Thermal Stabilization of Bacillus subtilis Family-11 Xylanase by Directed Evolution*

We used directed evolution to enhance the thermostability of glycosyl hydrolase family-11 xylanase from Bacillus subtilis. By combining random point mutagenesis, saturation mutagenesis, and DNA shuffling, a thermostable variant, Xylst, was identified which contained three amino acid substitutions: Q7H, N8F, and S179C. The half-inactivation temperature (the midpoint of the melting curves) for the Xylst variant compared with the wild-type enzyme after incubation for 10 min was elevated from 58 to 68 °C. At 60 °C the wild-type enzyme was inactivated within 5 min, but Xylst retained full activity for at least 2 h. The stabilization was accompanied by evidence of thermophilicity; that is, an increase in the optimal reaction temperature from 55 to 65 °C and lower activity at low temperatures and higher activity at higher temperatures relative to wild type. To elucidate the mechanism of thermal stabilization, three-dimensional structures were determined for the wild-type and Xylst enzymes. A cavity was identified around Gln-7/Asn-8 in wild type that was filled with bulky, hydrophobic residues in Xylst. This site was not identified by previous approaches, but directed evolution identified the region as a weak point. Formation of an intermolecular disulfide bridge via Cys-179 was observed between monomers in Xylst. However, the stability was essentially the same in the presence and absence of a reducing agent, indicating that the increased hydrophobicity around the Cys-179 accounted for the stability.

Xylanases (EC 3.2.1.8) catalyze the hydrolysis of xylan, a major constituent of hemicellulose. The enzyme is the focus of much attention because of its potential for use in industrial processes, including paper and pulp industries and food and feed industries (1,2). Many efforts have been made to improve the properties of the glycosyl hydrolase family-11 (previously family G) xylanase to handle industrial tasks. In particular, thermostability is a major target of such modifications (3)(4)(5)(6)(7). However, the majority of these studies have been performed by sitedirected mutagenesis based on high resolution three-dimensional structures, and solutions that identify functional changes are well beyond our predictive ability with these approaches (8,9). We, therefore, used a directed evolution technique, an alternative approach to protein engineering that can complement the weaknesses of the sitedirected approach (8). Our approach included a global search for weakness using random mutagenesis, an extensive search for the best-fit residue by saturation mutagenesis, and rapid fitness improvement by DNA shuffling. By using these techniques, we attempted to identify positions that confer thermostability and that have not been identified by other methods.
Directed evolution does not require three-dimensional structures but does require a quick and sensitive screening system (10). From this point of view, the high-throughput screening systems developed for this enzyme are a poor fit except for the clear zone (halo) assay (11). With this assay, xylanase-producing bacteria form a clear zone (halo) around the colonies when grown on agar plates containing xylan (or dye-coupled xylan such as Remazol Brilliant Blue to increase sensitivity). Formation of a complex between the dyes and sugars enables us to detect using the naked eye enzyme-producing colonies as a clear zone. Although these assays are simple and quick, the method is limited since the activity needs to be detected under physiological conditions. Therefore, we first attempted to establish a more flexible high-throughput xylanase activity screening system. To achieve this, we modified the dinitrosalicylic acid (DNS) 2 method (12) to a 96-well format. We also used a commercial mild detergent, the BugBuster TM reagent (Novagen), for efficient and consistent extraction of enzymes from Escherichia coli. After setting up reliable screening conditions, we applied directed evolution and screened for increased thermostability. We successfully obtained a thermostable variant that retained full activity at 60°C for more than 2 h, whereas the wild type inactivated in ϳ5 min.

EXPERIMENTAL PROCEDURES
Reagents-Taq DNA polymerase was purchased from Takara BIO (Shiga, Japan); PfuTurbo was purchased from Stratagene (La Jolla, CA); DpnI was purchased from New England Biolabs (Beverley, MA); CircleGrow was purchased from BIO101 (Carlsbad, CA); BugBuster TM reagent and benzonase nuclease were purchased from Novagen (Madison, WI); the Diversify TM PCR random mutagenesis kit was purchased from Clontech (Mountain View, CA); oligonucleotides were purchased from Japan Bio Services (Saitama, Japan); competent E. coli JM109 cells were purchased from Toyobo (Osaka, Japan). Soluble birch wood xylan was kindly provided by Dr. Kazuhiko Ishikawa.
Mutagenesis-Plasmid pBsx2, which carries a thermolabile variant (a G510A nucleotide substitution occurs which causes a Met-169 to Ile amino acid substitution) of Bacillus subtilis xylanase (13), was cloned into pTD-tac (14). * 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. Error-prone PCR was carried out using a Clontech Diversify TM PCR random mutagenesis kit. A mutation condition, which should generate an error frequency of ϳ4 -5 base substitutions per 1000 bases per gene copy, was used. The reaction mixture contained 40 mM Tricine-KOH (pH 8.0), 16 mM KCl, 3.5 mM MgCl 2 , 3.75 g/ml bovine serum albumin, 0.64 mM MnSO 4 , 40 M dGTP, 1ϫ Diversify TM dNTP mix, 10 M concentrations each of flanking primers, 1 ng of pBsx2, and 5 units of Taq polymerase in a total volume of 50 l. The mixture was heated at 94°C for 30 s followed by 25 cycles of incubation at 94°C for 30 s and 68°C for 1 min. Products were purified and cloned back into the pTDtac vector. The plasmid library thus obtained was used to transform competent E. coli JM109.
Site-saturation mutagenesis was carried out using a pair of oligonucleotide primers. The sequences of oligonucleotides are listed in Table 1. The target amino acid position was coded by NNK (sense strand) and MNN (antisense strand), where N ϭ A, G, C, or T, K ϭ G or T, and M ϭ A or C. The reaction was carried out following the Stratagene QuikChange TM site-directed mutagenesis protocol (15). Briefly, a pair of primers was mixed in 1 ϫ PfuTurbo buffer, 50 ng of pBsx2, 10 M concentrations each of primer, 0.2 mM concentrations of each dNTP, and 2.5 units of PfuTurbo DNA polymerase in a total volume of 50 l. The solution was heated for 30 s at 95°C followed by 20 rounds of incubation at 95°C for 30 s, 55°C for 1 min, and 68°C for 10 min. After cooling down the mixture to 4°C, 10 units of DpnI was added to the products and incubated at 37°C for 2 h to digest the template pBsx2. The mixture was used to transform competent E. coli JM109.
DNA shuffling was carried out using the StEP method (16). The reaction contained ϳ0.2 g each of parent plasmid DNA, 1ϫ Taq buffer, 0.2 mM concentrations each of dNTP, 0.15 M concentrations of each flanking primer, and 2.5 units of Taq polymerase in a total volume of 50 l. The mixture was heated at 95°C for 5 min and then subjected to 80 rounds of thermal cycling at 94°C for 30 s and 55°C for 15 s in a Stratagene Robocycler. The product was purified and cloned back into the pTD-tac. The plasmid library thus obtained was used to transform competent E. coli JM109.
Library Screening-E. coli JM109 transformants were picked with a sterile toothpick and resuspended in separate wells of a 96-deep well flat-bottom block (Qiagen, Hilden, Germany) containing 1.2 ml of Cir-cleGrow and 100 g/ml ampicillin. After growing the cells at 37°C for 47 h, we mixed the cell suspension from each well with the Bug-Buster TM , which is formulated with non-ionic detergents that are capable of cell wall perforation without denaturing soluble proteins. To identify optimal conditions for protein extraction, we varied the concentration of the reagent. After the addition of the BugBuster TM reagent, the cell suspension sat for 30 min at room temperature. After a brief centrifugation (3000 ϫ g, 10 min), which was sufficient to remove precipitates and to obtain a non-viscous supernatant, activity was measured as described below. As the result, the activity plateaued when 0.2 ϫ Bug-Buster TM was added to the culture (i.e. a 160-l aliquot of cell suspension from each well with 40 l of the BugBuster TM ).
Activity was determined by the DNS method (12). Each cell extract (20 l) was mixed in 100 l of reaction mixture containing 0.1 M Mes-NaOH (pH 5.7) and 0.1% (w/v) soluble xylan. After incubation at 37°C for 30 min, 100 l of stop solution containing 1% (w/v) DNS, 30% (w/v) potassium sodium tartrate tetrahydrate, and 0.4 N NaOH was added. The mixture was heated at 99°C for 10 min, and activity was determined from the increase of absorbance at 540 nm on a Molecular Devices (Sunnyvale, CA) plate reader (VersaMax). For the thermostability assay, enzymes were heat-treated, and the residual activity was determined. The thermostability index was defined as the ratio of activities before and after the heat treatment.
Enzyme Purification-E. coli JM109 cells harboring wild-type or mutant xylanase genes were grown overnight in LB medium (1.6% Tryptone, 1.0% yeast extract, and 0.5% NaCl) containing 100 g/ml ampicillin at 37°C. The culture was then diluted 1/1000 in a fresh LB medium (1 liter) containing 100 g/ml ampicillin and 1 mM isopropyl-␤-d-thio-galactopyranoside and agitated at 37°C for 12 h. Cells were harvested by centrifugation (5000 ϫ g, 10 min, 4°C) and suspended in 25 ml of BugBuster TM containing 25 l of benzonase nuclease. Proteins were extracted by agitation at room temperature for 20 min. Cell debris was removed by centrifugation (20,000 ϫ g, 20 min, 4°C), and supernatant was applied to a Resource S (Amersham Biosciences) column (6.4 mm inner diameter ϫ 3 cm) pre-equilibrated with 20 mM sodium acetate buffer (pH 5.5). The column was rinsed with 20 ml of the buffer, and proteins were eluted with a linear gradient of NaCl from 0 to 0.2 M. Active fractions were pooled and dialyzed against 20 mM Tris-HCl buffer (pH 7.0). Then proteins were applied to a TSK gel CM-5PW (Tosoh, Tokyo, Japan) column (7.5 mm inner diameter ϫ 7.5 cm) preequilibrated with 20 mM Tris-HCl buffer (pH 7.0). The column was rinsed with 20 ml of the buffer, and proteins were eluted with a linear gradient of NaCl from 0 to 0.4 M. Active fractions were pooled and dialyzed against 20 mM Tris-HCl buffer (pH 7.0). The concentration of purified protein was determined using the extinction coefficient (⑀ 280 ϭ 4.08) as reported previously (17).
Thermal Inactivation-Half-lives of thermal inactivation of wildtype and Xyl st enzymes were determined by incubating the enzymes (0.1 mg/ml) at 60 or 65°C in 20 mM sodium acetate buffer (pH 5.5) containing 0.2 M NaCl and 1 mM DTT. At various time intervals, 10-l aliquots were removed and diluted into 90 l of assay solution containing 0.1 M Mes-NaOH (pH 5.7) and 0.1% (w/v) soluble xylan for the measurement of residual activity at 37°C. Inactivation profiles of wild-type and Xyl st enzymes were plotted by incubating the enzymes (0.1 mg/ml) at various temperatures (from 30 to 85°C) for 10 min followed by determining the residual activity at 37°C using 10 l of enzymes in 90 l of assay solution containing 0.1 M Mes-NaOH (pH 5.7) and 0.1% (w/v) soluble xylan.
X-ray Analysis-The initial screening of the crystallization conditions was carried out according to the sparse-matrix sampling method

TABLE 1 Oligonucleotide sequences used for site-saturation mutagenesis
The target amino acid position was coded by degenerated codon NNK (sense strand, upper sequence) and MNN (antisense strand, lower sequence), where N ϭ A, G, C, T, or K ϭ G or T, and M ϭ A or C.

Target sites
Oligonucleotide sequences The initial conditions under which crystals appeared were refined by varying the pH of the buffer and the concentration of the precipitant. Diffraction data of the wild-type xylanase crystal were collected at beamline BL6A and the Xyl st crystal were collected on NW12 of the Photon Factory, KEK, Tsukuba, Japan at a wavelength of 1.0000 Å. Before data collection, the crystal was soaked in the crystallization solution containing 20% glycerol then mounted on a nylon fiber loop and immersed in liquid nitrogen for freezing. The diffraction data were processed using HKL2000 (20) and the CCP4 program suite (21).
Structures of wild-type xylanase and Xyl st were determined by a molecular replacement method with the program MOLREP (22) in the CCP4 package using the structure of Bacillus circulans xylanase (PDB codes 1XNB) as a search model. CNS (23) and REFMAC5 (24) were used for structural refinement. A randomly chosen 5.0% of the data were used to calculate the free R factor (25).

RESULTS AND DISCUSSION
High-throughput Screening System for Xylanase Thermostability-Because Miller (12) reported the use of DNS reagent to determine the concentration of reducing sugars released, the DNS method has routinely been used in determination of glycosidase activities. In this study we modified the DNS method by downscaling all the reaction components to fit the 96-well format. In addition, for consistent extraction of cytosolic proteins from E. coli with high efficiency and to minimize well-to-well or plate-to-plate fluctuations, we used the BugBuster TM protein-extracting reagent. Although the reagent is usually used in a 1ϫ concentration, we identified that a 0.2ϫ concentration was the best. Increasing the concentration of the reagent had an adverse effect; the extraction efficiency of the enzyme did not increase, whereas the solution became more viscous. Because pipetting viscous solution is highly problematic, it was at least in our case not advantageous to use the regular 1ϫ concentration.
To test the validity of our screening system, we performed site-saturation mutagenesis at amino acid position 169. The residue is fully buried in the molecule, and the "solutions" that confer thermostability may be limited; thus, some hydrophobic residues will be accommodated. We previously found that the thermal stability of the enzyme is greatly reduced by altering the wild-type residue Met to Ile. 3 We, therefore, used the thermolabile M169I variant and employed saturation mutagenesis to test if our screening was sensitive enough to solve the "problem" correctly.
Eight clones were randomly picked from the saturation mutagenesis library to check for the randomness of the library. DNA sequencing revealed that the site was occupied with 2ϫ Asp (GAT), 1ϫ Phe (TTT), Trp (TGG), Lys (AAG), Ala (GCT), Val (GTG), and Ile (ATT), ensuring the randomness of the library. We then screened for activity and thermostability of 360 randomly picked clones. Approximately half of the clones retained the activity (Fig. 1a). We next screened 164 active clones for thermostability by incubation at 55°C for 5 min (Fig. 1b). Fourteen thermostable clones that showed essentially the same stability were picked and subjected to DNA sequencing. As a result, the position was occupied with Met (5ϫ ATG) or Leu (2ϫ TTG, 2ϫ CTG, and 5ϫ CTT). Taking the different types of codons in Leu-169 variants into account, our screening system is accurate and sensitive enough to identify thermostable variants of xylanase.
Directed Evolution-The thermolabile M169I variant was used as a starting enzyme to enhance thermostability by directed evolution. To avoid sacrificing activity at low temperature, which is often seen in naturally thermophilic enzymes (26), simultaneous screening for activity at 25°C and thermostability at elevated temperatures was performed.
At first, error-prone PCR was carried out for the entire sequence of the xylanase gene. A total of 744 clones were screened for thermostability by incubation at 55°C for 5 min. Five thermostable variants, 2D6, 4C3, 4A6, 4D12, and 6B5, were identified, and their DNA sequences were determined (Table 2). Interestingly, some mutations were localized at three sites. Mutations at 169 were identified in 4A6 (I169V), 4D12 (I169L), and 6B5 (I169M), mutations at residues 7 and 8 were identified in 4C3 (Q7H) and 4A6 (N8Y), and mutations at residues 151/152 were identified in 4A6 (A152T) and 6B5 (N151S). To identify true-positive amino acid substitutions, we then conducted recombination experiments of those variants by DNA shuffling (15).
Two StEP libraries were created to recombine variants 2D6/4C3 and 4A6/4D12/6B5. As for the 2D6/4C3 StEP library, 168 clones were screened for thermostability by incubation at 65°C for 5 min. Two thermostable variants that surpassed the thermostability of parents  Relative activities and thermostabilities of xylanase variants from a library prepared by site-saturation mutagenesis at amino acid 169. a, plotted in descending order. b, activity versus thermostability. Activity was measured in 100 l of assay solution containing 0.1 M Mes-NaOH (pH 5.7) and 0.1% (w/v) soluble xylan by using 20 l of cell extract at 37°C. For thermostability assay, enzymes were heat-treated (55°C, 5 min) before the measurement of residual activity in the same assay solution. Thermostability was defined as the ratio of activities after and before the heat treatment. Values were relative to that for the wild-type enzyme, respectively. Dotted lines indicate the mean values for activity and stability of eight wild-type clones. S.E. were Ͻ4%.
were identified. DNA sequencing showed that both variants contained the same mutations, Q7H (21A3 C) and S179C (535A3 T). As for the 4A6/4D12/6B5 StEP library, 192 clones were screened for thermostability by incubation at 60°C for 5 min. Two thermostable variants that surpassed the thermostability of parents were identified. DNA sequencing revealed that one variant contained N8Y (22A3 T), N151S (452A3 G), and I169M (507A3 G), and the other contained N8Y (22A3 T) and I169M (507A3 G). Because the thermostability was nearly the same for the two variants, the latter variant was used for further study. Although the sites 151 and 152 appeared to be a weak point of the molecule, they did not contribute to enhancing stability in the context of other mutations, N8Y and I169M. The mutations might be false positive mutations, and DNA shuffling did not pick such nonadaptive mutations. There were three amino acid substitutions both in 6B5 and 4A6, and the rest of the mutations (N114K and I169M in 6B5; N8Y and I169V in 4A6) might play a more important role in thermostabilization.
Next, we employed site-saturation mutagenesis (27) to optimize the local sequence at positions 7, 8, and 179. Positions 7 and 8 were regarded as a single site, and we performed two mutagenesis experiments. For positions 7/8, a total of 744 clones were screened for thermostability by incubation at 55°C for 5 min. Two thermostable variants were identified that contained mutations Q7H (CAT)/N8F (TTT) and Q7L (TTG)/N8V (GTT), respectively. The former variant showed slightly higher stability. As for position 179, 192 clones were screened, and 9 thermostable variants were identified. Of these, seven variants contained the same mutations S179C (all in the TGT codon). The rest of the variants, S179Y (TAT) and S179M (ATG), showed a slightly lower stability.
Individually optimized sequences (Q7H, N8F, and S179C) were then recombined by DNA shuffling (16). As a result of the screening of 192 clones for thermostability by incubation at 65°C for 5 min, 3 thermostable variants were obtained, all of which contained amino acid substitutions Q7H, N8F, and S179C (one variant contained an additional synonymous mutation, t66a). Based on the triple amino acid substitution variant, site-directed mutagenesis was performed to convert Ile-169 to Met, to obtain a variant Xyl st .
Thermostability of Purified Enzymes-To determine thermostability, we purified the wild-type and Xyl st enzymes. Because the Xyl st contained a single Cys residue at a surface of the molecule, we checked the possibility of dimer formation through intramolecular disulfide bridges. Proteins (wild type and Xyl st ) were incubated in the presence and absence of DTT (1 mM) and subjected to SDS-polyacrylamide gel electrophoresis. As shown in Fig. 2, Xyl st formed in part a dimer in the absence of DTT, indicating that the enzyme is prone to forming a dimer upon prolonged incubation. To simplify the inactivation experiments, we added DTT thereafter to both Xyl st and wild type.
First, the enzymes were heated at various temperatures for 10 min, and residual activities were determined. As shown in Fig. 3, half-inactivation temperatures (the midpoint of the melting curves) shifted ϳ10°C upward for Xyl st (wild type, 58°C; Xyl st , 68°C). At 60°C, the wild-type enzyme was inactivated, whereas Xyl st retained full activity. At 65°C, the wild-type enzyme completely lost activity, but Xyl st retained activity. A slight increase in activity was observed for the Xyl st enzyme especially at high temperature around 60°C. This might be caused by heat activation of the enzyme via structural rearrangement as reported previously (28).
Next, half-lives of the enzymes upon heating were determined. As observed in the previous experiment (Fig. 3), the stability of the two enzymes was widely different, and thus, the half-life of the enzymes was not determined under the same temperature. At 60°C, the wild-type enzyme was inactivated within 10 min (half-life, ϳ5 min), whereas Xyl st was fully active for at least 2 h (Fig. 4A). At 65°C, the wild-type enzyme was inactivated rapidly, and the half-life could not be determined (Fig.  4B). The half-life of Xyl st was ϳ20 min at 65°C.
Temperature Dependence of the Activity-The specific activities of wild-type and Xyl st were determined over a range of temperatures from 20 to 85°C. As shown in Fig. 5, stabilization of Xyl st was accompanied by an increase in the optimal temperature of ϳ10°C. The activity of Xyl st was slightly lower than that of wild type below 55°C but higher at elevated temperatures, indicating that the Xyl st acquired a thermophilic nature after evolution. In our previous studies of temperature adapta-   tion of psychrophilic subtilisin S41 to a high temperature (27,29), we obtained a variant whose activity shifted upward over a wide range of temperature.
Crystal Structure Analysis and Possible Explanation for Thermostability-To elucidate the effect of mutation on thermal stability and activity by comparing wild-type xylanase and Xyl st on the basis of structural information, we determined their crystal structures. Crystals of wild-type xylanase were grown from 0.1 M imidazole buffer (pH 7.7-7.9), 1.0 -1.1 M potassium/sodium tartrate. Xyl st was crystallized in a different form under conditions of 0.1 M Mes buffer (pH 5.9 -6.1), 1.1-1.2 M ammonium sulfate, 10% dioxane, and 25 mM DTT. In the wild-type crystal, five molecules exist independently in the asymmetric unit of the cell, and two molecules are contained in the asymmetric unit of the Xyl st crystal. Their structures were determined by the molecular replacement method and refined against the 20 -1.4 Å intensity data for wild-type xylanase and 20 -1.9 Å for Xyl st . The crystallographic R-factor and free R-factor of wild-type xylanase were 0.197 and 0.217, respectively. The crystallographic R-factor and free R-factor of Xyl st were 0.244 and 0.269, respectively. Table 3 summarizes the crystallographic data, statistics of x-ray data collections, and the result of the structural refinement of wild-type xylanase and Xyl st .
The root mean square deviations of all atoms between pairs within five molecules of wild-type xylanase crystal existing in the asymmetric unit range from 0.17 to 0.80 Å, showing no large perturbations incurred by packing in the crystal. Exceptional large deviations (ϳ3.0 Å) occur only on turn regions around residue 120, which is located at the edge of the central groove, protruding from the molecule and exhibiting poor electron density relative to the rest of molecule. Each of the five molecules of B. subtilis xylanase determined in this study can be superposed onto B. circulans xylanase, with the root mean square displacements ranging from 0.42 to 0.73 Å, indicating that their overall three-dimensional structures are practically identical.
Two independent Xyl st molecules exist in the asymmetric unit to form an intermolecular dimer by disulfide bond at the mutated residue, Cys-179. They are related by a non-crystallographic two-hold axis as shown in Fig. 6. Only a limited number of surface residues are involved in the interaction between each monomer, suggesting that the threedimensional structure of Xyl st is not affected by intermolecular dimerization in the crystal. The root mean square deviations between wildtype xylanase and Xyl st range from 0.38 to 0.8 Å. Furthermore, structural differences at three mutated positions (His-7, Phe-8, and Cys-179) are about the same as those of other positions except for around amino acid position 120, significantly showing that amino acid substitution introduced to Xyl st induces no large changes to its three-dimensional structure.     6 shows the location of the mutations. In crystals, the enzyme formed a dimer connecting via the newly introduced Cys-179. Because the wild-type enzyme did not form the dimer in crystal, the mutation must have promoted intermolecular disulfide-bond formation under a crystallization condition. The Cys-179 is surrounded by hydrophobic Val-57 and Ala-59 (Fig. 7). In our site-saturation mutagenesis experiment at residue 179, we identified not only the Cys variant but also the Tyr variant as the second most stable enzymes. Taking the hydrophobic environment around the residue and hydrophobic nature of Cys residue into account (30), the enzyme may have been stabilized by increasing hydrophobic contact between monomers. In fact, the addition of a reducing agent did not greatly alter the stability. Fig. 8 is a close view of the mutation site at amino acid positions 7 and 8. The cavity seems larger for the wild type (Fig. 8A) than for Xyl st (Fig. 8B). Packing the cavity with hydrophobic residues should have contributed to enhancing stability. It is also reported that hydrophobic interaction at surface of a protein can stabilize the molecule (31,32). This finding is in agreement with the result in which the Leu-7/Val-8 variant was identified in the saturation mutagenesis library.
Amino Acid Sequence Comparison to Natural and Engineered Thermostable Xylanases-We next made amino acid sequence comparisons with a naturally thermophilic xylanase from Bacillus D3 (33). The Bacillus D3 xylanase shares 73% identity with the B. subtilis xylanase, but the difference in thermostability is distinct. The B. subtilis enzyme is not stable at temperatures above 60°C, whereas Bacillus D3 xylanase is stable even at 75°C. Position 7 in B. subtilis xylanase (Gln-7) is identical to that of the Bacillus D3 enzyme, which was changed to His after evolution. Residue 8 is Asn in the B. subtilis enzyme, whereas the corresponding residue is Tyr in the Bacillus D3 enzyme. Although the position was changed to Phe in Xyl st , we also identified the Tyr variant in the error-prone PCR library ( Table 2). As for position 179, the B. subtilis enzyme has Ser, and the corresponding residue is Tyr in Bacillus D3. In our site-saturation mutagenesis experiment, we identified the Tyr variant as well as the Cys variant. This observation agrees with our structural studies, in which a major contribution to thermostability was concluded to be the result of increased hydrophobicity between monomers.
We next compared these findings with the results of site-directed mutagenesis studies of glycosyl hydrolase family-11 xylanases (3-6). Among the three sites, only the Ser-179 was targeted by site-directed mutagenesis (3). In their study, Wakarchuk et al. (3) altered the Ser of B. circulans xylanase to Cys in an attempt to introduce an intermolecular disulfide bridge. The variant, in fact, was prone to forming a dimer at 77% efficiency. Similar to their result, the S179C variant of B. subtilis xylanase also formed a dimer in solution (Fig. 2) and in crystals (Fig. 8). However, the difference in thermostability with or without the reducing agent was nominal. In addition, screening of the site-saturation mutagenesis library identified not only the Cys variant but also the Tyr and Met variants as having near-equivalent stability. Therefore, disulfide bridges may not account for the thermostability, but the hydrophobic nature of the side chain may be critical.
In this study we used directed evolution and evolved the enzyme stability without using the three-dimensional structure. Error-prone PCR quickly identified residues 7/8 as a weak point, and subsequent saturation mutagenesis further improved the stability. By sharp contrast, no studies based on a targeted design have identified the region as a weak point of the molecule even with high resolution crystal structure analysis. Directed evolution is, thus, advantageously used in protein engineering even when the high resolution crystal structure is available. In the crystal structure, the enzymes formed a dimer, and Cys-179 was located at the dimer interface. The figures were produced using MOLSCRIPT (34) and RASTER3D (35).