and III-PI, which contained small amounts of other activities, was comprised of a single band at m ϭ 201 kDa and three other smaller, weakly staining bands. None of the isoform protein bands stained with periodic acid-Schiff stain following SDS-PAGE of 14 -50 g of IA, IB, or III-PII. None of the following pretreatments altered band appearance on SDS-PAGE: incubation at pH 3.9 or at 9.1 immediately prior to electrophoresis at pH 7.0, or heating the samples at 100°C with SDS and MCE by the method of Laemmli (44,45), with and without adding 6 M urea to the cooled samples prior to electrophoresis.
On nondenaturing PAGE the three native isoforms displayed a single major protein band and several weaker, enzymatically active ones which were most prominent in isoform IA. The weight-average molecular mass of native isoforms IA, IB, and III-PII, determined by scanning densitometry, was 135, 229, and 295 kDa, respectively. The molecular mass determined by gel filtration was 265, 417, and 530 kDa, respectively, with a ratio of 1.00:1.57:2.00. A monomer of 133 kDa would satisfy this size relationship within a 4.3% error if it were present as a dimer, a trimer, and a tetramer in IA, IB, and III-PII. The size of such a monomer is virtually identical to the weightaverage molecular mass of native isoform IA (135 kDa).
Amino acid analysis (Table IV) revealed that overall there was no statistically significant difference in the mol % composition of 16 amino acids among the three isoforms, the mean CV being Ϯ19.0% with a range of Ϯ1.9% to Ϯ35%. The variation among six of the amino acids was less than the inherent variation in the assay (CV ϭ Ϯ15.1%, see "Methods"). In isoform IB the mol % composition showed slightly greater differences from IA and III-PII (CV ϭ Ϯ17 and Ϯ19%) than between IA and III-PII (CV ϭ Ϯ15%), but the differences were statistically insignificant.
N-terminal sequence analysis of the first 20 amino acids (Table V) revealed an identical sequence in isoforms IA and III except that Ala in position 6 of isoform III was present as a minor amino acid in position 6 of isoform IA. A putative Nglycosylation sequence was present in positions 3-5. By contrast, the N-terminal sequence in isoform IB differed com-  pletely from that of IA and III. Consistent with the evidence from SDS-PAGE that some IB was present in IA, 13 of 20 minor amino acid peaks encountered during sequencing of IA were identical to those of isoform IB. A BLAST search (54) of the protein data base of the National Center for Biotechnology Information, the National Library of Medicine, revealed homologies in the N-terminal sequence of isoform III with the Nterminal sequence of sialidase isoforms of Bacteroides fragilis SBT3182 (55), another human commensal enteric bacterium. No matches were found with isoform IB in the BLAST search.
Substrate Specificity: Mucin Glycoproteins (Fig. 5)-␣-Gal-NAcase activity in IA, IB, and III rapidly cleaved GalNAc from human blood group AB (Fig. 5A) and group A (Fig. 5B) mucin glycoprotein subunits with a decrease in blood group A antigen titer, a rise in H antigen titer, and no change in B antigen titer. TLC of the saccharide cleavage products revealed a single spot corresponding to GalNAc in each digestion mixture.
Range of pH Activity and pH Stability (Fig. 6, Table VII)-Isoforms IA and IB were active over the range 4.7-8.5. They     had identical pH activity curves, with optimum pH ϭ 5.8 -6.8 and 70 -80% of maximum activity present at pH 7.4. By contrast, in two separate studies the pH activity curve of III-PII was shifted 0.5 pH units more acidic; it was active over the range 4.5 to 8.0, and its pH optimum was 5.6 -6.1, with 36% of maximum activity present at pH 7.4. All three isoforms were stable for 4 h at 4°C over the pH range 5.0 -9.5 but were unstable below pH 4.5. Isoform III-PII was slightly more heat resistant than IA or IB (Table VII). At 55°C the activity of all three isoforms was rapidly lost, but at 50°C 47% of III-PII activity remained after 120 min compared with Ͻ10% of IA and IB. Thermal stability was unaltered by addition of 1 mg/ml bovine serum albumin.
Effect of Added Reagents and Ions (Table VIII)-Addition of 0.1% v/v Triton X-100 to the purified isoforms did not increase their activity as it did with cruder preparations (data not shown). Treatment with 0.1 M EDTA inhibited activity of IB by 95-96%, and the activity of IA by 46% but had no inhibitory effect on III-PII. In a separate study, SDS-PAGE of IB after EDTA treatment revealed no change in protein band number, appearance, or migration rate. Neither 10 mM dithiothreitol nor 10 mM MCE affected activity. Treatment with 10 mM iodoacetamide, N-ethylmaleimide, or p-hydroxymercuribenzoate inhibited isoform III-PII more than IA and IB, and isoform III-PII was also inhibited more by HgCl 2 and CuSO 4 . By contrast, addition of 1 mM MgCl 2 , MnCl 2 , NiCl 2 , ZnCl 2 , CaCl 2 , Co(NO 3 ) 2 , FeSO 4 , or ferric ammonium citrate nonspecifically enhanced the activity of all three isoforms in the order III-PII Ͼ IB Ͼ IA, the differences being statistically significant (p Ͻ 0.05). Stability During Storage-The activity of each isoform was stable over 25-31 months' storage at 4°C and pH 6 in ESBA buffer at protein concentrations ranging from 66 to 450 g/ml, and repeat SDS-PAGE after 24 months showed no change in protein pattern or migration. Freezing, thawing, and lyophili-TABLE V N-terminal amino acid sequences in isoforms IA, IB, and III The top row of IA and IB displays the most probable amino acids in the sequence, and the second row displays other amino acids detected at each position, with an X indicating an indeterminate residue. Bold lettering indicates amino acids in positions identical with isoform III whose sequence was unambiguous. The N-terminal sequence of sialidase isoforms of B. fragilis SBT3182 is aligned with those of isoform III to show positions of identity (double dots) and of functionally similar amino acids (single dots).   (57), is one of a subset of Ruminococcus and Bifidobacterium strains averaging 1% of human fecal bacteria (29,34,58) that is functionally defined as the major producer of extracellular glycosidases degrading ABH and Lewis blood group active oligosaccharide chains of mucin glycoproteins and cell membrane glycolipids (31, 33, 58 -60). The mono-and disaccharides resulting from mucin degradation by their exo-and endoglycosidases are utilized by larger fecal populations that do not degrade these chains (59,61). Of the five strains characterized to date only the two R. torques strains produce strong blood group A-degrading ␣-GalNAcase activity (58 -60). In probable adaptation to the enteric environment, the extracellular ␣-GalNAcase and ␤-galactosidase of R. torques resist degradation by pancreatic proteases (61). ␣-Gal-NAcase in cell-free culture supernatants of R. torques is produced constitutively in the range of 70 -190 units/liter and is free of proteinases, hemolysins, or toxins (58). However, as with purification of extracellular glycosidases produced by other Gram-positive bacteria (5, 62-65), separation of ␣-Gal-NAcase from other glycosidases resisted other attempts until we employed hydrophobic interaction chromatography at high ionic strength, resulting in selective elution of isoforms IA and IB free of sialidase, ␤-GlcNAc'ase, and ␣-L-fucosidase. Hydroxyapatite chromatography was a useful next step since glycosidases in the undialyzed eluate fractions adhered to it despite high NaCl concentrations.
Extracellular ␣-GalNAcase produced by R. torques exists in at least three isoforms differing in hydrophobicity, size, and enzymatic properties. Isoform III was the largest and most hydrophobic isoform, accounting for half of the total activity recovered from phenyl-Sepharose. Isoform III resembled isoform IA in having the same N-terminal amino acid sequence, but it differed from IA and IB in having a more acidic pH optimum, greater thermal stability, greater sensitivity to metal ions and alkylating agents, and anomalous retardation during gel filtration on Sephacryl S-300. Anomalous retardation of ␤-GlcNAc'ase on derivitized dextran gels has also been observed by us 2 and by others (66 -68); it has been ascribed to hydrophobic interactions as well as to charge effects (69).
Isoform I, the least hydrophobic form, separated during hydroxyapatite chromatography into two isoforms, IA and IB, which differed in molecular size, N-terminal amino acid sequence, and sensitivity to EDTA. The larger, IB, had more Asp and less Ala than IA. It was inactivated by treatment with EDTA without alteration of the subunit size on SDS-PAGE, and activity was not restored by addition of eight metal cations, alone or in combination. This suggests that one or more chelatable metal cations is an intrinsic part of the tertiary structure of IB whose removal results in irreversible changes in catalysis or conformation (70 -72).
The subunit size of isoforms IA, IB, and III, 160 -200 kDa, is similar to the 180 -355-kDa subunit sizes reported for purified extracellular glycosidases of Streptococcus pneumoniae and C. perfringens (12,73,74). Such large subunit sizes may be a property of extracellular glycosidases produced by Gram-positive bacteria in contrast to ␣-GalNAcases from other sources (10, 16 -18, 21).
Despite differences in isoform properties, the apparent molecular mass of the native isoforms determined by gel filtration bore a ratio of 1.96:3.09:3.93 to the 135-kDa weight-average apparent molecular mass of native isoform IA, suggesting that a segment of approximately 135 kDa was present as a dimer, trimer, and tetramer in native isoforms IA, IB, and III, respectively. Because the isoforms differ in properties and resistance to disulfide bond cleavage, it is unlikely that noncovalent aggregation or linkage via disulfide bonds account for the putative multimeric structure. A model that could account for the observations is one in which the three respective isoforms consist of 2, 3, and 4 tandem repeats of a 135-kDa segment containing an active site sequence and linked by shorter intervening sequences. Tandem repeats were recently reported by  6. pH activity curves of isoforms IA, IB, and III-PII with pnp-␣-GalNAc substrate, expressed as percent of maximum activity. Cumulative values of two studies. In both studies the pH activity curve of isoform III-PII was 0.5 pH unit more acidic. Isoform IA, q; IB: f, III-PII; Ⅺ.
Clarke and coworkers (74) in the 144-kDa subunit of S. pneumoniae ␤-GlcNAc'ase (74). This contained two tandem repeats of a 335-amino acid sequence which itself contained a highly conserved 30-amino acid sequence that was likely to be part of the active enzymatic site. The model proposed for the ␣-Gal-NAcase isoforms would need to account for the different Nterminal sequence and EDTA sensitivity of isoform IB as well as its slightly different amino acid composition. The overall similarities of isoform IB with the other isoforms suggests that it is not the product of a separate gene. A stronger possibility is that the isoforms are derived by post-translational proteolytic processing of a single gene precursor polypeptide in which the dimeric and tetrameric isoforms IA and III are cleaved by the same linkage-specific endopeptidase, whereas the trimeric isoform IB is cleaved at a different site by a second endopeptidase, leading to its different N-terminal sequence and properties. This is analogous to the observation (75) that the heterodimeric ␣ and ␤ subunits of Escherichia coli acylase arise by separate, highly specific cleavages from a single large precursor polypeptide during post-translational processing.
The observation that the N-terminal amino acid sequence of ␣-GalNAcase isoforms IA and III of R. torques shares homologies with the N-terminal sequence of sialidase isoforms of B. fragilis is of biological interest since both species share the same human enteric habitat at roughly similar fecal population densities (76), yet differ widely in phenotype and phylogeny. The possibility of a common origin during evolution of these microbial hydrolases is an intriguing question that may be answered when their complete structures are determined and compared.