Mutational Analysis of the Major Loop of Bacillus1,3-1,4-β-d-Glucan 4-Glucanohydrolases

The carbohydrate-binding cleft of Bacillus licheniformis 1,3-1,4-β-d-glucan 4-glucanohydrolase is partially covered by the surface loop between residues 51 and 67, which is linked to β-strand-(87–95) of the minor β-sheet III of the protein core by a single disulfide bond at Cys61–Cys90. An alanine scanning mutagenesis approach has been applied to analyze the role of loop residues from Asp51 to Arg64 in substrate binding and stability by means of equilibrium urea denaturation, enzyme thermotolerance, and kinetics. The ΔΔG Ubetween oxidized and reduced forms is approximately constant for all mutants, with a contribution of 5.3 ± 0.2 kcal·mol−1 for the disulfide bridge to protein stability. A good correlation is observed between ΔG U values by reversible unfolding and enzyme thermotolerance. The N57A mutant, however, is more thermotolerant than the wild-type enzyme, whereas it is slightly less stable to reversible urea denaturation. Mutants with a <2-fold increase inK m correspond to mutations at residues not involved in substrate binding, for which the reduction in catalytic efficiency (k cat/K m ) is proportional to the loss of stability relative to the wild-type enzyme. Y53A, N55A, F59A, and W63A, on the other hand, show a pronounced effect on catalytic efficiency, with K m > 2-fold andk cat < 5% of the wild-type values. These mutated residues are directly involved in substrate binding or in hydrophobic packing of the loop. Interestingly, the mutation M58A yields an enzyme that is more active than the wild-type enzyme (7-fold increase in k cat), but it is slightly less stable.

Bacillus licheniformis 1,3-1,4-␤-glucanase is a retaining glycosidase (2), acting by general acid/base catalysis in a double displacement mechanism (15). Glu 138 has been proposed as the proton donor residue and Glu 134 as the catalytic nucleophile (16,17). The three-dimensional structure, recently refined at 0.18-nm resolution by x-ray crystallography (18), is almost identical to that of the hybrid H(A16M) between Bacillus amyloliquefaciens and Bacillus macerans (19) and the B. macerans (20) enzymes. It has a jelly-roll ␤-sandwich fold, with the carbohydrate-binding cleft located on the concave face of a ␤-sheet formed by seven antiparallel ␤-strands (see Fig. 1). The Bacillus enzymes are unrelated to the plant 1,3-1,4-␤-glucanases in both sequence similarity (family 17 of glycosylhydrolases) and three-dimensional structure (␣/␤-barrel structure), clearly indicating that the identical substrate specificities have arisen by convergent evolution (21). On the other hand, the Bacillus enzymes show structural similarities to plant legume lectins and family 7 cellulases. Cellobiohydrolase I from Trichoderma reesei (22) has a very similar fold, with most of the ␤-sandwich residues in the protein core being superimposable, but it has long loops shaping the substratebinding tunnel that are missing in the 1,3-1,4-␤-glucanase structure.
Except for the Bacillus brevis isozyme, all Bacillus 1,3-1,4-␤-glucanases possess a single disulfide bond at Cys 61 -Cys 90 (B. licheniformis numbering) that connects a ␤-strand (residues 87-95) with a loop from residues 51 to 67 (see Fig. 1). This major loop is located on the concave side of the molecule, is solvent-exposed, and partially covers the active-site cleft. Even though no three-dimensional structure of an enzyme-inhibitor complex with a carbohydrate inhibitor filling the entire binding cleft is yet available, the three-dimensional structure of a covalent complex between the hybrid H(A16M) and epoxybutyl ␤-cellobioside (19) and the molecular model of an enzymesubstrate complex made by computational methods (23) indicate that some loop residues might interact with a substrate occupying distant subsites on the nonreducing end of the binding site cleft.
Here we use the technique of alanine scanning mutagenesis (24) to analyze the role of loop residues (from Asp 51 to Arg 64 ) in B. licheniformis 1,3-1,4-␤-glucanase in substrate binding and stability by means of equilibrium urea denaturation, enzyme thermotolerance, and kinetics. Previous studies of the disulfide bond at Cys 61 -Cys 90 have shown the deleterious effect of cysteine-to-alanine mutations on protein stability and activity, but no effect of disulfide bond reduction on activity (25). These results suggested that the loop has little flexibility and that the disulfide bond is not required to keep the structural integrity of the loop. Other hydrophobic interactions may position the loop to shape the active-site cleft.
Chemicals and Enzymes-Urea (molecular biology-grade) was purchased from Sigma; dithiothreitol, 3,5-dinitrosalicilic acid, and 5,5Јdithiobis(2-nitrobenzoic acid) were from Fluka. Restriction endonucleases and T4 DNA ligase were from Boehringer Mannheim, and DeepVent ® polymerase was from New England Biolabs Inc. ␣-35 S-ATP was purchased from Amersham Corp. DNA sequencing was performed with the T7 sequencing kit from Pharmacia Biotech Inc. Oligonucleotides were synthesized by Boehringer Mannheim. Barley ␤-glucan was from Megazyme (Sydney, Australia). All buffers and solutions for kinetic and urea denaturation experiments were degassed prior to use.
Protein Expression and Purification of Wild-type and Mutant Enzymes-Proteins were purified from the supernatant of E. coli TG1 cultures harboring the mutagenized plasmids basically as described before (28) with an additional purification step of fast protein liquid chromatography on an ion-exchange TSK CM-3SW column in 5 mM acetate buffer, pH 5.6, and elution with a linear gradient of 0 -0.4 M NaCl in the same buffer. The proteins were analyzed by SDS-polyacrylamide gel electrophoresis as described (29) and by fast protein liquid chromatography on a TSK CM-3SW column at pH 5.6. Enzyme concentrations were determined by absorbance at 280 nm using A 1 mg/ml ϭ 14.5 absorbance units for the wild-type enzyme and by the Bradford protein assay (30) for the mutants using the wild-type enzyme as a standard. Spectrophotometric and kinetic measurements were performed on a Varian Cary 4 spectrophotometer with a Peltier temperature control system.
Equilibrium Urea Denaturation-Unfolding was monitored by fluorescence spectroscopy in a Perkin-Elmer LS50 spectrofluorometer, with excitation at 282 nm (3-nm slit) and the emission spectra being recorded from 270 to 440 nm (8-nm slit) and measured at 340 nm, in thermostatted cuvette holders at 37°C. For each data point collected, wild-type or mutant 1,3-1,4-␤-glucanases in citrate/phosphate buffer, pH 7.2, were diluted to 1 g/ml in degassed urea solution in the same buffer and incubated overnight at 37°C. To obtain reduced enzymes, the protein stock solutions (25 g/ml) were made 200 mM dithiothreitol and incubated for 30 min at 25°C before being added to the denaturant solution. The final dithiothreitol concentration was 10 mM, enough to avoid reoxidation for at least 24 h as shown by free sulfhydryl titration with 5,5Ј-dithiobis(2-nitrobenzoic acid) (after removal of the excess dithiothreitol by means of a pD10 G25 desalting column from Pharmacia).
Enzyme Inactivation and Thermotolerance Measurements-For determination of enzyme thermotolerance, samples of 50 g/ml enzyme in 50 mM sodium acetate buffer, pH 6.0, 20 mM CaCl 2 were incubated at 65 or 70°C. Aliquots of 80 l were withdrawn at various time intervals (until complete inactivation) and immediately diluted 5-fold in ice-cold water. The residual activity was determined at 45°C using 4-methylumbelliferyl 3-O-␤-cellobiosyl-␤-D-glucopyranoside (3 mM assay concentration) and barley ␤-glucan (5 mg/ml assay concentration) in citrate/ phosphate buffer, pH 7.2, 0.1 mM CaCl 2 . The enzymatic half-life (t 50 ) was calculated by fitting the first phase of the plot residual activity versus incubation time to a single exponential decay.

RESULTS
Enzyme Expression and Purification-Point mutations to alanine in the loop residues from Asp 51 to Arg 64 ( Fig. 1) were prepared by site-directed mutagenesis by PCR. The mutant proteins were purified up to 95% as judged by SDS-polyacrylamide gel electrophoresis following the procedure described for the wild-type enzyme (28). Expression and purification yields were similar for all the mutant and wild-type enzymes. Proteins were stored in their oxidized form. Reduction of the disulfide bond at Cys 61 -Cys 90 (reduced enzymes) was done just before their use as described under "Materials and Methods." Analysis of Stability by Equilibrium Urea Denaturation-The stability of the enzymes reported in this study was examined by urea denaturation assuming a two-state transition using the model of Clarke and Fersht (35). Unfolding was monitored by measuring the dependence of fluorescence intensity on urea concentration. The data were analyzed as described previously for other ␤-glucanase mutants (25) (49).
where F is the measured fluorescence, a F and a U are the intercepts and b F and b U are the slopes of the base lines at low ( . Previous experiments with the wild-type enzyme (25) have shown the urea-induced denaturation to be reversible and independent of protein concentration.
The calculated values for m and [D] 50% are given in Table I This difference has proved to be significant by a Student's t test (⑀ ϭ 0.05), indicating that the presence or absence of the disul-fide bond has a significant and constant effect on the unfolding behavior.
Since individual m values for each mutant are subjected to large standard errors, we used the corresponding m av value to calculate the free energies of unfolding in the absence of denaturant for the oxidized and reduced enzymes, respectively. Then, Equation 2 becomes Equation 3.
The difference in stability between two enzymes is evaluated as shown in Equation 4, where a and b are the mutant and wild-type enzymes, respectively, in their oxidized or reduced forms (⌬⌬G U(ox) or ⌬⌬G U(red) ). The calculated ⌬G U H2 O and ⌬⌬G U values are listed in Table II.
The energetic contribution of the disulfide bridge to stability (⌬⌬G U S-S ) was calculated using Equation 4 in which a and b are the reduced and oxidized forms of the same protein, respectively. The calculated values in Table II (36), some of the alanine mutants are more thermolabile. 45°C was found to be the highest temperature for which all the proteins studied showed a linear progress curve during the initial 15 min of reaction. Substrate inhibition was observed at high concentrations, so the data were fitted to an uncompetitive substrate inhibition model by nonlinear regression (36). Calculated values for k cat and K m are summarized in Table III. Analysis of Enzyme Thermal Stability-A measure of the enzyme thermotolerance can be obtained by deducing t 50 at a specified temperature. Residual activity of the enzymes was measured after various periods of incubation at a given temperature by steady-state kinetics with 4-methylumbelliferyl 3-O-␤-cellobiosyl-␤-D-glucopyranoside substrate at 45°C. Preliminary experiments with the wild-type enzyme at 50 and 500 g⅐ml Ϫ1 at 65 and 70°C in sodium acetate buffer, pH 6.0, showed that extensive protein aggregation took place at high enzyme concentration and that a very low t 50 (Ͻ10 min) was obtained at 70°C. Therefore, we chose a protein concentration of 50 g⅐ml Ϫ1 and an incubation temperature of 65°C as standard assay conditions. The plot of residual activity versus incubation time follows a double exponential curve, with the value of t 50 being in the first phase of the inactivation decay. Mutants Y53A, N55A, F59A, and W63A could not be analyzed under these conditions due to their low activity, which required enzyme concentrations above 500 g⅐ml Ϫ1 . Values of t 50 are summarized in Table I. DISCUSSION The carbohydrate-binding cleft of the 1,3-1,4-␤-glucanase of B. licheniformis is partially covered by the surface loop between residues 51 and 67, which is linked to ␤-strand-(87-95) of the minor ␤-sheet III (18) by the single disulfide bridge at Cys 61 -Cys 90 . The technique of alanine scanning mutagenesis has been applied to analyze the role of loop residues (Asp 51 -Arg 64 ) in substrate binding and stability as well as the contribution of the disulfide bridge to stability.
Equilibrium Urea Denaturation-Unfolding transition curves are described by two parameters in a two-state model: [D] 50% is a measure of the midpoint of the transition region, and m is a measure of the steepness of the transition region and reflects the cooperativity of the unfolding process. A clear distinction is observed between reduced and oxidized forms in terms of m values, with the oxidized enzymes having a steeper transition. Common to all models that have been proposed to describe the dependence of the free energy of unfolding on denaturant concentration (37,38) is the premise that denaturants alter the equilibrium N 7 U through a preferential interaction with the denatured state. Schellman (37) proposed that the parameter that can describe the differential interaction of the native and denatured state to denaturants is the different solvent-accessible area between both states, A N and A U , respectively (Equation 5),

TABLE III Kinetic parameters for wild-type and mutant 1,3-1,4-␤-glucanases (4methylumbelliferyl 3-O-␤-cellobiosyl-␤-D-glucopyranoside as substrate)
Conditions were as follows: citrate/phosphate buffer, pH 7.   where K represents a thermodynamic constant. According to this model, the smaller m value of the reduced enzymes (Table  I) must reflect a smaller value of (A U Ϫ A N ), only explained by a decrease in A U (as large variations in A N are not expected). However, the reduction of the disulfide bridge is more likely to produce an increase in A U , accounted for by a more extended denatured conformation (39,40). Thus, the results are unlikely to be explained in this way and reveal a more complex meaning of the m parameter.
The spatial distribution of the mutated residues in the crystallographic structure of the wild-type enzyme (18) suggests that the destabilizing effect is larger near the N-and C-terminal ends of the loop (Table IV). This observation is in agreement with the idea that the loop edges are rigid, with a central part being more flexible and the C-terminal end being more tightly packed as judged by the side chain solvent accessibility and van der Waals interaction data shown in Table IV. No correlation was found between the experimental ⌬⌬G U values and the free energy of transfer of amino acid side chains from water to octanol (corrected or not for solvent-exposed area) or the number of atoms inside a sphere around C-␣ of the mutated amino acid residue (Table IV). Such correlations have been shown to work properly in a number of proteins (41)(42)(43)(44) for series of mutants in hydrophobic regions of the protein structure. This is not the case for the 1,3-1,4-␤-glucanase mutants probably because the loop is partially solvent-exposed and some of the residues are hydrophilic.
Contribution of Disulfide Bridge to Protein Stability-The calculated values for ⌬⌬G U S-S in Table III indicate a constant stabilizing effect of 5.3 Ϯ 0.2 kcal⅐mol Ϫ1 for the disulfide bridge in all mutant and wild-type enzymes. This value is larger than that previously estimated for the wild-type enzyme (0.7 kcal⅐mol Ϫ1 (25)). The m value for the oxidized wild-type enzyme deviates from the general trend observed for the mutants, and it is much closer to the m value for the reduced wild-type enzyme. In the absence of mutant data, our first estimation was performed using an m av value of 1.31 kcal⅐mol Ϫ1 ⅐M Ϫ1 for both forms of the wild-type enzyme. However, the large number of mutants studied here clearly shows a significant difference between oxidized and reduced forms. Even though the behavior of the wild-type enzyme might be different, the general trend observed here allows us to conclude that the disulfide bond has a larger contribution to protein stability.
Thermotolerance-Enzyme thermotolerance was determined at 65°C as the incubation time required to irreversibly inactivate the enzyme to 50% of its initial activity (t 50 ). A good correlation was observed between ⌬G U (from equilibrium urea denaturation) and t 50 for the mutants in their oxidized form (Fig. 3), except for N57A, which is surprisingly more thermotolerant than the wild-type enzyme. Even though direct comparison of thermal stability and urea denaturation is not possible in general (kinetic versus equilibrium experiments), the results may be rationalized considering a fast irreversible process from the denatured state at high temperature (Reaction 1), where k 1 , k 2 , and k 3 are rate constants and I is the irreversibly denatured state. Recent examples have shown that mutations in mobile loops or at labile residues may yield proteins that are more resistant to thermal denaturation, whereas reversible denaturation is similar to the wild-type proteins (45)(46)(47). The higher thermotolerance of the N57A mutant and its deviation from the above correlation may be the result of lowering the irreversible denaturation rate constant k 3 as a consequence of removing a labile solvent-exposed Asn residue. Asparagine residues are often involved in several degradative covalent reactions in proteins such as deamidation, isoaspartate formation, and peptide bond cleavage at high temperatures or in low pH environments, with the effect being more pronounced when the Asn residue is next to a Gly residue in the amino acid sequence (48). Asn 57 is next to Gly 56 in a highly solvent-exposed region of the loop (Table IV). Another Asn mutant in the loop (N60A) fits the correlation between ⌬G U and t 50 , but it is a buried residue with no Gly residue next to it. Finally, Asn 55 is also solventexposed, but the thermotolerance of the mutant N55A could not be determined due to its low activity. For the purpose of engi- a Percentage of buried amino acid side chain in the wild-type enzyme calculated from Connolly surfaces on the three-dimensional structure. b Observed hydrogen bond interactions (distance (Å) between heteroatoms in parentheses). c Amino acid residues having at least one atom with van der Waals contact with the side chain of the residue being mutated.
neering more heat-stable enzymes, preventing irreversible thermal inactivation may be more important than increasing stability.
Effects on Enzyme Kinetics-The k cat and K m values of the 1,3-1,4-␤-glucanase mutants (Table II) show that most of the mutations have an effect on enzyme activity. It could be a direct effect of removing an amino acid side chain involved in substrate binding (or interacting with an essential catalytic residue) or an indirect effect of local rearrangements produced by the mutation (which are also reflected in a decrease in protein stability). k cat /K m values are plotted against stability data (as ⌬G U H2 O values) for the oxidized enzyme forms in Fig. 4. Inspection of this plot suggests that the mutants can be classified in four groups. Group A (wild-type, D51A, G52A, G56A, N60A, T62A, and R64A) is formed by those enzymes showing a good correlation between catalytic efficiency and enzyme stability. For these mutants, the decrease in catalytic efficiency is mainly due to k cat since K m values are Ͻ2-fold larger than the wildtype K m value. Moreover, the mutated residues in this group have no specific role in substrate binding as proposed from the structure of the modeled enzyme-substrate complex. Therefore, the reduction in k cat /K m is interpreted as the result of local rearrangements in the protein structure induced by the mutations, which also have a proportional effect on protein stability. Group B mutants (S54A and N57A) slightly deviate from the correlation. They show Ͻ2-fold reduction in k cat and almost no effect on K m as compared with the wild-type enzyme. Group C (Y53A, N55A, F59A, and W63A) is composed of mutants that have a pronounced deleterious effect on enzyme activity, with K m Ͼ 2-fold and k cat Ͻ 5% of the wild-type kinetic parameters. These mutated residues are directly involved in substrate binding or in hydrophobic packing of the loop. 2 Tyr 53 forms a hydrogen bond with the 3-OH of the glucopyranose unit of the substrate in subsite ϪIII, whereas the amide nitrogen of Asn 55 hydrogen-bonds with the 6-OH of the glucopyranose unit in subsite ϪII according to the modeled enzyme-substrate complex for the B. licheniformis enzyme 2 or the modeled complex for the hybrid H(A16M) ␤-glucanase (23). On the other hand, Phe 59 has the aromatic side chain pointing toward the core of the protein, and it has a strong hydrophobic (stacking) interaction with Trp 213 , which belongs to a ␤-chain of the major ␤-sheet on the concave face of the molecule. This interaction might be important to position the loop and to create a hydrophobic environment in this portion of the cleft when the sub-strate binds. Trp 63 also has the aromatic side chain interacting with the main core and very close to Phe 59 , contributing to the structural integrity of the loop. Finally, group D comprises the single mutant M58A, which is more active, with a k cat value 7-fold higher than that of the wild-type enzyme. This surprising result was unpredictable from a simple structural analysis since the side chain of Met 58 does not interact with the substrate or with any essential catalytic residue in the threedimensional structure of the free enzyme. However, replacement of the side chain by a smaller methyl group (M58A) might allow some readjustments of the loop in the enzyme-substrate complex that may have a favorable effect on transition state stabilization. Up to date, only the structure of a covalent complex between the hybrid H(A16M) 1,3-1,4-␤-glucanase and epoxybutyl ␤-cellobioside has been solved by x-ray crystallography (19). Inhibitor binding has no significant effect on the active-site geometry as observed by comparing the structures of the covalent complex and the free H(A16M) enzyme. However, the cellobiose unit of this suicide inhibitor only fills subsites ϪII and ϪIII, with subsite ϪI being occupied by an alkyl chain instead of a glucopyranose ring. Since ring distortion in subsite ϪI is expected, the structure of this covalent complex is not a good model to analyze the structural effects that the reported mutations may have on substrate binding and on transition state stabilization. New three-dimensional structures of enzyme-inhibitor complexes (or inactive mutant-substrate complexes) are required to evaluate small but significant structural changes that might occur upon ligand binding.
When applying an alanine scanning mutagenesis strategy, comparison between catalytic efficiency and enzyme stability provides a useful method to identify those residues that have an important role in ligand binding or in structural packing of the protein. Taking as a reference the mutants for which the reduction in catalytic efficiency is proportional to the loss of protein stability, mutations that deviate from this correlation indicate that these residues are involved in substrate binding or in maintaining the active-site structure. It is remarkable that two mutants, N57A with increased thermostability and M58A with higher catalytic efficiency, have been obtained. The effects of these mutations were unpredictable with the current knowledge of protein structure/function relationships, supporting the fact that scanning and random mutagenesis strategies are useful approaches to obtain proteins with improved properties for biotechnological applications.  from Table II), and t 50 values are the enzymatic half-lives at 65°C (data from Table I). wt, wild-type enzyme.  from Table III), and ⌬⌬G U values are from Table II. wt, wild-type enzyme.