Mutational and structural analysis of aglycone specificity in maize and sorghum beta-glucosidases.

Plant beta-glucosidases display varying substrate specificities. The maize beta-glucosidase isozyme Glu1 (ZmGlu1) hydrolyzes a broad spectrum of substrates in addition to its natural substrate DIMBOA-Glc (2-O-beta-d-glucopyranosyl-4-hydroxy-7-methoxy-1,4-benzoxaxin-3-one), whereas the sorghum beta-glucosidase isozyme Dhr1 (SbDhr1) hydrolyzes exclusively its natural substrate dhurrin (p-hydroxy-(S)-mandelonitrile-beta-d-glucoside). Structural data from cocrystals of enzyme-substrate and enzyme-aglycone complexes have shown that five amino acid residues (Phe198, Phe205, Trp378, Phe466, and Ala467) are located in the aglycone-binding site of ZmGlu1 and form the basis of aglycone recognition and binding, hence substrate specificity. To study the mechanism of substrate specificity further, mutant beta-glucosidases were generated by replacing Phe198, Phe205, Asp261, Met263, Phe377, Phe466, Ala467, and Phe473 of Glu1 by Dhr1 counterparts. The effects of mutations on enzyme activity and substrate specificity were studied using both natural and artificial substrates. The simple mutant replacing Phe198 by a valine had the most drastic effect on activity, because the capacity of this enzyme to hydrolyze beta-glucosides was almost completely abolished. The analysis of this mutation was completed by a structural study of the double mutant ZmGlu1-E191D,F198V in complex with the natural substrate. The structure reveals that the single mutation F198V causes a cascade of conformational changes, which are unpredictable by standard molecular modeling techniques. Some other mutations led to drastic effects: replacing Asp261 by an asparagine decreases the catalytic efficiency of this simple mutant by 75% although replacing Tyr473 by a phenylalanine increase its efficiency by 300% and also provides a new substrate specificity by hydrolyzing dhurrin.

In maize (Zea mays L.), the major function of ␤-glucosidases (DIMBOA-glucosidases) 1 is in defense against the European corn borer and other pests (3). The catalytically active form of both maize and sorghum ␤-glucosidases is a 120-kDa homodimer or its multimers (9,10), which are also confirmed by x-ray crystallography (11). The primary structures of maize and sorghum ␤-glucosidases contain the highly conserved peptide motifs TFNEP and ITENG, which contain the two glutamic acids (Glu 191 and Glu 406 , numbering of ZmGlu1, Fig. 1) involved in catalysis as the general acid/base catalyst and the nucleophile, respectively, in all family 1 ␤-glucosidases (12). These residues form part of a slot-like active site (13) and are required in the two steps of the substrate hydrolysis (14). In the glycosylation step, the nucleophilic Glu 406 attacks the anomeric carbon (C-1) of the substrate and forms a covalent glycosylenzyme intermediate with concomitant release of the aglycone after protonation of the glucosidic oxygen by the acid catalyst Glu 191 (14). In the deglycosylation step, Glu 191 acts as a base catalyst to activate a water molecule, which performs a nucleophilic attack on the covalent glycosyl-enzyme, releasing the glucose and regenerating the nucleophilic Glu 406 . The two catalytic glutamic acids (i.e. the nucleophile and the acid/base catalyst) of ZmGlu1 are positioned within the active site at expected distances (ϳ5.5 Å) for this mechanism (15).
Although much progress has been made in understanding the mechanism of catalysis and defining the roles of the two catalytic glutamates (14,16), one fundamental question about ␤-glucosidase-catalyzed reactions still remains to be answered: what determines the precise substrate specificity, including the site and mechanism of aglycone binding? Until recently, there was little or no information as to how ␤-glucosidases recognize their substrates and interact with them, specifically the aglycone moiety, which is the basis of tremendous diversity in * This work was supported by a CNRS-National Science Foundation joint program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The natural substrates and is responsible for subtle substrate specificity differences among ␤-glucosidases. Recently, using the sorghum ␤-glucosidase Dhr1 and maize ␤-glucosidase Glu1 chimeras, Cicek et al. (18) have determined that the aglycone (i.e. substrate) specificity-determining sites are different in ZmGlu1 and SbDhr1. These two enzymes provide an ideal model system to address questions related to substrate specificity, because they represent extremes in substrate specificity. SbDhr1 hydrolyzes only its natural substrate dhurrin, whereas ZmGlu1 hydrolyzes a broad spectrum of artificial and natural substrates, including its natural substrate DIMBOA-Glc, but it does not hydrolyze dhurrin. Cicek et al. (18) showed that specificity for dhurrin hydrolysis resides mostly in a C-terminal octapeptide ( 462 SSGYTERF 469 ) of SbDhr1 where SbDhr1 and ZmGlu1 sequences differ from each other by four amino acid substitutions, although specificity for DIMBOA-Glc hydrolysis is not within the homolog of the aforementioned octapeptide in ZmGlu1, nor within the extreme 47-amino acid long C-terminal domain of ZmGlu1. The recently resolved crystal structures of ZmGlu1 (11,15) provided new insights into the identity of the amino acids within the active site that are involved in aglycone recognition and binding (i.e. substrate specificity). The data showed that the aglycone moiety of the substrate is sandwiched between Trp 378 on one side and Phe 198 , Phe 205 , and Phe 466 on the other. The major mechanism of aglycone recognition and binding appears to be aromatic stacking and -interactions between aromatic aglycones and the above mentioned amino acids. Moreover, several others amino acids (e.g. Pro 377 , Asp 261 , Met 263 , Ala 467 , and Tyr 473 ) were identified to be potentially involved in substrate specificity (11).
The purpose of the study described in this paper was to investigate further the mechanism of substrate (aglycone) recognition and binding in ␤-glucosidases using the maize isozyme ZmGlu1 and the sorghum isozyme SbDhr1 as model systems.
To this end, we have produced site-directed mutants of ZmGlu1 and chimeric enzymes, which had been produced by reciprocal domain-swapping between ZmGlu1 and SbDhr1 and used in a previously published study (18). The targeted residues were chosen based on their localization in the active site aglycone binding pocket, as derived from the structural data, and on amino acid sequence comparisons between ZmGlu1 and Sb-Dhr1 (Fig. 1). Subsequently, the mutant F198V, having the most drastic effect on enzyme activity, was submitted to a crystallographic structure analysis in complex with the natural substrate. We demonstrate that at least three residues were critical in ZmGlu1 substrate (i.e. aglycone) specificity and that a single amino acid change allows this protein to hydrolyze dhurrin. Implications of these findings on general ␤-glucosidase specificity are also discussed.

EXPERIMENTAL PROCEDURES
General Methods-Manipulation of DNA and Escherichia coli were performed using standard procedures (19). DNA synthesis by the PCR was performed with the high fidelity Pfu TM Turbo DNA polymerase (Stratagene). DNA was purified from gel with the QiaQuick gel extraction kit (Qiagen). All restriction enzymes were from Stratagene and T4 DNA ligase from Promega. All constructs were sequenced for confirmation in the VBI Core Laboratory Facility (Blacksburg, VA).
Mutagenesis of ␤-Glucosidases-The criteria for choosing the targeted residues was their localization in the active site aglycone binding pocket of the ZmGlu1 protein and amino acid sequence comparisons between ZmGlu1 and SbDhr1. The double mutant ZmGlu1-E191D,F198V was produced and overexpressed for the purpose of the structural analysis in complex with the natural substrate. For that reason, it was necessary to additionally mutate the acid base residue, to be sure to obtain a completely inactive enzyme. Mutated cDNAs were constructed by the PCR-mediated overlap extension using a pair of complementary primers (sense:pS and antisense:pA) as described (20) and vector-specific primers T7 prom (sense) (5Ј-TAATACGACTCACTATAGGG-3Ј)) and T7 term (antisense) (5Ј-ATGCT-AGTTATTGCTCAGCGGT-3Ј). The oligonucleotides used in this study are listed in Table I. The DNA templates used in this study were PET21a-ZmGlu1, pET21a-Chim16, and pET21a-Chim21, which were previously used for expression in E. coli and subsequent purification in our laboratory (11). The T7 prom -pA and T7 term -pS PCR products were gel purified and combined in the second PCR step using T7 prom and T7 term oligonucleotides to obtain the full-length mutated cDNA. The resulting PCR product was gel purified, digested with NheI and XhoI restriction enzymes and cloned into the expression plasmid pET28a (Novagen). This plasmid allows the fusion of His-tagged residue extension at the N-terminal part of the mutated ␤-glucosidases, allowing purification on a Ni 2ϩ column (see below).
Expression and Purification of Recombinant ␤-Glucosidases by Ni 2ϩ Column Chromatography-Wild type, chimeric and mutated ␤-glucosidases were produced in E. coli BL21(DE3) pLys cells (F-ompT hsdSP r B-m B-gal dcm) (Stratagene) under the control of the T7 RNA polymerase promoter in the expression plasmid pET28a (Novagen). Cells were grown at 37°C until A 600 nm ϭ 0.5, and recombinant protein expression was induced by the addition of 0.1 mM isopropyl-1-thio-␤-Dgalactopyranoside for 4 h at 22°C. Cell lysis was performed using 10 ml of extraction buffer (100 mM Tris-HCl, 50 mM NaCl, pH 8.0, 0.2 mM phenylmethylsulfonyl fluoride, 200 units of DNase I) for 1 g of cell pellet using a French Press (3 ϫ 15,000 psi), and the extract was centrifuged at 30,000 ϫ g for 20 min. The cell debris pellet was extracted 2 more times with 5 ml of extraction buffer. All protein-containing supernatants were pooled, applied on the His-Bind TM (Ni 2ϩ resin) column, and purified as recommended by Novagen. Ammonium sulfate was added to purified protein to a final concentration of 1.0 M and the solution was centrifuged at 16,500 ϫ g for 15 min. The supernatant was applied to HiTrap Phenyl-HP (Amersham Biosciences) column and the protein was further purified by fast protein liquid chromatography using a 0.8 to 0.2 M ammonium sulfate (pH 7.0) gradient. Purified ␤-glucosidase preparations were checked for purity by SDS-PAGE on a standard 10% polyacrylamide gel under reducing conditions (19) and their protein concentration was determined using the Bio-Rad protein assay (Bio-Rad) and bovine serum albumin as standard.
Enzyme Assays-Kinetic parameters, K m and k cat (V max /E tt ) for the wild-type enzymes (ZmGlu1 and SbDhr1), chimeric constructs (chimeras 16 and 21), and mutated ␤-glucosidases were determined by varying the substrate concentration from 0.078 to 25 mM in citrate-phosphate (100/50 mM) buffer (pH 5.8) for the artificial substrates oNPGlc and pNPGlc and from 0.039 to 2.5 mM for 4MUGlc. For the natural substrate dhurrin, the range of substrate concentration was from 0.01 to 2.5 mM. oNPGlc, pNPGlc, and 4MUGlc assays were performed in quadruplicate in a microcentrifuge tube for 10 min at 37°C by mixing 25 l of substrate with 25 l of enzyme solution (25 to 100 ng of enzyme), and the reaction was stopped by boiling the samples at 100°C for 3 min. Then, the glucose release was determined by the peroxidase/glucoseoxidase-coupled reaction (22) by adding 100 l of peroxidase/glucoseoxidase enzymes and 50 l of ABTS (2,2Ј-azinobis-3-ethylbenzthiazolinesulfonic acid). The reaction mixture was incubated at 37°C for 15 min, and the absorbance was read in a microtiter plate reader at 410 nm. Because dhurrin is not resistant to the enzyme-heat inactivation step, dhurrinase activity was monitored by mixing 25 l of substrate, 25 l of enzyme solution, 100 l of peroxidase/glucose/oxidase enzymes, and 50 l of ABTS in a microtiter plate. The absorbance was read at 410 nm after incubation at 37°C for 1 h.
Crystal Structure Determination of ZmGlu1-E191D,F198V in Complex with DIMBOA-Glucoside-The crystals were grown under similar conditions as the native enzyme (15). Two l of the protein solution was mixed with 2 l of reservoir containing 0.1 M Hepes (pH 7.5), 22% PEG 4000, and 5% isopropyl alcohol. The crystals belong to space group P2 1 and have unit cell parameters a ϭ 58.1 Å, b ϭ 114.1 Å, and c ϭ 80.1 Å, ␤ ϭ 93.88°. The ZmGlu1-E191D,F198V⅐DIMBOA-Glc complex was obtained in soaking experiments, in which 92 l of the crystallization solution was mixed with 5 l of glycerol (the cryoprotectant) and 3 l of ligand (10 mM) solution. Subsequently, 45.5 l of this mixture was supplemented with 2.5 l of glycerol and 2 l of ligand solution. The crystals were soaked in these two solutions for 1 h and then frozen in a stream of N 2 at 100 K. The crystals diffracted to a resolution of 1.9 Å.
Data were collected at the synchrotron ESRF (Grenoble, France) on beam-line ID14-EH4. The data set was treated with DENZO (23) and scaled with SCALA, part of the CCP4 package (24). The statistics on the data collection are given in Table II.
The molecules of the ZmGlu1-E191D,F198V mutant were positioned in the new unit cell with respect to the native ␤-glucosidase by molecular replacement using AMoRe (25) and the native maize ␤-glucosidase (Protein Data Bank accession code 1e1e) as the model. Two solutions were obtained, which correspond to the two molecules in the asymmetric unit related by a non-crystallographic 2-fold symmetry, with a correlation coefficient of 70% and an R factor of 31.1% in the resolution range of 15.0 to 4.2 Å. The complex was analyzed with SIGMAA (24) weighted F obs Ϫ F calc maps calculated with model phases. The final R and R free factors for the complexed ZmGlu1-E191D,F198V are 18.4 and 22.1%, respectively.
A model of the ZmGlu1-E191D,F198V mutant structure was produced with the homology modeling program MODELLER (26) starting with the crystal structure of ZmGlu1-E191D (Protein Data Bank code 1E4L) and introducing the single mutation F198V into the sequence to be modeled (data not shown).

N-terminal His-tagged Glu1
Is Fully Active in Vitro-Because our goal was the characterization of a large number of ␤-glucosidases, we chose to use polyhistidine-tagged ␤-glucosidases to speed up protein purification. Thus we first had to demonstrate that the addition of such a His 6 extension does not affect ␤-glucosidase stability and activity in vitro. For this, we chose the ZmGlu1 ␤-glucosidase, and cloned its cDNA into an expression plasmid, pET28a, using the NheI/XhoI restriction sites to fuse a polyhistidine extension at the N-terminal end. The recombinant protein was then produced in E. coli and purified on a Ni 2ϩ column. The in vitro activity of the His 6tagged and the native ZmGlu1 ␤-glucosidases were tested on the artificial substrates oNPGlc and pNPGlc. No significant differences in stability and kinetic parameters were observed between the 2 forms (data not shown). Subsequently, all ␤-glucosidases used in this study were produced with an N-terminal His 6 tag.
Substrate Specificity and Kinetics of Wild-type Enzymes and Site-directed Mutants on Artificial Substrates-The catalytic activity of wild-type and mutant enzymes was assayed toward three artificial substrates (oNPGlc, pNPGlc, and 4MUGlc) in solution and the kinetic parameters (K m , k cat , and k cat /K m ) were determined. The data are summarized in Table III.
The results obtained on Chim16 and Chim21 in this study are similar to those previously published (18) in that both chimeric enzymes show increased catalytic efficiency, as compared with the parental ZmGlu1, in pNPGlc hydrolysis (284 and 358% reported here, compared with 340 and 380% increase in the previous study (18)). However, the reported kinetic parameters of parental enzymes and chimeras are slightly different (i.e. for ZmGlu1: First, we analyzed the effect of substitutions at the three phenylalanines forming one wall of the aglycone binding site of the slot-like active site on the activity and substrate specificity of mutant enzymes. Of the three single mutants, the F205L mutant shows no significant effect on pNPGlc hydrolysis other than doubling the k cat . In contrast, the F198V and F466S single mutations had drastic effects on catalytic efficiency. For example, single mutant F466S exhibits a marked increase in relative efficiency in comparison to wild-type ZmGlu1 on all substrates tested (from 28 to 123%), because of increased substrate turnover (e.g. 4 times better for pNPGlc hydrolysis). As for the F198V mutant, the k cat of 4MUGlc hydrolysis is reduced by about 90% resulting in a 82% reduction in relative efficiency, whereas the hydrolysis of other artificial substrates is drastically reduced. For example, the F198V mutant is unable to hydrolyze pNPGlc and shows only a negligible activity toward oNPGlc (7% of the relative efficiency of the wild-type enzyme). Furthermore, adding the double F198V,F205L substitutions to Ch16 or Ch21 results in a 90% decrease in relative efficiency compared with the parental chimeric enzymes on all three substrates tested. It is also interesting to note that both k cat and K m are affected by these substitutions in the chimeric enzymes, where K m was increased by 2-fold for pNPGlc and up to 5-fold for oNPGlc with a concomitant decrease of the k cat . These results point out the critical role of Phe 198 in substrate recognition and binding by ZmGlu1, even though the generally observed decrease in activity stays context specific, which was also observed by Zouhar et al. (27). When reciprocal mutations (Dhr1-V196F and L203F) were made in SbDhr1, the mutant enzymes were still not able to hydrolyze any of the artificial substrates tested (cf . Table III).
Again, Chim16 and Chim21 show increased relative efficiency in the hydrolysis of all substrates tested. These chimeras introduce 9 and 4 amino acid substitutions, respectively, from SbDhr1 to the parental ZmGlu1, where the 4 substitutions of Chim21 are shared with Chim16 (Fig. 2). Among these substituted amino acids, only Phe 466 and Ala 467 are located in the aglycone binding site, although Tyr 473 is only indirectly but profoundly affecting substrate specificity because it forms a hydrogen bond with Trp 378 , which in turn is directly involved in aglycone binding. We therefore analyzed only the kinetic parameters of the F466S, A467S, and Y473F mutants. Among these residues, the A467S substitution in ZmGlu1 has no detectable effect on substrate specificity and catalytic efficiency in the case of pNPGlc or oNPGlc hydrolysis, and only a lower K m is obtained for 4MUGlc (0.49 mM F466S,A467S versus 0.79 mM for F466S). As mentioned before, the F466S substitution did not produce as clear effects on substrate specificity and catalytic efficiency as did the Y473F substitution. The single Y473F mutation in ZmGlu1 affects profoundly the catalytic behavior of ZmGlu1 because the k cat of this mutant is increased by 110 to 450% over the wild-type enzymes, depending on substrates tested. Nevertheless, this enhanced turnover rate is partly compensated by a slight increase of the K m of this mutant, giving a 40 to 250% better relative efficiency over the wild-type enzyme. Such enhanced activity is very much similar to that found for the Chim16 and Chim21, both of which include the Y473F substitution (Table III), leading us to conclude that the Tyr to Phe substitution at position 473 accounts for most if not all of the increased catalytic efficiency. It should be noted that the F469Y substitution in SbDhr1 does not produce any change in the catalytic efficiency of this mutant on the artificial substrates tested (Table IV).
Finally, additional mutations were made at residues Asp 261 , Met 263 , and Pro 377 that are found around the active site of ZmGlu1. Again, these amino acids were substituted by their homologs in SbDhr1 and their kinetic parameters were determined. Of these, the D261N substitution leads to a drastic decrease in relative efficiency of ZmGlu1 on all substrates tested, but little effect on the K m , whereas the M263F substitution had a slight effect on the k cat . This last mutation has opposite effects depending on the substrate considered because the relative efficiency on pNPGlc and 4MUGlc hydrolysis is reduced by about 40 and 50% whereas it shows an increase of about 60% on oNPGlc hydrolysis (Table III). This last result has also been observed by Zouhar et al. (27) on pNPGlc with a 70% decrease in the turnover rate of this mutant as opposed to 62% decrease in our study. At last, the P377A mutation leads to changes (decreases) in both k cat and K m with no decrease in the overall relative efficiency. The reciprocal substitution A375P in SbDhr1 does not yield any significant or detectable change in the artificial substrate hydrolysis.

Substrate Specificity and Kinetics of Wild-type Enzymes and Site-directed Mutants on the Natural Substrate Dhurrin-
The catalytic efficiency of the mutants described above was also tested on dhurrin (Table IV), the natural substrate of the sorghum SbDhr1. Dhurrin competitively inhibits maize Zm-Glu1 (19), and this enzyme hydrolyzes dhurrin with a lower catalytic efficiency only when a C-terminal octapeptide ( 466 FAGFTERY 477 ) is exchanged with the homologous octapeptide of SbDhr1 ( 462 SSGYTERF 469 ). This peptide substitution leading to amino acid substitutions at four sites (bold) is found in Chim21 and Chim16 (with an extra 22 amino acids substitution, cf. Fig. 2) enables ZmGlu1 to hydrolyze dhurrin (18) 2 with less than 10% of the relative efficiency of SbDhr1 (7.4% for Chim21 and 9.5% for Chim16). We investigated the role of 3 substitutions found in the above mentioned 8-amino acid long peptide fragment and located in Phe 466   and SbDhr1-A375P led to a drastic decrease of dhurrin hydrolysis, mainly because of a decrease in the turnover rate (75% decrease in both mutants), implicating these 2 residues in dhurrin specificity. These results are interesting because the ZmGlu1-F198V,F205L double mutant and ZmGlu1-P377A single mutant are not able to hydrolyze dhurrin, suggesting that aglycone specificity of ZmGlu1 and SbDhr1 reside in different amino acids within the aglycone binding site of the active site. Analysis of the Crystal Structure of ZmGlu1-E191D,F198V in Complex with DIMBOA-Glucoside-The structure at 1.9-Å resolution of ZmGlu1-E191D,F198V contains 980 residues, 493 water molecules, and two substrate molecules, one in the active site of each subunit of the homodimeric enzyme. The final R and R-free factors are 18.4 and 22.1%, respectively. The final electron density map clearly defined all atoms of the residues located in the active site pocket (Fig. 3a) and the difference Fourier map revealed the location of a substrate molecule (Fig.  3b). Interestingly, the glycone moiety was better defined by density in this structure than in the inactivated ZmGlu1-E191D, whereas the density, coming from the aglycone moiety is less visible here, indicating that the aglycone is possibly disordered. The overall structure of the mutant ZmGlu1-E191D-F198V is identical to the native structure (17) as well as to the simple mutant ZmGlu1-E191D (11). All forms of the enzymes have the classical (␤/␣) 8 barrel fold where ␤-strands and ␣-helices within each ␤/␣ repeat are connected by loops at the top of the barrel. The quaternary structure of ZmGlu1-E191D,F198V is a 120-kDa homodimer, as is its parental wildtype enzyme (15). Three structural differences are encountered, besides the point mutation introducing a valine residue into the aglycone pocket instead of a phenylalanine, in that two residues undergo a conformational change and the position of one residue is affected in this structure with respect to the structure of the single mutant ZmGlu1-E191D. These are Glu 464 , which forms a hydrogen bond to the OH6 group of the glycone moiety, and Phe 205 and Phe 466 in the aglycone pocket (Fig. 3c). In contrary, no such changes were observed in the model created by homology modeling of the single mutant. These conformational changes lead to a completely different position of the substrate molecule in the active site pocket of  the double mutant ZmGlu1-F198V,E191D structure, when compared with the structure of the complex ZmGlu1-E191D/ DIMBOA-glucoside (11) (Fig. 3d). Interestingly, conformational changes are observed for both the glycone binding pocket (Glu 464 ) as well as the aglycone binding pocket. Apparently, the phenylalanine in position 198 influences the arrangement of the three-residue cluster Phe 198 , Phe 205 , and Phe 466 , because both Phe 205 and Phe 466 change positions in the structure of the F198V mutant. The re-positioning of these residues influence the glutamic acid Glu 464 (two residues from Phe 466 ), which also changes its relative position within the glycone binding pocket. As a consequence to all these rearrangements, the substrate apparently binds differently in the glycone binding pocket, is rather flexible in the aglycone pocket, and the glycosidic bond is not correctly positioned for the catalytic cleavage to take place (Fig. 3b). DISCUSSION In this study, we have produced site-directed mutants of a maize (ZmGlu1) and a sorghum (SbDhr1) ␤-glucosidase and we have determined the structure of the most affected mutant of ZmGlu1 in complex with the natural substrate. These two enzymes share 70% sequence identity and represent extremes in substrate specificity (18) in that the former hydrolyzes a wide spectrum of substrates in addition to its natural substrate DIMBOA-Glc, whereas the latter hydrolyzes its natural sub-  . The map is a SIGMAA weighted 2F obs Ϫ F calc synthesis contoured at 1, showing the well defined density around the residues undergoing a positional shift in this mutant structure. b, the initial difference Fourier map (F obs Ϫ F calc ) located in the active site pocket, contoured at a 2 level, showing the presence of a substrate molecule. c, structural superimposition of ZmGlu1-E191D (in blue) and ZmGlu1-E191D, F198V (the residues are color coded with standard atom-type colors), both in complex with DIMBOA-Glc (omitted from this figure). The F198V mutation (highlighted in blue) induces a positional change of Phe 205 , Phe 466 takes a different rotameric conformation, and Glu 464 in the glycone binding pocket is also displaced. d, the same structural superimposition as in c, viewed in a different orientation, illustrating the drastic change in the position of the DIMBOA-glucoside molecule in ZmGlu1-E191D,F198V (red substrate molecule and atom type colors for the residues) with respect to ZmGlu1-E191D (dark blue, substrate molecule, and blue, residues). strate dhurrin only. The biochemical characterization of these mutants showed that some exhibit a novel substrate specificity by hydrolyzing dhurrin, whereas others exhibit improved or reduced catalytic efficiency on certain substrates, providing new and significant information on mechanism of substrate (i.e. aglycone) specificity. For example, it was found that a single amino acid substitution (Y473F) in ZmGlu1 enabled the hydrolysis of dhurrin, which is not hydrolyzed by wild-type ZmGlu1. The same mutant also improved catalytic efficiency toward such artificial substrates as pNPGlc. The level of hydrolysis is comparable with those of Chim16 and Chim21, indicating that this amino acid might be the critical residue in the ability of Chim16 and Ch21 to hydrolyze dhurrin. The higher K m found for these 2 chimeras (101 and 76 M, respectively) compared with that of ZmGlu1-Y473F (47 M) might be because of the other substitutions, namely A467S (with a K m of 92 M). The improved catalytic efficiency provoked by the Y473F mutation may be explained by its role in stabilizing Trp 378 . In the native ZmGlu1 structure Tyr 473 is involved (via the phenolic OH group) in a hydrogen bond with the amide group of Trp 378 . The side chain of this latter residue therefore has no conformational freedom. Changing Tyr 473 into a phenylalanine removes this hydrogen bond and the side chain of Trp 378 may be more flexible and therefore adapts better to the different orientations of the aglycones of artificial substrates and dhurrin. The influence of the flexibility of the Trp 378 side chain on the binding of substrates is also demonstrated by the mutation of Pro 377 by an alanine, introducing a higher flexibility to the loop carrying Trp 378 , which leads to a decrease of K m for artificial substrates of the enzyme ZmGlu1-P377A, and a lower k cat as a consequence of a bad positioning of the glycosidic bond.
In contrast, another single amino acid substitution (F198V) in the same enzyme resulted in an almost inactive ␤-glucosidase (Tables III and IV). The structural analysis shows that the presence of valine instead of phenylalanine at position 198 induces a rearrangement of three amino acids, namely Phe 205 , Phe 466 , and Glu 464 , which are all involved in either the glycone or aglycone binding pocket. This rearrangement leads to a different binding mode of the substrate and most probably is the reason for the loss of catalytic activity upon mutation. Our structural study of the F198V,E191D mutant clearly points to the limits of molecular modeling, because the cascade of rearrangements would have been difficult to predict by a model without taking into account very local and very precise energetic terms. The model calculated with a standard homology modeling algorithm did not reveal these conformational changes for the point mutation F198V.
It is interesting to note that, whereas the F198V substitution has a drastic effect on ZmGlu1 activity, the reverse substitution has no effect on SbDhr1 for the substrates hydrolyzed by the maize enzyme. It appears clear that more than one or two substitutions (the V196F,L203F has no effect) are required to change SbDhr1 substrate specificity. Therefore it seems likely that other amino acids, whether located in the active site or not, are responsible for the strict specificity in SbDhr1, whereas the corresponding residues have little or no role in ZmGlu1 substrate specificity. The structural analysis of Sb-Dhr1 and its inactive mutant in complex with dhurrin, which are in progress, will shed more light on this aspect.
The mutation changing Asp 261 into an asparagine mainly has an effect on the catalytic efficiency versus artificial substrates. In the structure of ZmGlu1, Asp 261 is close to the catalytic acid/base Glu 191 and displays no hydrogen bonds either with the glucose or the aglycone. Therefore, the loss of catalytic efficiency may be explained by a stabilizing role of Asp 261 on the different protonation states of the acid/base catalyst. Interestingly, a recent mutational study of a human cytosolic ␤-glucosidase has also demonstrated that this position is critical for catalysis to take place in human cytosolic ␤-glucosidase, even though it is a phenylalanine (Phe 225 in human cytosolic ␤-glucosidase) in that case (28). Apparently, this position plays an important role in maintaining the local environment of the acid/base catalytic glutamate optimal, counterbalancing the effects of the other surrounding residues.
The site-directed mutagenesis approach used here, whereas giving interesting information on the precise role of individual residues on substrate specificity, reached a limit in the goal of transforming ZmGlu1 into SbDhr1 or vice versa. To attain such a transformation, it is also necessary to know the critical determining residues in SbDhr1. However, from our study it appears most likely that more than a handful of residue substitutions are necessary for the inversion of specificity and that local influences of these substitutions are difficult to predict. Whereas site-directed mutagenesis has been shown to be very efficient for removing chemically "essential" residues, it often led to altered catalytic mechanisms instead of abolished activity, because of the plasticity in enzyme active site (for review, see Refs. 29 and 30). Such "malleability" could be expected for non-chemically essential residues, as the ones involved in substrate specificity. To counteract the plurality of residues that might be involved in this specificity, a strategy using directed evolution, instead of site-directed mutagenesis, could reasonably be developed. This technique has been successful in changing a ␤-galactosidase in a ␤-fucosidase (31) or improving the low-temperature catalysis in the hyperthermostable Pyrococcus furiosus ␤-glucosidase CelB (32). For these reasons, we think that directed evolution between SbDhr1 and ZmGlu1 will speed up the process of understanding the aglycone specificity in these enzymes, and ␤-glucosidases in general.