Directed evolution of beta -glucosidase A from Paenibacillus polymyxa to thermal resistance.

The beta-glucosidase encoded by the bglA gene from Paenibacillus polymyxa has a half-life time of 15 min at 35 degrees C and no detectable activity at 55 degrees C. We have isolated random mutations that enhance the thermoresistance of the enzyme. Following a directed evolution strategy, we have combined some of the isolated mutations to obtain a beta-glucosidase with a half-life of 12 min at 65 degrees C, in the range of resistance of thermophilic enzymes. No significant alteration of the kinetic parameters of the enzyme was observed. One of the mutants isolated in the screening for thermoresistant beta-glucosidase had the same resistance to denaturation as the wild type. This mutation caused the accumulation of enzyme in E. coli, probably due to its lower turnover. The structural changes responsible for the properties of the mutant enzymes have been analyzed. The putative causes increasing thermoresistance are as follows: the formation of an extra salt bridge, the replacement of an Asn residue exposed to the solvent, stabilization of the hydrophobic core, and stabilization of the quaternary structure of the protein.

The ␤-glucosidase encoded by the bglA gene from Paenibacillus polymyxa has a half-life time of 15 min at 35°C and no detectable activity at 55°C. We have isolated random mutations that enhance the thermoresistance of the enzyme. Following a directed evolution strategy, we have combined some of the isolated mutations to obtain a ␤-glucosidase with a half-life of 12 min at 65°C, in the range of resistance of thermophilic enzymes. No significant alteration of the kinetic parameters of the enzyme was observed. One of the mutants isolated in the screening for thermoresistant ␤-glucosidase had the same resistance to denaturation as the wild type. This mutation caused the accumulation of enzyme in E. coli, probably due to its lower turnover. The structural changes responsible for the properties of the mutant enzymes have been analyzed. The putative causes increasing thermoresistance are as follows: the formation of an extra salt bridge, the replacement of an Asn residue exposed to the solvent, stabilization of the hydrophobic core, and stabilization of the quaternary structure of the protein.
A key factor in the rapid growth of protein engineering is the increasing industrial demand for enzymes selected to work under defined conditions, both for traditional and novel applications. Despite important advances in protein chemistry, current understanding of the structure-function relationship is still very limited. This precludes in most cases knowledgebased modification of the protein structure to achieve a desired function. An alternative approach, directed evolution, has been proposed as a strategy for protein engineering. Directed evolution follows the traditional scheme of alternating steps of mutation and selection used for the development of industrial strains with improved properties, but in this case the subject of the mutagenesis is a defined gene or set of genes, and the selection is applied to the function(s) encoded. This method has some requirements as follows: the gene encoding the protein to be modified has to be functionally expressed in a suitable microbial host, and there must be an effective screening procedure for the desired activity or function (1).
The properties of the ␤-glucosidase encoded by the bglA gene of Paenibacillus (formerly Bacillus) polymyxa are very well suited to the requirements of directed evolution. This enzyme (henceforth BglA) can be assayed for Escherichia coli colonies grown on Petri dishes, with different chromogenic substrates. Enzymes with novel properties can be screened for using conditions which differentiate the wild-type enzyme from a mutant with a given property (i.e. resistance to thermal denaturation) (2). The three-dimensional structure of BglA is known (3), which helps investigation of the structural basis of the newly isolated mutations.
Protein thermal resistance is an important issue beyond its obvious industrial interest, because the understanding of the basic physicochemical principles necessarily aids the comprehension of other aspects of the structure-function relationship. Recent reports based on experimental data and theoretical considerations point to increased number of hydrogen bonds and salt bridges as the two main factors contributing to thermal stability (4 -10). Nevertheless, it seems clear that there is no unique paradigm of thermostable structure. Other factors are also important such as increased internal hydrophobicity, minimization of the surface/volume ratio, fewer internal cavities, stabilization of the secondary structure elements, and minimization of loops (6,11,12). This paper describes the continuation of previous work in which we isolated and characterized thermoresistant forms of BglA, obtained after random chemical mutagenesis (2,13). We now report the use of an alternative, PCR 1 -based random mutagenesis procedure to obtain new mutants. These novel mutant forms of BglA have been combined, by a directed evolution strategy, to produce a highly stable thermoresistant ␤-glucosidase.

EXPERIMENTAL PROCEDURES
Basic Procedures-Conditions for maintenance and expression of the bglA gene from P. polymyxa in E. coli have been described previously (2). The cloning vector used was ⌬pUC18. This plasmid was obtained from pUC18 (14) by deletion of a 575-base pair fragment corresponding to the lacZ gene. The use of the modified vector was convenient to avoid the detection of artifactual ␤-glycosidase activity when 5-bromo-4chloro-3-indolyl-␤-D-glucoside (X-Gal) was used as a substrate.
Mutagenesis and Screening for Thermoresistant ␤-Glucosidase Activity-Random mutagenesis of the bglA gene was carried out by errorprone PCR amplification (15,16). The template DNA for the mutagenic amplification was plasmid pLGBGA (2), which contains the bglA gene cloned in the pUC18 vector. Direct and reverse pUC universal primers were used for the amplification. Four parallel reactions were carried out for each mutagenic experiment. Each one of these four reactions contained an excess of a particular dNTP (0.5 mM) respective to the other dNTPs (0.2 mM). The reaction mixtures were subjected to 25 cycles of amplification with Biotaq DNA polymerase (Stratagene). The amplified material was digested with EcoRI and HindIII and ligated into the ⌬pUC18 vector. The ligation mixtures were used to transform E. coli DH5␣ selecting for ampicillin-resistant colonies. The transformant colonies were screened for thermoresistant ␤-glucosidase activity (17). * This work was supported by Comisión Interministerial de Ciencia y Technologia Grants ALI97-0362 and BIO97-0642. 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.
Structural Analysis-The structural models of the different mutations were constructed on the basis of the reported crystal structure of native ␤-glucosidase A (3), with Protein Data Base code 1BGA. Amino acid sequence changes were incorporated using the graphics program O (19) running on a Silicon Graphics workstation. Side chain conformation was chosen from the data base of more common conformers (20) and adjusted manually.

Isolation of Random Mutations That Increase Thermal Resistance of BglA-
The open reading frame of the bglA gene of P. polymyxa was amplified by PCR under conditions devised to introduce about 3-10 nucleotide changes per kilobase. The amplified material was ligated into ⌬pUC18 and used to transform E. coli DH5␣ selecting for Amp R colonies. The transformants were screened for ␤-glucosidase activity, after incubation at a temperature that caused inactivation of the wild-type enzyme. Plasmid was extracted from positive clones (presumptively expressing thermoresistant BglA), and the mutations responsible for the observed phenotype were mapped within suitable restriction fragments and identified by DNA sequencing. Table I lists the mutations thus characterized and the corresponding amino acid substitutions. The repeated recovery of the same mutations in independent mutagenesis experiments indicated that we had obtained all thermoresistant mutations that could be recovered by this procedure. Two of the mutations GAG 3 AAG (E96K) and ATG 3 ATA (M416I) (where boldface letters indicate the mutated base) had been previously isolated by chemical mutagenesis (2). The other four were characterized for the first time. The PCR mutagenesis allowed the recovery of two transversions, AAT 3 AAA (N437K) and AAT 3 CAT (N437H). This type of mutation is very rarely produced by conventional procedures of chemical mutagenesis. Since the three-dimensional structure of BglA is available (3), it was possible to map the mutations to their specific spatial location and to the element of secondary structure to which they are associated (Table I). Fig. 1 shows plots of thermal inactivation at 50°C for wildtype BglA and the six isolated mutants. One of these (T385A) did not show any increase of thermal resistance compared with the wild-type enzyme. The outstanding property of this mutant, which explained why it was isolated in the screening for thermoresistant enzymes, was a large increase of ␤-glucosidase content in the E. coli host.
BglA belongs to the well characterized family 1 glycosyl hydrolases (21,22). This family includes enzymes from thermophylic microorganisms. The three-dimensional structure of BglA and other enzymes of the family are known, including one from the thermophylic archaea Sulfolobus solfataricus (3,24). As an alternative approach to random mutagenesis, we tried to improve the thermoresistance of BglA by rational design. By using an alignment of sequences, we identified amino acids present in BglA and homologous enzymes from mesophilic species but absent in homologous sequences from thermophyles. Among these, we planned substitutions that could have a beneficial effect on thermoresistance. For instance, oxidation of Cys and Met or deamination of Asn and Gln cause enzyme denaturation, and the replacement of these residues can give rise to a more resistant enzyme. Based on these considerations we tested the effect of several amino acid substitutions: N52D, C117W, Q145E, Q256K, M290F, R331E, and M375V. None of them had any appreciable effect on the thermoresistance of the enzyme.
Thermoresistance and Physiological Stability-Crude extracts of E. coli DH5␣ cultures expressing the P. polymyxa bglA gene with the T385A mutation showed about 20 times more specific ␤-glucosidase activity than extracts of the same strain and plasmid, with the wild-type gene. Kinetic parameters of both mutant and wild-type enzymes were the same (Table II). The difference between the two cultures is illustrated in Fig. 2A that shows a Western blot of protein from crude extracts of parallel E. coli cultures expressing either one of the two bglA alleles. The presence of the mutant enzyme in a much higher amount explains why it was isolated in a hunt for thermoresistant enzymes. The T385A substitution was originated by an ACA 3 GCA replacement. The first codon is relatively infrequent in E. coli (RSCU ϭ 0.10), whereas the second one is much more common (RSCU ϭ 1.09) (25). Therefore, a possible cause for the increase of enzyme content in the mutant would be an improvement of the translation efficiency. To test this, the ACA wild-type codon was changed to the synonymous ACC codon (RSCU ϭ 1.91) by site-directed mutagenesis. This mutation caused a 5-10 times increase of the specific activity that was associated with a corresponding increase of the enzyme content in cell extracts (Fig. 2B). We also tested the change of the wild-type ACA codon to the infrequent GCC codon (RCSU ϭ 0.18). This double mutation also leads to the T385A substitution. Unexpectedly, specific activity and the amount of enzyme in cell extracts were similar to those observed for the ACA to ACC replacement (Fig. 2C). The increased expression of the protein in this mutant cannot therefore be attributed to muta-  1. Thermal inactivation of BglA and mutants. Time course of irreversible thermal inactivation at 50°C for wild-type BglA (q) and mutants as follows: T385A (‚), N411S Ⅺ, M416I (ƒ), E96K (E), N437H (f), and N437K (OE). Enzyme aliquots were diluted in prewarmed assay buffer. Samples were taken at different times and assayed for activity (2). tion to a more abundant codon.
Looking for clues to explain the structural basis of the effect associated to the T385A substitution, we have analyzed the environment of this position. The -R group of Thr 385 is oriented toward a cavity situated in the internal space of the protein (Fig. 3). A greater stability associated with the T385A replacement would agree with the more hydrophobic nature and the higher volume of the alanine side chain. It is known that refilling of internal cavities with hydrophobic groups protein stability (26). To test this idea, we carried out alternative amino acid replacements at positions: T385L, T385I, and T385F. All three replacements led to a similar increase of specific activity (5-10 times over that of the wild type), which was associated with corresponding increase of intracellular enzyme content (Fig. 2).
Combined Action of the Single Mutations-The different amino acid substitutions that had been identified as responsible for the increased thermoresistance of BglA were associated in different combinations, and the effect in the resulting enzymes was determined. Results presented in Table III     M416I/N437K/T385A combination showed maximal thermal resistance, like the E96K/M416I/N437K triple mutant (Table  III), as well as a high intracellular enzyme level, comparable with that of the T385A single mutant (Fig. 2A).
Structural Analysis of the Mutations-The crystal structure of BglA is known (3). The monomer displays the (␣/␤) 8 barrel topology characteristic of family 1 glycosyl hydrolases, with the active center placed in a wide cavity defined along the axis of the barrel. The native enzyme has an octameric quaternary structure, which can be described as a tetramer of dimers with 4-fold symmetry. The two active sites of each dimer face each other separated by the equator of the structure. Fig. 5A shows the secondary structure elements of the subunits and five of the isolated mutations with respect to the overall folding. The structural basis of the effect caused by the E96K mutation (Fig.  5B) has been characterized (13). Glu 96 is located at the N terminus of helix 2 of the barrel, and its replacement by Lys allows the formation of an ion pair with Asp 28 , situated in one of the loops surrounding the active center. The ion pair links segments that are 68 amino acids apart in the sequence.
Asn 411 is part of a loop on the surface of the monomer, near the active center cavity. Its side chain is only hydrogen-bonded to water molecules and is fully accessible to the solvent. The analysis of the amino acid composition of thermophilic proteins reveals a tendency toward a low content of certain "thermolabile" residues such as Asn, Gln, Cys, or Met (11,12,27). The substitution of Asn by Ser, also a polar residue, does not disturb the chemical environment of the loop and avoids the vulnerability of the Asn side chain.
The Met 416 residue is in a loop, totally buried in a hydrophobic cavity (Fig. 5C), and makes contacts to Leu 71 , Met 357 , and the aromatic rings of Tyr 432 and Tyr 435 . Its substitution by Ile should favor thermostability because of the improved packing efficiency conferred by the isoleucine side chain. Interestingly, Met 416 is a unique feature of BglA, since other family 1 enzymes have Ile, Val, or Phe at this position.
Single mutations of the Asn 437 residue are those with a greatest effect in thermoresistance (Table III). Asn 437 is located at the end of the last helix before the C terminus of the chain (Fig. 5, A and D). This helix is close to the loop His 40 -Asn 50 of a neighboring subunit, but the side chain of Asn 437 is not hydrogen-bound to this loop. The replacement of asparagine by histidine in the interfacial surface allows the formation of an intermolecular ion pair to the Asp 49 side chain. This might help in preventing protein inactivation by subunit dissociation, as it often occurs in oligomeric proteins. Furthermore, the longer side chain of lysine enables a bifurcated salt bridge from the terminal amine moiety of lysine to both Asp 49 and Asn 47 of the other subunit, thus introducing a stronger intermolecular interaction between subunits. This is consistent with the improved thermoresistance of the N437K as compared with N437H mutant. DISCUSSION Advances in protein chemistry and the resolution of an elevated number of three-dimensional structures of thermoresistant enzymes have revealed determinant factors of this property. It should be possible, at least in theory, to increase the thermal resistance of an enzyme by a planned modification of its primary structure. Nevertheless, this is a difficult approach that has only been successfully accomplished in a few cases (28 -30). In this work we have used an alternative way, the isolation and characterization of randomly generated mutations and their combination, to achieve a substantial increment of the thermoresistance of the enzyme. This requires an effective method to generate a wide range of random mutations and a procedure for their identification. The error-prone PCR mutagenesis proved to be useful, allowing us to expand the range of mutations of the bglA gene previously obtained by chemical mutagenesis (2).
One of the mutations isolated in the screening for thermoresistant enzyme variants conferred unexpected properties. Thermal resistance and kinetic parameters of the enzyme with the T385A replacement were the same as those of the wild type. The only significant difference was that the enzyme was present in the cell in a much higher amount. Our results indicate that the main cause of this effect is the amino acid replacement and not other factors such as codon change. The high content of the T385A enzyme in the cell led to a substantial level of residual activity, after the incubation at high temperature used in the screening for thermoresistant activity, and explains why this mutant was isolated. Proteins have a certain degree of instability as a physiological requirement to allow turnover. An accumulation of enzyme will result from any mutation that would displace the conformational equilibrium toward a state less prone to be degraded by the proteolytic machinery. The properties of the T385A mutation suggest that it increase the "physiological stability" of the enzyme.
Not all the structural modifications of an enzyme that may improve a desired characteristic (i.e. higher temperature resistance) are expected to suit physiological conditions. This fact is illustrated by some of the mutants studied here. The combination of E96K, M416I, and N437K led to a highly thermoresistant enzyme, but it was produced by E. coli in much lower amount than the wild type. This effect can be envisaged as the opposite of that caused by the T385A substitution.
Regardless of the astounding impression caused by the existence of living organisms thriving at the temperature of boiling water, the structural difference between homologous proteins that have optimal performances at very different temperatures may be rather trivial. Several reports in the literature have shown that the change of a few amino acids was sufficient to achieve important changes in thermostability (28,30,31). In other instances, the comparative analysis of threedimensional structure of highly thermoresistant proteins and their mesophilic homologs led to the identification of a few key residues responsible for the difference (8,(32)(33)(34)(35). The conformation of proteins is maintained by a large number of weak interactions. The net free energy of stabilization of a protein (that is the difference between the stabilizing and destabilizing forces) is relatively small. The additional stabilizing energy required to achieve the properties of highly stable proteins is the same order, equivalent to only a few weak interactions (27). Therefore, it is conceivable that the combination of just a few mutations may lead to the recovery of a thermoresistant enzyme. Our results agree with these observations. We have isolated a few thermoresistant mutations and found the structural basis of their effect. The combination of only four of these mutations caused a remarkable increase of the thermal resistance of BglA: from no detectable activity at 55°C to a half-life time of 12 min at 65°C. This suggests a cooperative rather than additive effect of the mutations, although a precise definition of their interactions will require spectroscopic or calorimetric assessment of the protein stability.