Barley (cid:98) - D -Glucan Exohydrolases with (cid:98) - D -Glucosidase Activity PURIFICATION, CHARACTERIZATION, AND DETERMINATION OF PRIMARY STRUCTURE FROM A cDNA CLONE*

Two (cid:98) -glucan exohydrolases of apparent molecular masses 69,000 and 71,000 Da have been purified from extracts of 8-day germinated barley grains and are designated isoenzymes ExoI and ExoII, respectively. The sequences of their first 52 NH 2 -terminal amino acids show 64% positional identity. Both enzymes hydrolyze the (1,3)- (cid:98) -glucan, laminarin, but also hydrolyze (1,3; 1,4)- (cid:98) -glucan and 4-nitrophenyl (cid:98) - D -glucoside. The complete sequence of 602 amino acid residues of the mature (cid:98) -glucan exohydrolase isoenzyme ExoII has been deduced by nucleotide sequence analysis of a near full- length cDNA. Two other enzymes of apparent molecular mass 62,000 Da, designated (cid:98) I and (cid:98) II, were also purified from the extracts. Their amino acid sequences are sim- ilar to enzymes classified as (cid:98) -glucosidases and although they hydrolyze 4-nitrophenyl (cid:98) -glucoside, their substrate specificities and action patterns are more typ- ical of polysaccharide exohydrolases of the (1,4)- (cid:98) -glu-can glucohydrolase type. Both the (cid:98) -glucan exohydrolase isoenzyme ExoI and the (cid:98) -glucosidase isoenzyme (cid:98) II release single glucosyl residues from the nonreducing ends of substrates and proton-NMR shows that ano- meric configurations are retained during hydrolysis by both classes of enzyme. These results raise general ques- tions regarding the distinction between polysaccharide exohydrolases and glucosidases, together with more specific questions regarding the functional roles of the two classes of enzyme in germinating barley grain.

Cell walls in the starchy endosperm of barley and other agriculturally important cereal grains are composed predominantly of (1,3;1,4)-␤-glucans and arabinoxylans (reviewed by ). In the case of barley, these two polysaccharides account for more than 90% by weight of walls of the starchy endosperm, with (1,3;1,4)-␤-glucan levels of approximately 70% and arabinoxylan levels of about 20% (Fincher, 1975;Ballance and Manners, 1978). In the germinating grain the cell walls represent a physical barrier between hydrolytic enzymes, such as ␣-amylases and endopeptidases which are secreted from the aleurone cells surrounding the starchy endosperm, and their starch and storage protein substrates which are packaged inside the starchy endosperm cells. Following germination, walls of starchy endosperm cells are extensively degraded and eventually disappear (Selvig et al., 1986). The removal of the walls is mediated by two (1,3;1,4)-␤-glucanases (EC 3.2.1.73), designated isoenzymes EI and EII (Woodward and Fincher, 1982a), and by the combined action of xylan endoand exohydrolases, and ␣-arabinofuranosidases (Preece and MacDougall, 1958;Slade et al., 1989).
Because of the quantitative importance of (1,3;1,4)-␤-glucans in barley endosperm cell walls and difficulties encountered with the extreme instability of the endo-xylanases, most attention has been focussed on the (1,3;1,4)-␤-glucanases in studies of cell wall degradation in germinating grain. The genes encoding the two isoforms are subject to independent regulation. The isoenzyme EI gene is expressed mainly in the scutellum of germinating grain, but also in young leaves and roots of barley seedlings, whereas expression of the isoenzyme EII gene appears to be restricted to the aleurone layer of germinating grain Fincher, 1992a, 1992b). Both isoenzymes release the characteristic tri-and tetrasaccharides 3-O-␤-cellobiosyl-D-glucose (G4G3G red ) and 3-O-␤-cellotriosyl-D-glucose (G4G4G3G red ) as the major oligomeric products from (1,3;1,4)-␤-glucans of the starchy endosperm walls (Parrish et al., 1960;Woodward and Fincher, 1982b). Given that cell wall (1,3;1,4)-␤-glucans may represent up to 18% of the total glucosyl residues stored in the starchy endosperm of barley (Morall and Briggs, 1978), it is highly likely that glucosyl residues bound in the tri-and tetrasaccharide products of (1,3;1,4)-␤-glucan endohydrolase action would eventually be released as free glucose and translocated as an energy source to the developing seedling. Prime candidates for the release of glucose from oligosaccharides derived from (1,3;1,4)-␤-glucans would be ␤-glucosidases or ␤-glucan exohydrolases, both of which have been detected in crude extracts of germinating barley grain ( Manners and Marshall, 1969;Simos et al., 1994;Leah et al., 1995).
In the work described here we have purified two ␤-glucan exohydrolases, which can hydrolyze both (1,3;1,4)-␤-glucans and (1,3)-␤-glucans, and two ␤-glucosidases. The chemical and enzymatic properties of the enzymes have been defined and a cDNA encoding one ␤-glucan exohydrolase isoform has been characterized. Comparisons of the substrate specificities and action patterns of the two classes of enzyme suggest that both classes release glucose from the nonreducing ends of (1,3;1,4)-␤-oligoglucosides, but that the ␤-glucan exohydrolases more rapidly catalyze the complete hydrolysis of these oligosaccharides.
Enzyme Extraction-Barley (Hordeum vulgare L., cv. Clipper) grain was surface-sterilized as described by Hoy et al. (1980). Grains (3.5 kg), maintained at approximately 40% (v/v) moisture content in the antibiotic solution, were germinated in the dark at 19 Ϯ 2°C for 8 days. The homogenization of grain and ammonium sulfate precipitation were as described previously (Hrmova and Fincher, 1993).
Rates of oligosaccharide hydrolysis were assayed at a final substrate concentration of 1 mM in 0.1 M sodium acetate buffer, pH 5.25 (isoenzyme ExoI) or pH 5.0 (isoenzyme ␤II), containing 160 g/ml BSA. The substrates were incubated with 35.1 and 15.3 pkat (based on 4-nitrophenol released from 4-NPG) of purified isoenzymes ExoI and ␤II, respectively, and hydrolysis rates were determined reductometrically. Products released from G3G4G red and G4G3G red after 15 min were concentrated and applied to Kieselgel 60 thin-layer plates, developed in ethyl acetate/acetic acid/water (2:1:1 by volume), and reducing products detected with the orcinol reagent (Hrmova and Fincher, 1993).
Inhibition of the ␤-glucan exohydrolase isoenzyme ExoI and the ␤-glucosidase isoenzyme ␤II by D-glucono-1,5-lactone and glucose was monitored by incubating 35.1 and 45.6 pkat isoenzymes ExoI and ␤II, respectively, in 0.1 M sodium acetate buffer, pH 5.25 (isoenzyme ExoI) or pH 5.0 (isoenzyme ␤II), containing 160 g/ml BSA. Rates were determined during the first 10 min of hydrolysis (Legler, 1990) after the addition of a fixed amount of enzyme to incubation mixtures containing different concentrations of 4-NPG and inhibitors. Kinetic data (K m and k cat ) and the dissociation constants (K i ) of enzyme-inhibitor complexes were processed by a proportional weighted fit, using a nonlinear regression analysis program based on Michaelis-Menten kinetics (Perella, 1988).
Protein Determination and Polyacrylamide Gel Electrophoresis-Protein content and SDS-PAGE were performed as described previously (Hrmova and Fincher, 1993).
Tryptic Digestions and Amino Acid Sequencing-Tryptic fragments of exo-␤-glucanase isoenzyme ExoI were generated and analyzed essentially as described by Chen et al. (1993). Selected peptides and enzyme fractions were sequenced in a Hewlett-Packard G1005A protein sequencer (Palo Alto, CA) using the Hewlett-Packard 3.0 sequencing routine, based on Edman degradation chemistry.
Isolation of cDNAs-A barley cDNA library (Clontech Laboratories Inc., Palo Alto, CA) had been prepared from poly(A) ϩ RNA of 5-day-old barley seedlings (H. vulgare L., cv. Boni) using both oligo(dT) and random priming. The cDNA was ligated into the ZAP II vector (Stratagene, La Jolla, CA). The library was screened by hybridization of nitrocellulose filter plaque replicas with a degenerate oligonucleotide which was end-labeled with [␥-32 ]ATP, and designed on the basis of NH 2 -terminal amino acid sequence data for the ␤-glucan exohydrolase isoenzyme ExoII. The sequence of the oligonucleotide probe was For subsequent screening with partial cDNA clones, cDNA inserts were labeled with deoxycytidine 5Ј-[␣-32 P]triphosphate, using random sequence nonanucleotides (Feinberg and Vogelstein, 1983). Hybridization with oligonucleotide or cDNA probes was performed as described previously .
The cDNA inserts of positive clones were rescued into plasmid pBSII SK(Ϫ) and both strands were sequenced using the dideoxynucleotide chain termination procedure (Sanger et al., 1977). Computer analyses were performed with the University of Wisconsin Genetics Computer Group software (Devereux et al., 1984). 1 H NMR Spectroscopy-1 H NMR spectra were generated on a Bruker ACP300 spectrometer in 5-mm external diameter tubes at a probe temperature of 303 K. Transients (32) were collected into 32K data points using a spectral width of 4 KHz and a relaxation delay of 4 s. A Gaussian weighting factor was applied to the data before Fourier transformation. The spectra were referenced against an external standard of sodium 2,2-dimethyl-2-silapentane-5-sulfonate at 303 K. The ␤-glucan exohydrolase isoenzyme ExoI (400 g, 5.8 nmol) and ␤-glucosidase isoenzyme ␤II (111 g, 1.8 nmol) were freeze-dried and redissolved in 0.1 M sodium acetate. The substrates laminarin (10 mg) and cellopentaose (3 mg) were dissolved in 80 mM sodium acetate. Enzymes and substrates were freeze-dried from 99.75% D 2 O three times to remove exchangeable protons and finally redissolved in 99.996% D 2 O. The p 2 H of the solutions were adjusted with DCl to 5.25 (isoenzyme ExoI) and 5.0 (isoenzyme ␤II) immediately prior to the NMR measurements.
Experiments were performed at a final concentration of 1.4% (w/v) laminarin and with 93 nkat isoenzyme ExoI, or 0.3% (w/v) cellopentaose and 4.5 nkat isoenzyme ␤II. Spectra of both substrates were recorded without the presence of enzyme and, following the addition of 50 l of enzyme in D 2 O, at regular time intervals for 90 min (laminarin) and 180 min (cellopentaose).

Purification of Exo-␤-glucanases-
The procedures developed to purify the exo-␤-glucanases from extracts of germinated barley are shown in Scheme 1. Enzyme yields, specific activities, and purification factors are presented in Table I. Unpurified grain extract and fractions precipitated with ammonium sulfate were assayed for (1,3)-␤-glucanase, (1,3;1,4)-␤glucanase, and ␤-glucosidase using the substrates laminarin, barley (1,3;1,4)-␤-glucan and 4-NPG, respectively. For each assay, most activity was detected in the material precipitated by 40 -80% saturated ammonium sulfate, which was used for subsequent purification steps. Exploratory experiments at this stage showed that the exo-␤-glucanases to be purified could hydrolyze both laminarin and 4-NPG, but that the (1,3)-␤glucan endohydrolases, which were present at very high levels (Hrmova and Fincher, 1993) did not hydrolyze 4-NPG. Because the major objective of the work was to purify exo-␤-glucanases rather than the (1,3)-␤-glucan endohydrolases, 4-NPG was used to assay exo-␤-glucanases in all subsequent purification procedures, despite the fact that it was not the preferred substrate and that it would also detect ␤-glucosidases.
Following dialysis against 20 mM Tris-HCl buffer, pH 7.6, containing 10 mM sodium azide and 3 mM 2-mercaptoethanol, the 40 -80% ammonium sulfate fraction was applied to a 4.5 cm ϫ 20-cm column of DEAE-cellulose equilibrated in 20 mM Tris-HCl buffer, pH 7.6. Most activity passed through the column (Table I) and was immediately transferred to a 3.2 cm ϫ 24-cm column of CM-Sepharose equilibrated in 50 mM sodium acetate buffer, pH 4.9. After unbound proteins were eluted, a linear gradient (3.8 liters) of 0 to 0.6 M NaCl in the same buffer was SCHEME 1. Summary of purification procedures for exo-␤-glucanases and ␤-glucosidases from extracts of germinated barley. applied to the column at a linear flow rate of 76 cm h Ϫ1 . Five peaks of activity against 4-NPG were detected ( Fig. 1). Of these, peaks I and IV were active against laminarin, (1,3;1,4)-␤-glucan, and 4-NPG, and proved to contain exo-␤-glucanases. Peak III was most active against 4-NPG and was subsequently shown to contain the ␤-glucosidases (Scheme 1, Fig. 1). Peak II was subjected to further purification by chromatofocusing, but no ␤-glucan exohydrolase activity could be detected (data not shown). Peak V contained a third isoform of ␤-glucosidase, which had an isoelectric point of 9.3, but the fractions were contaminated with ␤-amylase and have not been purified further.
Fractions associated with peaks I (fractions 42-50, Fig. 1), III (fractions 63-72), and IV (fractions 127-137) were pooled and concentrated by ultrafiltration. During ultrafiltration, fractions from peaks I and IV were equilibrated in 25 mM ethanolamine-HCl buffer, pH 9.4, while fractions from peak III were equilibrated in 25 mM triethylamine-HCl buffer, pH 11. Pooled fractions were loaded onto 1 cm ϫ 28.5-cm columns of PBE 94 or PBE 118 resins equilibrated in the buffers described above and eluted at a flow rate of 31 cm h Ϫ1 with 9.1% (v/v) Polybuffer 96 adjusted to pH 7 with concentrated HCl (peaks I and IV), or with 2.2% (v/v) Pharmalyte 8 -10.5 adjusted to pH 8 with HCl (peak III).
Two peaks of activity against 4-NPG, designated peaks A and B, were detected when the pooled fractions of peak I from the CM-Sepharose column (Fig. 1) were eluted from the chromatofocusing column at pH 7.8 and 7.6 ( Fig. 2). The large protein peak eluting in fractions 1-20 contained (1,3)-␤-glucan endohydrolase isoenzymes GI and GII (Hrmova and Fincher, 1993) but showed no activity against 4-NPG (Fig. 2). Both peaks (ExoIA and ExoIB) with 4-NPG activity were also active against laminarin and barley (1,3;1,4)-␤-glucan. When examined by SDS-PAGE under reducing conditions, the peaks contained major proteins of 68 -69 kDa, together with several contaminating proteins (data not shown). The fractions in the peaks were concentrated by ultrafiltration in 50 mM sodium acetate buffer, pH 5.0, containing 2 M NaCl and subjected separately to hydrophobic interaction chromatography on a 1 cm ϫ 9-cm column of phenyl-Sepharose equilibrated in the same buffer. The enzymes bound weakly to the column and were eventually eluted at a flow rate of 115 cm h Ϫ1 with the equilibration buffer. The contaminating proteins bound to the phenyl-Sepharose and were eluted with a 0 -2 M NaCl gradient in the same buffer. However, SDS-PAGE showed that the active enzyme fractions from the phenyl-Sepharose column still contained minor contaminating proteins, which were subsequently removed by size exclusion chromatography on a 1.5 cm ϫ 96-cm Bio-Gel P-100 column equilibrated in 50 mM sodium acetate buffer, pH 5.0, containing 200 mM NaCl and 1 mM dithiothreitol, at a flow rate of 0.5 cm h Ϫ1 .
After size exclusion chromatography of the ExoIA and ExoIB preparations, the active fractions were concentrated by ultrafiltration. Amino acid sequence data showed that their NH 2terminal sequences were identical for the first 52 residues. While it is formally possible that the isoenzyme ExoI and ExoII preparations represent separate isoforms, we believe that these more probably correspond to a single protein and that the resolution of two peaks during chromatofocusing (Fig. 2) may result from partial deamination of glutamine or asparagine residues, or from different levels of post-translational modification. These enzyme fractions are designated ␤-glucan exohydrolase isoenzyme ExoI.
When the pooled fractions of CM-Sepharose peak IV ( Fig. 1) were subjected to chromatofocusing, two peaks of 4-NPG activity were eluted at pH values of 8.0 and 7.7 and although SDS-PAGE showed that the major proteins had apparent molecular mass values of 70 -71 kDa, some contaminating proteins were still present (data not shown). The contaminating proteins were removed by size exclusion chromatography on the Bio-Gel P-100 column, as described above. Following the separate purification of the chromatofocusing peaks on Bio-Gel P-100, active fractions were concentrated by ultrafiltration and their NH 2 -terminal amino acid sequence determined. Both proteins had identical sequences for 52 NH 2 -terminal amino acids and we conclude that they too probably represent a single protein in which deamination or other post-translational modifications result in slightly different mobilities on the chromatofocusing column (data not shown). These enzymes are designated ␤-glucan exohydrolase isoenzyme ExoII. The NH 2terminal amino acid sequences of the ␤-glucan exohydrolase isoenzymes ExoI and ExoII are compared in Table II.
Purification of ␤-Glucosidases-The pooled fractions of peak III from the CM-Sepharose column (Fig. 1) were also resolved into two peaks of 4-NPG activity after chromatofocusing (Fig. 3), but in this case the fractions under the peak exhibited no activity against laminarin or (1,3;1,4)-␤-glucan. It seemed likely, therefore, that these were ␤-glucosidases. Amino acid sequence analyses revealed a single amino acid difference in the sequence of their 50 NH 2 -terminal amino acids (Table II) and the enzymes are designated ␤-glucosidase isoenzyme ␤I (pI 8.9) and isoenzyme ␤II (pI 9.0) (Fig. 3). However, after pooling the fractions SDS-PAGE revealed the presence of low levels of contaminating proteins in each; these contaminating proteins were removed by size exclusion chromatography on Bio-Gel P-100, as described above. The amino acid sequence of the ␤-glucosidase isoenzymes corresponds with those already published for barley ␤-glucosidases (Simos et al., 1994;Leah et al., 1995) and is similar to other ␤-glucosidase sequences in the data bases (Table II).
Purity and Yields-The final enzyme preparations were examined by SDS-PAGE, which revealed single protein bands at the loadings used (Fig. 4). For the ␤-glucan exohydrolases, isoenzyme ExoI was represented by a single band of apparent M r 69,000, while the value for isoenzyme ExoII was 71,000 (Fig. 4). The ␤-glucosidases also exhibited single bands following SDS-PAGE analysis, but in this case their apparent M r values were 62,000 (Fig. 4). The purities of all four enzyme preparations were confirmed by NH 2 -terminal amino acid sequence analysis, in which no secondary sequences could be detected. The high degree of purity of the enzymes enabled the sequence of 50 or more amino acids to be determined (Table II).
The final recoveries of the enzymes were generally low when expressed on a percentage basis (Table I), primarily because of the highly selective pooling of column fractions that was necessary to separate the enzymes from contaminating proteins, particularly after CM-Sepharose ion exchange chromatography and size exclusion chromatography. Furthermore, we believe the high pH values (pH 11) used during chromatofocusing (Fig.  3) caused considerable losses in enzyme activity. The yields are also lower than those reported by Leah et al. (1995), although the enzymes were purified from extracts of different tissues, using different procedures. Nevertheless, milligram quantities of each enzyme were obtained from extracts of 3.5 kg of germinated barley. The final purification factors were also relatively low (Table I) but this is because of the presence, in the crude extract, of a battery of enzymes capable of hydrolyzing the 4-NPG substrate. Thus, the initial specific activity was much higher than that attributable to any single enzyme and the final purification achieved therefore appears correspondingly lower.
In contrast, the ␤-glucosidases hydrolyze 4-NPG but have no detectable activity on laminarin or (1,3;1,4)-␤-glucan (Table  III). However, isoenzyme ␤II had a relatively high activity against cellopentaose and the rate of hydrolysis increased markedly with DP in the (1,4)-␤-oligoglucoside series (Fig. 5B). These enzymes preference for (1,4)-␤-glucosyl residues at the nonreducing terminus of oligomeric substrates was further demonstrated by their ability to hydrolyze G4G3G red at much higher rates than G3G4G red (Figs. 5 and 6). It should also be noted that isoenzyme ␤II hydrolyzed cellobiose more rapidly than laminaribiose (Fig. 5B), but that the preparation of Leah et al. (1995) hydrolyzed laminaribiose more rapidly than cellobiose. We are unable to explain this difference at this stage.
The availability of the two (1,3;1,4)-␤-oligoglucosides G4G3G red and G3G4G red , in which the arrangements of (1,3)and (1,4)-linkages differ, enabled us to investigate whether glucose units were released from the reducing or nonreducing ends of the substrates. For both types of enzyme initial reaction products on G4G3G red were glucose and laminaribiose (G3G red ), while for G3G4G red the initial products were glucose and cellobiose (G4G red ) (Fig. 6). Thus, both enzyme types remove glucose from the nonreducing termini of these substrates. The faster rate of hydrolysis of G4G3G red by the ␤-glucosidase, compared with its rate against G3G4G red , also confirmed the marked preference of this enzyme for (1,4)-␤-linkages (Fig. 6), as already observed in its relatively high activity against (1,4)-␤-oligoglucosides (Fig. 5B).
Inhibition by D-Glucono-1,5-lactone and Glucose-␤-Glucan exohydrolases and ␤-glucosidases have often been distinguished on the basis of differential inhibition with D-glucono-1,5-lactone; ␤-glucosidases are generally more susceptible to inhibition (Reese et al., 1968). Kinetic parameters calculated from inhibition data are shown in Table IV, where it is apparent that K i values for both D-glucono-1,5-lactone and glucose inhibition of ␤-glucan exohydrolase isoenzyme ExoI are significantly lower than for the isoenzyme ␤II (Table IV). Thus, the D-glucono-1,5-lactone shows a greater affinity for the ␤-glucan exohydrolase.
In the case of the barley ␤-glucan exohydrolase isoenzyme ExoI, values for K i (glucose)/K m (4-NPG) of 6.3 and for K i (glucose)/K i (lactone) of 4,190 are exceptionally low and high, re-spectively. Previously recorded values are in the 20 -100 range and approximately 1000, respectively (Legler et al., 1990). This indicates that the nitrophenyl group in 4-NPG makes a large contribution to binding (K i /K m ϭ 6.3) (Legler et al., 1990).
ably resulted from mutarotation of newly released ␤-anomers ( Fig. 7A; Chen et al. (1995)). The decrease in the doublet signal at about ␦ ϭ 4.79 ppm, J ϭ 8.5 Hz corresponds to the reduction of inter-residue anomeric protons of the substrate (Fig. 7A). At the same time the doublet at ␦ ϭ 4.52, J ϭ 7.9 Hz increased and this signal corresponds to gentiobiose released from laminarin during hydrolysis. 2 During the hydrolysis of cellopentaose by the ␤-glucosidase isoenzyme ␤II, the ␣-anomeric protons of the substrate appear as a doublet at ␦ ϭ 5.23 ppm, J ϭ 3.7 Hz, while the ␤-anomeric protons of the substrate appear as a doublet at ␦ ϭ 4.66 ppm, J ϭ 7.7 Hz (Fig. 7B). Following the addition of the ␤-glucosidase, ␤-anomeric protons of released glucose are observed as a doublet at ␦ ϭ 4.65 ppm, J ϭ 8.1 Hz. The apparent triplet in this region results from the two overlapping doublets described above (Fig. 7B). It is noteworthy that after the addition of the enzyme a shift in the HDO peak from ␦ ϭ 4.72 to ␦ ϭ 4.71 ppm was observed (Fig. 7B). This shift, which commonly appears following the addition of enzymes to reaction mixtures (Withers et al., 1986), was not caused by pH or temperature fluctuations or by salt effects, because these parameters were held constant. The shift might result from changes in viscosity of the reaction mixture following the addition of enzyme.
The apparent triplet centered around ␦ ϭ 4.52 ppm arises from overlapping doublets of inter-residue anomeric protons of both intrachain and nonreducing terminal glycosidic linkages of the cellopentaose substrate. As hydrolysis proceeds this triplet decreases in intensity (Fig. 7B). The signal corresponding to the ␣-anomeric protons of glucose at ␦ ϭ 5.23, J ϭ 3.7 Hz increased in intensity during hydrolysis and reflects mutarotation of the ␤ anomeric protons of the glucose initially released from the substrate (Fig. 7B). Based on these data, it is clear that both the barley ␤-glucan exohydrolase and the ␤-glucosidase catalyze hydrolysis of their substrates with retention of anomeric configuration in the glucose product released (Fig. 7).
Characterization of a cDNA Encoding ␤-Glucan Exohydrolase Isoenzyme ExoII-In the initial screening of the cDNA library a single 600-base pair cDNA clone was identified among 6 ϫ 10 4 plaques and nucleotide sequence analysis showed that it encoded the NH 2 -terminal region of the barley ␤-glucan exohydrolase isoenzyme ExoII. The cDNA was subsequently labeled and used to re-screen 10 5 plaques, two of which hybridized strongly with the probe. One was 1.6 kilobases in length and a 1-kilobase fragment from its 3Ј-end was used to screen another 3 ϫ 10 5 plaques. This resulted in the isolation of a near full-length cDNA encoding the ␤-glucan exohydrolase isoenzyme ExoII; the sequences of the overlapping regions of the three cDNAs were identical.
Both strands of the 2,105-base pair cDNA were sequenced (Fig. 8). To assist with the nucleotide sequencing, purified ␤-glucan exohydrolase isoenzyme ExoI was hydrolyzed with trypsin and, after separation of tryptic peptides by high performance liquid chromatography, nine fragments were subjected to amino acid sequence analysis (data not shown). This isoform was chosen because initial purification protocols resulted in much lower yields of isoenzyme ExoII, which was therefore not available at the time for the generation of tryptic peptides for amino acid sequence analysis. However, subsequent minor modifications to the procedure led to much higher relative yields of isoenzyme ExoII (Table I). Nevertheless, the peptides from isoenzyme ExoI were similar to blocks of amino acid sequences deduced from the cDNA and were therefore useful in ensuring no errors were made with the reading frame during nucleotide sequencing.
The cDNA has an open reading frame which extends from nucleotides 26 to 1897 and the amino acid sequence deduced from nucleotides 92-247 corresponds exactly with that determined directly from the 52 NH 2 -terminal amino acids of ␤-glucan exohydrolase isoenzyme ExoII (Table II). On this basis, we conclude that the cDNA encodes ␤-glucan exohydrolase isoenzyme ExoII.
The region from the codon specifying the NH 2 -terminal aspartic acid residue at nucleotide 92 to the stop codon at nucleotide 1898 encodes a polypeptide of 602 amino acid residues, with a calculated M r of 65,102 and an estimated pI of 7.9. Four potential N-glycosylation sites are present, and are encoded by sequences beginning at nucleotides 206, 749, 983, and 1580 (Fig. 8). Immediately 5Ј to the codon which specifies the NH 2terminal aspartic acid residue is a sequence corresponding to a putative signal peptide of 22 residues (Fig. 8). The coding region of the mature polypeptide is followed by a TGA stop codon, which becomes part of a 208-nucleotide pair 3Ј-untranslated region, and is followed by a polyadenylic acid tail of 30 residues (Fig. 8). A potential polyadenylation signal AATAAA begins 80 nucleotides upstream from the start of the polyadenylate tail. The coding region of the ␤-glucan exohydrolase cDNA has an overall (G ϩ C) content of 54% and some preference for the use of G or C in the wobble base position of codons (65%). 2 M. Hrmova and G. B. Fincher, unpublished data.

DISCUSSION
Two basic ␤-glucan exohydrolases of apparent M r 69,000 and 71,000 have been purified from extracts of 8-day germinated barley grain by fractional precipitation with ammonium sulfate, ion exchange chromatography, chromatofocusing, hydrophobic interaction chromatography, and gel filtration chromatography (Scheme 1). The purification procedure allowed the separation of ␤-glucan exohydrolases from (1,3)-and (1,3;1,4)-␤-glucan endohydrolases, both of which were present at high levels in the initial grain extracts. The exohydrolases have been designated isoenzymes ExoI and ExoII in order to distinguish them from the (1,3)-␤-glucan endohydrolase isoenzymes GI-GVII and the (1,3;1,4)-␤-glucan endohydrolase isoenzymes EI and EII (Woodward and Fincher, 1982a;Xu et al., 1992;Høj and Fincher, 1995). The purified enzymes hydrolyze 4-NPG, but have been identified with polysaccharide exohydrolases rather than ␤-glucosidases because they also hydrolyze polymeric substrates such as laminarin and (1,3;1,4)-␤-glucan (Table III), releasing glucose as the primary degradation product. Furthermore, the use of (1,3;1,4)-␤-oligoglucosides shows clearly that the glucose units are released from the nonreducing termini of the oligomeric substrates (Fig. 6). The ␤-glucan exohydrolases can be readily distinguished from two ␤-glucosidases that were also purified from the grain extracts (Scheme 1), not only on the basis of differences in substrate specificity (Table III), but also because of their different isoelectric points, molecular weights, and NH 2 -terminal amino acid sequences . No linkage specificity has been assigned to the ␤-glucan exohydrolases, because they are able to release glucose from (1,3)-␤-glucans, (1,3;1,4)-␤-glucans, and (1,2)-, (1,3)-, (1,4)-, (1,6)-, and 1,3;1,4-␤-oligoglucosides (Table III). 2 Two isoforms of the barley ␤-glucan exohydrolases were resolved during the purification procedure. The sequences of their NH 2 -terminal amino acids show considerable divergence; only 33 of the first 52 residues are identical (Table II). The complete primary structure of the ␤-glucan exohydrolase isoenzyme ExoII has been deduced from a near full-length cDNA clone (Fig. 8). The nucleotide sequence reveals the presence of a putative signal peptide of 22 residues, suggesting that the enzyme is secreted from cells in which it is expressed. The putative signal peptide region is followed by a region that encodes a protein with 602 residues, a calculated molecular mass of 65,102 Da, and a deduced pI of 7.9. These values may be compared with an apparent M r of 71,000 estimated from SDS-PAGE (Fig. 4) and an isoelectric point of 8.0 estimated by chromatofocusing (data not shown). The difference between the molecular weight calculated from the deduced amino acid sequence and that measured by SDS-PAGE may result from glycosylation of the mature enzyme, which carries four potential N-glycosylation sites (Fig. 8).
When the barley ␤-glucan exohydrolase isoenzyme ExoII sequence (Fig. 8) was compared with sequences in the DNA and protein data bases, a sequence tagged site (clone ABC254; accession number L43939) from barley was found to have a nucleotide sequence identical with nucleotides 1853-2016 of the cDNA isolated here (Fig. 8). The ABC254 sequence is also in the GrainGenes data base and has been mapped to the long arm of barley chromosome 1 (7H), close to the centromere (Langridge et al., 1995). It is therefore possible to positively identify the ABC254 probe as part of the gene encoding the barley ␤-glucan exohydrolase isoenzyme ExoII, and to incorporate this gene into high density genetic maps.
An important question raised in the work described here FIG. 7. Partial 1 H NMR spectra of anomeric protons of laminarin and hydrolysis products released from it by the action of ␤-glucan exohydrolase isoenzyme ExoI (A), and of cellopentaose and products released by the action of ␤-glucosidase isoenzyme ␤II (B). The reference t ϭ 0 spectra were recorded before the addition of enzyme, and the other spectra were recorded at time t (min) after the addition of enzyme. The large resonance signal at approximately ␦ ϭ 4.71 ppm arises from residual HDO.
relates to the differences between ␤-glucan exohydrolases and ␤-glucosidases, and how they should be classified within current Enzyme Commission groups. Jermyn (1961) andlater Fry (1995) have pointed out that the commonality between these enzymes lies in their ability to cleave nonreducing terminal glucosyl residues from their substrates. Their specificity for the "aglycone" portion of the substrate and the type of linkage hydrolyzed is quite variable and their classification according to the two subtypes as recognized by the Enzyme Commission is somewhat arbitrary. The barley ␤-glucan exohydrolase isoenzymes ExoI and ExoII release glucose from the nonreducing end of substrates, with retention of anomeric configuration (Figs. 6 and 7). In some respects they are similar in specificity and action pattern to plant (1,3)-␤-glucan exohydrolases that have been classified in the EC 3.2.1.58 group (Cline and Albersheim, 1981;Lienart et al., 1986;Labrador and Nevins, 1989), but they are capable of hydrolyzing substrates with linkages other than those in the (1,3)-position. While they are clearly polysaccharide exohydrolases, rates of hydrolysis of oligomeric substrates do not increase dramatically with increasing DP (Fig. 5A), but this may simply reflect subsite binding energies and the organization of the substrate-binding site (Varghese et al., 1994;Hrmova et al., 1995).
Classification of the barley ␤-glucosidases within existing EC groups (Webb, 1994) is also difficult. Although their amino acid sequences show high degrees of similarity to a wide range of ␤-glucosidases (EC 3.2.1.21) in the protein and DNA data bases, they hydrolyze cellodextrins at significantly higher rates than the aryl-glucoside 4-NPG and, more importantly, the rate of hydrolysis increases with DP of the cellodextrin substrate (Table III; Fig. 5B). The ␤-glucosidases release glucose from the nonreducing ends of substrates (Fig. 6), with retention of anomeric configuration (Fig. 7B) and show a marked preference for (1,4)-␤-linkages (Figs. 5B and 6). Based on the assumption that ␤-glucosidases prefer substrates of the type G-O-X (where X could either be another glucosyl residue for which linkage position would not be crucial, or a non-glycosyl aglycone) and that rates of hydrolysis would not increase, or actually decrease with increasing DP (Reese et al., 1968;Fadda et al., 1994;Sun et al., 1995), the isoenzymes ␤I and ␤II would not be classified as true ␤-glucosidases. Rather, their specificities and action patterns are characteristic of 1,4-␤-glucan glucohydrolases (EC 3.2.1.74), as noted for the celD gene product from Pseudomonas fluorescens (Rixon et al., 1992). If this is correct, then the classifications of several other putative ␤-glucosidases might need to be revised. The transition state analogue D-glucono-1,5lactone has been used diagnostically to inhibit ␤-glucosidases (Reese et al., 1968;Legler, 1990), but this compound inhibits the barley ␤-exohydrolases more strongly than the ␤-glucosidases (Table IV). It might be concluded from these results that the level of inhibition is also a function of subsite binding energies in the substrate-binding regions of the enzymes.
The final consideration here relates to the functions of the ␤-glucan exohydrolases and ␤-glucosidases in germinating barley grain or in young seedlings. It should be noted that the enzymes could be derived from the aleurone, scutellum, starchy endosperm, coleoptile, young leaves, or young roots, all of which are likely to be present in the 8-day germinated barley grain used for enzyme isolation. The availability of cDNA clones for both classes of enzyme ( Fig. 8; Leah et al. (1995)) FIG. 8. Nucleotide sequence and derived amino acid sequence of the cDNA encoding barley ␤-glucan exohydrolase isoenzyme ExoII. The putative signal peptide is shaded and the arrow indicates the NH 2 -terminal amino acid residue of the mature enzyme. The four potential N-glycosylation sites are underlined, and the potential polyadenylation signal is shown by double underlining. opens the way for defining their precise tissue location. ␤-Glucosidases are likely to have many functions in cellular processes during normal growth and development of plants, depending on their substrate specificities (Leah et al., 1995). Maize ␤-glucan exohydrolases have been implicated in cell wall loosening in elongating coleoptiles (Hoson and Nevins, 1989), in which high levels of this class of enzyme are correlated with significant decreases in cell wall (1,3;1,4)-␤-glucan (Sakurai and Masuda, 1978). Although it is difficult to envisage how exohydrolases might mediate in wall loosening (Fry, 1995), there is certainly little or no (1,3;1,4)-␤-glucan endohydrolase of the EC 3.2.1.73 group in elongating barley coleoptiles (Slakeski and Fincher, 1992a).
A second possible function for the enzymes is to participate in the complete depolymerization of endosperm cell wall (1,3; 1,4)-␤-glucan to glucose. The major oligosaccharides released from this polysaccharide by (1,3;1,4)-␤-glucan endohydrolases are G4G3G red and G4G4G3G red (Woodward and Fincher, 1982b). Both the ␤-glucan exohydrolases and the ␤-glucosidases can hydrolyze these oligosaccharides (Figs. 5 and 6;Leah et al. (1995)), although the rate of hydrolysis of (1,3)-␤-glucosyl linkages by the ␤-glucosidases is relatively slow. The third possible role for the enzymes, in particular the ␤-glucan exohydrolases, could be related to the protection of germinating grain or young seedlings from pathogen attack . (1,3)-␤-Glucan endohydrolases (EC 3.2.1.39) are classified among the pathogenesis-related proteins and could protect the young plant from microbial infection through their ability to hydrolyze the (1,3;1,6)-␤-glucans found in fungal cell walls. The ␤-glucan exohydrolases purified here are also able to hydrolyze (1,3)-and (1,3;1,6)-␤-glucans (Table III). 2 In any case, the availability of cDNAs for the exo-␤-glucanase will enable the time course and regulation of its gene expression to be examined, both in normal growth and development, and after pathogen attack, and this will undoubtedly provide clues as to the functions of the ␤-glucan exohydrolases and the ␤-glucosidases in germinated grain and young barley seedlings.