Scope and Mechanism of Carbohydrase Action STEREOCOMPLEMENTARY HYDROLYTTC AND GLUCOSYL-TRANSFERRING ACTIONS OF GLUCOAMYLASE AND GLUCODEXTRANASE WITH a- AND ~-D-GLUCOSYL FLUORIDE*

Rhizopus niveus glucoamylase and Arthrotxwter globiformis glucodextranase, which catalyze the hydrolysis of starch and dextrans, respectively, to form D-glUCOSe of inverted (p) configuration, were found to convert both a- and @-D-glucosyl fluoride to B-D-glucose and hydrogen fluoride. Each enzyme directly hydrolyzes a-D-glUcOSyl fluoride but utilizes the @-anomer in reactions that require 2 molecules of substrate and yield glucosyl transfer products which are then rapidly hy- drolyzed to form 8-D-glucose. Various D-glUCOpYraYIO-syl compounds serve as acceptors for such reactions. Mixtures a-isomaltoside, J1.2 J1.2 P-isomaltoside, ppm, J1,2

occur without glycosidic bond cleavage has emerged as an effective means of gaining deeper understanding of the catalytic capabilities and mechanism of carbohydrases (1-8). For example, fresh insight was recently obtained into the functioning of P-amylase through the demonstration (7) that this classic exoglucanase utilizes both a-and B-maltosyl fluoride as substrates in reactions leading to the formation of p-maltose and hydrogen fluoride. &Amylase was found to hydrolyze a-maltosyl fluoride directly, and evidence was obtained suggesting that it catalyzes maltosyl transfer from P-mdtosyl fluoride to a second substrate molecule to form an a-lP-linked higher saccharide, which is then rapidly hydrolyzed to yield ,&maltose. Although the transfer product was not directly observed, the findings strongly suggested that ,B-amylase has a second (nonhydrolytic) mode of action in addition to its long known hydrolytic action and that the functional groups at the active site of the enzyme alternate in their catalytic roles to promote reactions which are not directly related by microscopic reversiblility. Indeed, Hehre et al. (7) envision such functional flexibility as an attribute of the catalytic groups of glycosylases in general.
We find now that the ability to utilize both a-and panomeric forms of a substrate is not confined to p-amylase but is shared by two D-glucosyl mobilizing exo-a-glucanases, the glucoamylase of Rhizopus niveus and the glucodextranase of Arthrobacter globiformis. These enzymes catalyze reactions with a-and /i-D-glucopyranosyl fluoride that are complementary to each other in the same sense as observed ( 7 ) with &amylase acting on a-and P-maltosyl fluoride. Present findings with glucoamylase and glucodextranase, however, go beyond those obtained with j?-amylase in an important respect. Glucosyl transfer products have been recovered and characterized in each case, thus providing unequivocal evir dence of the ability of inverting exoglycanases to catalyze glycosylation reactions beyond hydrolysis and reversal (condensation) reactions, R. niueus glucoamylase is representative of the extensively studied fungal glucoamylases which hydrolyze the terminal (u-1,4-and a-1,6-D-glucosidic linkages of starch and related glucosaccharides (9-11) to produce ~-D-glUCOse (12)(13)(14)(15) and which similarly hydrolyze a-D-glucosyl fluoride (16)(17)(18). The R. niueus enzyme also catalyzes the condensation of glucose to form a-1,4-and u-1,6-linked glucosaccharides (19,20), using fl-D-ghCOSe as the specific donor (20). Aside from catalyzing condensations, however, glucoamylases have been considered unable to effect glucosyl transfer reactions (11, [21][22][23][24][25][26]. Likewise, A. globiformis glucodextranase (27,28), which hydrolyzes the terminal a-1,6-and a-l,4-~-glucosidic linkages of dextrans and related glucosaccharides to form p-D-glUCOSe, has shown no sign of glucosyl-transferring activity when acting on dextran (27).

6017
Based on the stereospecificity of the reactions catalyzed with a-and /?-D-glucosyl fluoride, a detailed description is presented of the mechanism whereby these exoglucanases effect hydrolytic or glucosyl transfer reactions, respectively, with substrates of a-or /?-form. The significance of the alignment of the present results with the concept that glycoside hydrolases and glycosyltransferases are interrelated glycosylases (catalysts of the interchange of a glycosyl residue and a proton) with functionally flexible catalytic groups is discussed.

EXPERIMENTAL PROCEDURES
Enzymes-Twice recrystallized, a-amylase-free glucoamylase from R. niueus  [a]u22.5' +96.7' (water)): gave -95% of the theoretical amounts of fluoride and D-glucose on hydrolysis by 0.02 N sulfuric acid (100 "C, 10 min) but was accompanied by -2.5% of free fluoride. To remove the latter, 400 mg of the a-D-glUCOSyl fluoride were chromatographed on a column (1.6 X 30 cm) of dry Silica Gel 60 (E. Merck), using absolute ethano1:ethyl acetate (2:5) as solvent. The pure product (217 mg), crystallized from methanol, had ~0 . 1 % free fluoride ion. "F NMR spectra, recorded in 0.1 M acetate-d4/D20 buffer of pD 5.3, showed a chemical shift of -152.06 ppm (relative to external trichlorofluoromethane in acetone-&) and JI,F 53.4 Hz, J~. F 26.8 Hz, in agreement with the values reported by Hall et al. (33). 'H NMR spectra are described under "Results." /I-D-Glucopyranosyl Fluoride-Tetra-0-acetyl P-D-glucopyranosyl fluoride was synthesized from tetra-~-acetyl-a-D-g~ucopyranosyl bromide (Pierce Chemical Co.) and silver fluoride (Ventron Corp., Danvers, MA) by an established method (31, [34][35][36] Pure amorphous p-D-glucopyranosyl fluoride was prepared by deacetylating 0.3-g (860-pol) samples of the tetraacetate with 3.0 ml of 0.015 M sodium methoxide in dry methanol (0 "C, 3 h). On completion of the reaction, the mixture was dried upon 1.5 g of Silica Gel 60 under vacuum at 25 "C, then added to a dry packed column of the same absorbant and developed with absolute ethanolethyl acetate  Stock solutions Of p-D-glUCOSyl fluoride in methanol, kept at -20 "C and protected from moisture, remained stable for weeks. As required, the compound was obtained by drying measured volumes of such solutions in a Rotovap apparatus (28 "C) immediately before use. The compound is labile in aqueous solutions; the rate of spontaneous release of fluoride ion in 0.05 M acetate buffers between pH 4.5 and 5.6 was found to be O.2%/min at 30 "C and O.O13%/min at 0 "C.
Methyl-a-D-glucopyranoside-Commercial preparations were found to contain small amounts of glucose and of an impurity migrating on paper chromatograms as methyl-a-isomaltoside (RglC 0.78). Preparations free from the latter impurity were obtained by chromatography on a column (2.8 X 15 cm) comprising a mixture of 50 g each of Darco GGO-activated carbon and Celite 535 (Johns-Manville Corp.). The column was washed with water and charged with 10 g of methyl-a-D-glucopyranoside (Pfanstiehl) dissolved in 30 ml of water. Development was with water, with fractions assayed for total sugar (37) and for purity (by paper chromatography, 3.5-mg samples). The final product (4.8 g) contained a trace of glucose as the only impurity visible chromatographically (0.3% by glucose oxidase assay). The absence of contamination by methyl-a-maltoside, which migrates at the same rate as glucose, was shown by the failure of material eluted at the glucose position to yield methyl-a-D-glucoside (Rglc 1.35) after treatment with glucoamylase.
Methyl-a-~-[U-"C]glucopyranoside-A stock solution in methanol was prepared from methyl-a-~-['~C]glucoside (184 mCi/mmol, purified by paper chromatography; Amersham Corp.) mixed with sufficient purified, unlabeled methyl-a-D-glucoside to contain 25 pmol of glucoside/ml and 0.2 pCi/pnol of glucoside. Purity of the label was checked by chromatographing a 1.2-pCi sample and eluting materials corresponding to methyl-a-D-glucoside (RRlc 1.35), glucose or methyla-maltoside (RgIc LO), and methyl-a-isomaltoside (& 0.78). Measurements of these eluates by scintillation counting after rechromatography indicated the presence of radioactive impurities amounting to 0.015% of total counts in the R,I, 1.0 region and 0.03% in the R,I, 0.78 region. For experimental use, required amounts of the I4C-labeled methyl-a-D-glucoside were obtained by evaporating known volumes of the stock solution under vacuum.
Additional Carbohydrates-Pure P-maltosyl fluoride was synthesized as recently described (7). D-Glucose, maltose, and l-glyceryl-a-D-glucopyranoside (41) were laboratory preparations, free from accompanying oligosaccharides. Isomaltose was a gift of Dr. Allene Jeanes. Other carbohydrates were of reagent grade.
General Methods-Thin layer chromatography was carried out with Silica Gel G according to Stahl (Brinkmann Instruments), with ethyl ether:petroleum ether (32) as the solvent for acylated compounds and absolute ethanolethyl acetate (25) for nonacylated compounds. Spots were visualized by the sulfuric acid-char method. Paper chromatography was performed in the descending manner with Whatman No. 1 or 3MM paper and I-butano1pyridine:water (643) as developer. Staining was by a silver nitrate dipping technique (42), with papers hung in the air for 12-15 min after application of the sodium hydroxide reagent.
Concentrations were carried out under diminished pressure in rotary evaporators at or below 32 "C unless otherwise noted. Drying was done in a vacuum oven at or below 35 "C. Melting points were determined on a Mel-Temp block (Laboratory Devices, Cambridge, MA) and are uncorrected. Optical rotations were measured using a Rudolph and Sons Model 70 polarimeter and 2-dm tubes. Total carbohydrate (as D-glucose) was determined by the phenolsulfuric acid method (37). Elemental analyses were performed by Galbraith Laboratories, Inc., Knoxville, TN.
Free D-glUCOSe was determined using the Somogyi and Nelson reagents (43,44) or, when specified, by the glucose oxidase method (special Glucostat reagent and chromogen, Worthington Biochemicals); both procedures were concurrently standardized with equilibrated D-glucose. Fluoride anion concentrations, in the presence or absence of a-or P-D-glucosyl fluoride, were measured in the presence of TISAB buffer (1 M sodium acetate buffer, pH 5.2, 1 M sodium chloride, 0.4% 1,4-cyclohexane bis(dinitril0tetraacetic acid) monohydrate) with a specific fluoride ion probe (Orion specific ion meter Model 407A and combination fluoride electrode Model 96-09) as previously described (6, 7).
Radioactivity Meas~rernents-~~C-labeled compounds were measured with a Beckman LS 230 scintillatioon counter. Test samples (0.1 ml) in water were vigorously shaken with 4.8 ml of Aquasol (New England Nuclear) in plastic minivials; all counts per minute values were corrected for background. Counting efficiency was 95-98% of theoretical.
NMR Spectra-"H and "F NMR spectra recorded at 100 MHz and 94 MHz, respectively, were obtained using a Jeol PFT-100 spectrometer interfaced with a Nicolet 1080 Series computer. The instrument was operated in the pulse Fourier transform mode. 'H NMR spectra were also recorded at 220 MHz on a Varian HR-220 spectrometer equipped with a Nicolet 1083 computer. "F spectra were recorded in acetone-de, or in acetate-&/DzO buffer of pD 5.3, with trichlorofluoromethane in acetone-& as reference. 'H NMR spectra were recorded in 0.1 M acetate-dd/DzO buffer of pD 5.3 or in deuterium oxide (99.7 atom % deuterium, Merck, Sharp, and Dohme) at ambient temperature unless otherwise noted; chemical shift measurements were made with respect to 3-(trimethylsily1)propanesulfonic acid sodium salt as an internal standard.
Column Chromatography-Columns (17 X 0.95 cm in diameter) comprising equal parts by weight of activated carbon (Darco G60) and Celite 535 were used to recover methyl-a-maltoside from mixtures containing a large amount of D-glucose. The water-washed column was charged with 20-100 mg of methyl-a-maltoside/D-glucose mixture. Elution at the rate of 0.5 ml/min was first carried out with water, with 10-ml fractions collected and monitored for reducing sugar (as D-glucose) (43,44). When the sugar content fell below 1 pg/ml, elution with 10% ethanol was begun and 5-ml fractions were collected. Monitoring of 14C-labeled methyl-a-maltoside in these fractions was done by scintillation counting; the peak, usually in fractions 5-8, provided -75% recovery of the glycoside.

RESULTS
In preliminary experiments, the glucoamylase of R. niveus and the glucodextranase of A. globiformis were found to catalyze the release of glucose (observed chromatographically) and of fluoride ion (detected with a fluoride ion electrode) from both a-and p-D-glUCOSyl fluoride. T o learn whether the reactions with the a-and p-anomer are distinguishable kinetically, the relation between rate of fluoride release and substrate concentration was examined. Two series of 0.60-ml digests at pH 5.6 (one for each anomer) were prepared for each enzyme. Glucoamylase digests comprised 4.8 to 40 mM a-D-glucosyl fluoride and 1 p g / d of enzyme; 5 to 40 mM p-D-glucosyl fluoride and 0.5 mg/ml of enzyme. Glucodextranase digests contained 3.6 t o 36 mM a-D-glUCOSY1 fluoride and 5.6 pg/ml of enzyme; 0.6 t o 6 mM p-D-glUCOSyl fluoride and 0.17 mg/ml of enzyme. Individual digests and substrate/buffer controls were set up at 2-min intervals, incubated at 30 "C for 26 min (Cr-D-glUCOSYl fluoride series) or for 12 min (p-D-glu-cosy1 fluoride series), then analyzed for fluoride ion concentration. Substrate utilization was below 10% in all instances. Specific initial rates of enzymically catalyzed fluoride release were calculated after correction for spontaneous hydrolysis of a-or p-D-glUCOSyl fluoride (0.5 and 3.5%, respectively) in the controls incubated without enzyme.
The results ( Fig. 1) show that, with both enzymes, a great difference exists between the actions on the two anomers.  (7), for 2 molecules of substrate to be bound to enzyme before reaction can occur. Glucodextranase is substantially more active than glucoamylase in catalyzing the release of fluoride from /3-Dglucosyl fluoride. At low substrate concentrations, both enzymes act more rapidly on the a-then on the p-anomer, e.g. at 6 mM, a-D-glUCOSY1 fluoride was attacked 2700 times faster than j3-D-glucosyl fluoride by glucoamylase (11 times faster by glucodextranase) .
As a control on the unusual kinetic findings with the panomer, initial rates of fluoride release were determined for a series of digests comprising 6 to 60 mM /3-D-glUCOSYl fluoride and 1 pg/ml of purified sweet almond P-glucosidase (assaying 40 units/mg with salicin) incubated (pH 5.6) at 30 "C for 12 min. In this case, a plot of u versus S was hyperbolic and that of u" versus s" was linear, allowing calculation of K , 53 mM and V,,, 263 pmol of F"/min/mg. Since the range of fluoride ion concentrations in the P-glucosidase digests was comparable to that found with the exo-a-glucanases, it is evident that the unusual kinetic order found with the latter (Fig. 1, c and  d ) is specific for glucoamylase and glucodextranase and is not an artifact attributable to an accelerated breakdown of p-Dglucosyl fluoride on exposure to higher concentrations of hydrogen fluoride.

Enhancement of Enzymic Utilization ofp-D -Glucosyl Fluoride by D -Glucopyranosyl Compounds-
The possibility that the reactions catalyzed by glucoamylase and glucodextranase with /?-D-glucosyl fluoride might require 2 molecules of substrate, one functioning as a glucosyl donor and the other as a glucosyl acceptor, was examined by testing the ability of various glycosides and sugars to increase the rate of these reactions by serving as supplementary acceptors.
Individual test mixtures (0.60 m l , pH 5.2) containing 11 mM p-D-glUCOSyl fluoride, 1 mg/ml of glucoamylase or buffer, and 25 or 100 mM potential acceptor (or containing 6 m~ p-Dglucosyl fluoride, 0.25 mg/ml of glucodextranase or buffer, and 25 or 100 mM potential acceptor) were set up at 2-min intervals. After incubation (30 "C, 14 min) each mixture was treated with 1.80 ml of TISAB buffer, and the fluoride ion concentration was immediately measured. Since all tests were made with freshly prepared solutions of /3-D-glUCOSyl fluoride (kept at 0 "C less than 15 min) and since experimental runs were limited to six test mixtures and two controls (p-D-glu-cosy1 fluoride plus enzyme; p-D-glUCOSyl fluoride plus buffer), substrate degradation was kept small (3.8%) and uniform throughout. Control values from one experimental run to the next were in excellent agreement, within 3% of one another.
As illustrated in Fig. 2, a variety of compounds having an unsubstituted D-glucopyranosyl residue acted to increase the rate of C-F bond cleavage of /3-D-glUCOSyl fluoride by both enzymes. In contrast, methyl-a-D-mannoside and thep-nitrophenyl-D-galactosides tested depressed the rate for both enzymes. With glucoamylase, a 2-to 5-fold rate enhancement was observed with 12 of the 13 D-glucopyranosyl compounds examined; with glucodextranase, a 1.4-to 4.5-fold rate enhancement was found with 10 of the 13. None of the compounds tested had any effect on the release of fluoride from p-D-glUCOSyl fluoride in the absence of enzyme. Although the data do not permit precise comparison of the effects of individual compounds for the two enzymes, large differences were noted in several cases. p-Nitrophenyl a-D-glucoside and sucrose had a large potentiating effect for glucoamylase but none for glucodextranase; p-nitrophenyl P-D-glucoside en- hanced p-D-glUCOSYl fluoride utilization by glucoamylase but strongly inhibited the reaction catalyzed by glucodextranase. On the other hand, maltose and (to a lesser extent) D-glucose showed larger effects with glucodextranase than with glucoamylase. Overall, the results strongly suggest that a-or p-Dghcopyranosyl compounds of various kinds serve as D-glUC0syl acceptor substrates for both glucoamylase and glucodextranase when these enzymes act upon p-D-glucosyl fluoride.

Configuration of the Glucose Formed from a-and p-D -
Gbcosyl Fluoride by the Actions of Gbcoamylase and G bcodextranase-To further characterize the reactions catalyzed by the inverting exo-a-glucanases with a-and P-Dglucosyl fluoride, a study was made of the configuration of the glucose produced in each case, using 'H NMR spectroscopy.
Reference spectra showed that the resonances of the anomeric protons of a-and P-D-glucose are readily distinguished from the proton signals of either a-or /?-D-gluCOSyl fluoride. 'H NMR spectra of the latter compounds at 100 and at 220 MHz, reported for the first time, showed each to be anomerically pure although the H-1 resonances have an unusual appearance in each case. At 100 MHz, the anomeric proton of a-D-glucosyl fluoride (Fig. 3A) appears as a doublet of doublets centered at 5.69 ppm with J1,2 -3 Hz and J 1 ,~ 53.1 Hz; the doublet at 5.95 ppm, however, shows a slightly smaller J1.2 coupling constant than that at 5.42 ppm. The anomeric proton of P-D-glucOsyl fluoride at 100 MHz (Fig. 3B) gives rise to a doublet of doublets centered at 5.23 ppm with 51.2 -7 Hz and JI,F 53.1 Hz; here, the multiplet at 5.49 ppm appears to be three resonances while that at 4.96 ppm is the expected doublet with J1,2 6.9 Hz. These anomalous differences in the J splitting patterns are eliminated at a higher field. Thus, at 220 MHz, the anomeric proton of the a-anomer (Fig. 3C)  Each mixture was transferred to a 5-mm NMR tube, and 'H NMR spectra were recorded at intervals during incubation at 21 "C. Each spectrum consisted of 64 free induction decays, using 3-s repetition times. The data were Fourier transformed.
By using methyl-cu-n-['*C]glucoside (0.05 &i/pmol) as acceptor, the amounts of these presumed transfer products formed in different digests could be accurately measured by scintillation counting of eluates from chromatograms of test and appropriate control mixtures.* Data gathered in this way allowed the definition of digest conditions best suited for isolating these minor reaction products.
The conditions finally used with each exoglucanase, together with the transfer product yields, are given in Table I. For product recovery, chromatograms of the digests were sectioned at the level of glucose/methyl-cy-maltoside (R,I, 1.0; -22-27 cm) and of methyl-cu-isomaltoside (R&I, 0.78; -17-22 cm) and eluted with methanol. The Rglc 1.0 material from the glucoamylase digests (and, separately, that from the glucodextranase digests) was dried and weighed; in each case, the radioactive component, presumed to be methyl-cy-mahoside, was then separated from the main reaction product (unlabeled glucose) by carbon-celite column chromatography (see "Experimental Procedures" on No. 1 paper. As indicated in Table I, calculations based on the measured radioactivity (assuming each product to contain a single glucose ['4C] residue) show that the glucoamylase digest provided 5.33 pmol of R,I, 1.0 product and 2.18 pmol of R,I, 0.78 product. The order of yields was reversed for the glucodextranase digest, from which 16.46 pmol of R,,, 0.78 product and 1.08 pmol of that having R,I, 1.0 were obtained. Since only about one-half of the available p-n-glucosyl fluoride was utilized in each digest, the two glucoamylase products represent -2% of the substrate utilized; the glucodextranase products represent -16%.

Characterization of the Transfer
Products-The separated radioactive products were examined by two methods. First, their 'H NMR spectra were recorded in D20 at 25 and 50 "C, in comparison with those of methyl-o-and /3-maltoside and of methyl-a-and P-isomaltoside. Spectra of the R,I, 1.0 isolate from either the glucoamylase or glucodextranase digest were found to be indistinguishable from those of methyl-a-maltoside. The downfield region of each had the appearance shown in Fig. 5 (A and B) for the glucoamylase R,,, 1.0 product. The doublet at 5.38 ppm, J1.2 -3 Hz is assignable to the equatorial anomeric proton of an unsubstituted cu-1,4-linked (or a-1,3linked) D-&COpyMnOSyl residue. Were this residue (u-1,2-or cY-1,6-linked, the doublet would have appeared at higher field, as reported for cr-kojibiose or a-isomaltose (45,46) and found for methyl-ol-isomaltoside in the present study. If the residue was of /?-configuration, the anomeric proton would not only have resonated at higher field but shown a larger coupling constant (J1,2 7-8 Hz) as in sophorose, laminaribiose, cellobiose, gentiobiose, and their methylglycosides (45,46). Resonance of the anomeric portion of the o-D-g)UCOpyDUIOSyl moiety linked to the methyl group appears as a doublet at 4.81 ppm, J1,2 -3 Hz, in spectra recorded at 50 "C (Fig. 5B); it is hidden by the HOD signal in spectra recorded at 25 "C. The methyl group resonance, present as a sharp singlet at 3.48 ppm, is not &&rated.
These findings are consistent with the methyl-n-maltoside structure of R,I, 1.0 glycoside synthesized by glucoamylase and by glucodextranase from p-D-glUCOSy1 fluoride and methyl-a-D-glucoside; only one other structure, methyl-a-nigeroside, could account for the 'H NMR fmdings. 'H NMR spectra of the R,,, 0.78 isolate from the glucodextranase digest and of that from the glucoamylase digest were essentially indistinguishable from spectra recorded with methyl-cY-isomaltoside.
Materials migrating at the rates of methyl-a-D-glucoside (R,,, 1.37) and of the product under test (&I, 1.0 or 0.78) were eluted, and the radioactivity of each was measured. The counts in the two eluates in every chromatogram accounted for the total radioactivity (-7600 cpm). As shown in Table II, the two 14C-labeled R,,, 1.0 products were completely hydrolyzed to methyl-a-D-['4C]glucoside and unlabeled glucose by glucoamylase.
Thus, whereas all counts were at R,I, 1.0 in chromatograms of the products in buffer, >99% of the radioactivity appeared at R,,, 1.37 following incubation with glucoamylase. Eluates at RRlc 1.0 were essentially devoid of radioactivity, showing that the nom-educing terminal glucose moiety of each product was derived from the  "Authentic methyl-a-maltoside and methyl-o-isomaltoside migrate at R,I, 1.0 and at R,I, 0.78, respectively, in the chromatographic system employed.
n Based on counts per minute of material eluted at R,)' 1.0 and R,,, 0.78, corrected for traces of radioactivity in eluates from chromatograms of incubated methyl-a-n-['4C]glucoside/buffer control mixtures.
b Incomplete hydrolysis appears due to the presence of an impurity of unknown composition, comprising 16% of the radioactivity of this isolate (see Table I glucopyranoside. An cY-1,3-interglucosidic linkage is improbable in view of the limited ability of glucoamylase to hydrolyze (23,47) or form (18) such linkage.
Of the R,,, 0.78 products, the one from the digests with glucoamylase was hydrolyzed by glucodextranase with the formation of methyl-cy-D-[i4C]glucoside having 82% of the radioactivity of the unhydrolyzed material, 18% was resistant and remained localized in the R,l, 0.78 region (it represents an impurity as noted in Tables I and II Suitable methyl-cu-D-["'C]glucoside/buffer controls were included. After incubation, 50 ~1 of each mixture was chromatographed; material at the levels of methyl-o-Dglucoside, methyl-cu-maltoside, and methyl-a-isomaltoside were eluted and the radioactivity was measured. The findings (Table III)   digests with fl-D-glucosyl fluoride at any time during incubation) showed only a small fraction of the products found in the digest with p-D-glucosyl fluoride. Thus, these products are primarily formed by direct glucosyl transfer from p-D-glucosyl fluoride to methyl-a-D-['4C]glucoside with inversion ofconfiguration. DISCUSSION The ability of a carbohydrase to utilize both a-and panomeric forms of the same compound as glycosyl substrates, initially demonstrated with P-amylase acting on a-and flmaltosyl fluoride (7), is shown to be an attribute of glucoamylase and glucodextranase as well. Study of the reactions catalyzed with a-and p-D-ghCOSy1 fluoride by these two "inverting exo-a-glucanases" reveals that each is the catalyst of two stereochemically complementary types of glycosylation reactions. Reactions with a-glucans, a-glucosaccharides, is consistent with the specificity of this enzyme which effects rapid maltose synthesis and slower isomaltose synthesis from P-D-ghCOSe (21) and which preferentially hydrolyzes a-1,4-glucosidic linkages but also slowly cleaves a-l,&linkages (22,23,47,48). Similarly, recovery of mainly the a-isomaltoside (16.5 pmol), plus a little amaltoside (1.1 pmol), from the glucodextranase digests is consistent with this enzyme's preferential hydrolysis of a-1,6glucosidic linkages and slow cleavage of a-1,4-linkages (28). Various control experiments show that the isolated glycosides are products of reactions catalyzed by the glucoamylase and glucodextranase themselves and are not formed by contaminating enzymes or by nonenzymatic reactions. It is also evident (Table III) that they arise directly from ,8-D-glucosyl fluoride as glucosyl donor and not indirectly, by way of condensation reactions, from glucose. 3 The presently observed actions of glucoamylase and glucodextranase on a-and p-D-ghCOSy1 fluoride cannot be accounted for by reaction models that depict these enzymes as " Although 'H NMR spectra which record the conversion of P-Dglucosyl fluoride to P-D-glucose by glucoamylase and glucodextranase (Fig. 4C)  catalyzing only the hydrolysis of glycosidicalIy linked substrates. According to one such model (25,26), Rhizopus glucoamylase binds the glucose residues of maltosaccharide substrates in individual subsites with the nonreducing end glucose residue bound at subsite 1 and with the enzyme's catalytic groups located between subsites 1 and 2. Using the dependence of Michaelis-Menten rate parameters on the chain length of substrates, the authors conclude that subsite 1 has apparent zero affinity for binding a glucose residue and also that there is a negligible probability of simultaneous binding of more than one substrate molecule at the active site. The first conclusion is difficult to reconcile with the observed high rate of hydrolysis of a-D-glucosyl fluoride, Vmax 55.6 pmol/min/mg (Fig. 1B). Based on M, = 58,000 (minimum) for R. niueus glucoamy1ase,4 this represents a substantially higher turnover number (53.7 s-l) than reported (25) for the hydrolysis of maltose ( K0 4.6 s-l) or higher maltosaccharides (ko 23-33 s-l) catalyzed by the comparable glucoamylase of Rhizopus delemar. It seems possible that the observed rate of hydrolysis of any maltosaccharide by the enzyme may in part measure the rate of departure of that part of the substrate which remains bound (unproductively) after glycosylic bond cleavage has occurred. With a-D-glucosyl fluoride, no such bound residue exists to slow the reaction. The finding that methyl-a-Dglucoside and other D-glucopyranosyl compounds enhance the utilization of ,8-D-glucoSy1 fluoride by glucoamylase is in keeping with the reported (25,53) high binding affinity of subsite 2 for compounds having one glucose residue. However, the catalysis of glucosyl transfer reactions under such conditions show the productive binding of 2 substrate molecules (donor and acceptor) at the enzyme's active center, presumably at subsites 1 and 2, respectively.
Clearly, a Michaelis-Menten mechanism is unable to account for the full catalytic capabilities of glucoamylase.
The mechanisms proposed for the actions of glucoamylase and glucodextranase with a-and p-D-glucosyl fluoride and various other substrates are illustrated in Scheme I which, for simplicity, is confined to certain reactions effected by glucoamylase. The hydrolysis or formation of a-1,6-glucosidl" linkages by both enzymes, for example, would proceed respectively by type I or II reactions similar to those in Scheme I. Evidence has been reported for the presence of a carboxyl group and a carboxylate anion in the active site of R. delemar glucoamylase (49), and these groups are considered essential for catalysis by the R. niveus enzyme as well (50). We have taken this as a reasonable model for the active site functional groups of both glucoamylase and glucodextranase, although no information is available on the structure of the latter. Scheme IA (I) illustrates the mechanism envisioned for the hydrolysis of maltose by glucoamylase, in which the carboxyl group acts as a general acid and the carboxylate anion acts as a general base assisting the attack of a water molecule on the substrate. This is written as a concerted reaction for reasons given below. The condensation of P-D-glucopyranose to form maltose, catalyzed by R. niveus glucoamylase (20), is illustrated in Scheme IA (11). In this case, the C-l hydroxyl group of the donor substrate is displaced by the C-4 hydroxyl group of the acceptor, with the functional roles of the catalytic groups reversed as required by the principle of microscopic reversibility.
In Schem IB, a mechanism for the hydrolysis of a-D-glucosyl fluoride by glucoamylase or glucodextranase is presented which proceeds by a pathway similar to that in Scheme IA (I) for the hydrolysis of maltose. This is consistent with the kinetics of a-D-ghCOSy1 fluoride hydrolysis and with the formation Scheme IC shows the postulated mechanism for the formation of methyl-a-maltoside from p-D-glUCOSyl fluoride and methyl-a-D-glucoside. This is similar to the condensation mechanism shown in Scheme LA ( I n for the formation of maltose from P-D-glucopyranose. Scheme IC is also consistent with the kinetics of @-D-glucosyl fluoride utilization which indicates a requirement for 2 molecules of substrate at the active site in order for reaction to occur. The fluorine atom of the donor molecule would be displaced by the C-4 hydroxyl of the second substrate molecule located at the acceptor site, forming P-maltosyl fluoride; this intermediate would rapidly hydrolyze to @-D-ghCOSe and @-D-glucosyl fluoride along the pathway shown in Scheme LA (I). The requirement for 2 bound @-D-glucosyl fluoride molecules to effect C-F cleavage favors a concerted mechanism, as shown, rather than a stepwise mechanism in which the fluoride ion of the donor departs fist, leaving a carbonium ion that is subsequently captured by the C-4 hydroxyl group of an incoming acceptor molecule. The apparent concerted displacement mechanism for @-Dglucosyl fluoride favors, although it does not prove, that the condensation Of @-D-glUCOSe to form maltose (20) also involves a concerted mechanism as in Scheme LA (10. From considerations of microscopic reversibility, the hydrolysis of maltose to give @-D-glUCOSe would then also be concerted, as shown. However, it is also possible that maltose and other glycosidically linked substrates may undergo stepwise reactions with carbonium ion intermediates. Indeed, as noted below, there are positive indications that a given enzyme may not act on all substrates with the same reaction mechanism. Regardless of the details of the proposed mechanisms, it is clear that, for both enzymes, the functional roles of the catalytic groups must be reversed not only in the hydrolysiscondensation reactions (Scheme LA) which are reversals of each other, but also in the stereospecifically different and essentially irreversible reactions with a-and p-D-glucosyl fluoride (Scheme I, B and C ) . In the case of glucodextranase, the apparent operation of a concerted mechanism illustrates a further capability arising out of the functional flexibility of the catalytic groups. Recent work (8) has shown that this enzyme catalyzes hydration and glycosyl transfer reactions with 2,6-anhydro-l-deoxy-~-gluco-hept-l-enitol by a mechanism which, in all probability, involves a carbonium ion intermediate. Thus, it would appear that glucodextranase is not limited to a single reaction mechanism with all substrates. Carbohydrases have often been characterized as operating by one particular mechanism, but this view may be too rigid. Dahlquist et al. (51), for example, concluded on the basis of the secondary isotope effect on the rate of hydrolysis of phenyl-@-~-[~H]glucoside by sweet almond @-glucosidase that, ''the mechanism of this enzyme would seem to be a classic example of the displacement mechanism suggested by KO&land (1953." However, the same enzyme catalyzes the hydration of D-glucal by a mechanism in which the direction of protonation of the substrate is opposite that assumed for p-D-glucosides and which appears to involve a carbonium ion intermediate (2). Similarly, @-galactosidase catalyzes reactions with D-galactal and with @-D-galactosides that differ from each other in protonation direction and mechanism (3), and P-amylase, which appears to act upon (Y-and @-maltosyl fluoride by a concerted reaction (7), has been found to catalyze the hydration of maltal by a process that apparently is carbonium ion-mediated. 5 The ability of these several enzymes to act on different substrates by different reaction mechanisms can be related to the functional flexibility of their catalytic groups. If such flexibility is an attribute of glycosylases in general (7), an increasing number of enzymes may be expected to be found capable of acting by more than one mechanism. Reactions requiring different mechanisms have, in fact, been reported for lysozyme (52) and for the saccharifying a-amylase of Bacillus subtilis (53-55).
Present findings, finally, support a concept of carbohydrase action that departs from traditional views in considering that glycoside hydrolases and glycosyltransferases form a class of interrelated glycosylases whose reactions effect a simple chemical change, the interchange of a glycosyl residue and a proton, glycosyl-X + H-X' + glycosyl-X' + H-X (1,18,56).
This concept informs us that a compound may need no more than the ability to be suitably aligned at the active site of an enzyme and the ability to yield a glycosyl residue on protonation in order to serve as a glycosyl donor. There are now many examples of enzymic glycosylation reactions catalyzed with substrates lacking a glycosidic bond and/or a-or @anomeric configuration. The fresh insight into enzymic mechanisms which the study of such reactions is able to provide suggests that the above unifying concept will find increasing adoption as a guiding principle in place of present models.